Notes

“Parent communication is the part of the job that gets no planning time and carries more emotional weight than almost anything else in the week.”

A teacher's timetable accounts for lessons. Their directed time accounts for meetings and duties. Nothing anywhere accounts for the message from a parent that arrives at nine in the evening, is two paragraphs long, contains one sentence that is unfair, and will occupy the teacher's mind until it is answered. That exchange is unbudgeted. It happens in the gaps, and the gaps are the teacher's own evening. The cost is not the typing. Most replies to most parents are short and straightforward. The cost is the small number that are not: the ones where the parent is angry, or frightened for their child, or has misread a situation, or is right and the teacher knows it. Those take fifteen minutes to write and two hours to stop thinking about. And they arrive without warning, at the end of a day that had already used up whatever composure was available. What makes it harder is that teachers are largely unsupported in this specific skill. Most received no training in difficult written communication with adults. There is rarely a shared house style, seldom a colleague reading a draft before it is sent, almost never a norm that says a reply can wait until tomorrow morning. So a teacher who has spent the day being extremely skilled with children is asked, alone and tired, to be extremely skilled with an adult who is upset with them. Two things reliably help, and neither is technology. The first is a delay: almost no parent message needs answering the same evening, and the version written the next morning is nearly always better, shorter, and less defended. The second is structure: acknowledge what they have raised, state plainly what happened, say what you will do, offer a conversation. Most difficult messages are asking to be heard rather than argued with, and a reply built on that structure is both faster to write and more likely to end the exchange rather than extend it. Where a tool can help is narrower than the marketing usually suggests. It cannot know the child, the history, or what the parent is really worried about, and it should not be trusted with the judgement about what to say. What it can do is absorb the drafting cost of the routine ninety per cent — the reply about the missing homework, the note about next week's trip, the same explanation of the same policy written for the eleventh time — so that the teacher's remaining energy is available for the small number of messages that genuinely need them. The test I would apply is a simple audit. Look at the last ten messages you sent to parents and ask how many required you specifically — your knowledge of that child, your judgement, your relationship with that family. Whatever that number is, the rest were a cost the job imposed on your evening for no return. That is the part worth removing, and the remainder is the part worth protecting. — the aime team

“The second time you teach a topic should be far easier than the first, and for most teachers it barely is.”

Consider what a teacher knows at the end of teaching a unit for the first time. They know which explanation landed and which one produced blank faces. They know the two examples that were worth the time and the one that wasted a lesson. They know the misconception that appeared in nearly every book, the point where the pacing collapsed, the homework that came back undone because it was set on the wrong night. That is an enormous amount of hard-won, specific, expensive knowledge. A year later, when the unit comes round again, almost none of it is available. Some of it lives in annotations on a plan, if the teacher was disciplined and if they can find the file. Most of it has simply evaporated, because human memory does not retain the texture of a unit taught eleven months ago across four other classes and two changes of timetable. So the teacher rebuilds, and rediscovers the same misconception, and wastes the same lesson on the same weak example. The second run is a slightly better version of the first, when it should be a substantially better one. This is the least discussed waste in the profession. Schools worry about knowledge lost when a teacher leaves. The larger loss happens annually, inside teachers who stay, as the specific learning from each unit fails to survive the year between deliveries. A profession where practitioners genuinely accumulated their teaching across a career would look very different after fifteen years than it currently does. None of the usual solutions work, because they all require the tired teacher to do more. Reflection sheets get filled in for a fortnight in September. Departmental review meetings capture the general and lose the specific. Writing detailed notes on a unit at the end of a term is exactly the moment nobody has capacity for it. The reason the knowledge is lost is not that teachers do not value it. It is that capturing it costs energy at precisely the point in the year when there is none left. The only version of this that works is one where the capture is a by-product rather than a task. If the system a teacher already plans in also holds what happened — what was taught, what was re-taught, where the pacing slipped, what the class actually demonstrated — then next year's version of the unit can start from last year's, annotated by reality, without anyone having to sit down in July and write it up. That is the whole idea behind the memory aime keeps: not a record for anyone else to inspect, but a teacher's own practice, held so that the next run genuinely starts ahead of the last one. The test I would apply is straightforward. Take a unit you have taught three or four times. Ask what, concretely, is available to you now from the last time you taught it — not a general impression, but the specific misconception, the pacing note, the example that failed. If the honest answer is "whatever I remember", then fifteen years of experience is being stored in a place that cannot hold it, and every year you are paying again for something you already bought. — the aime team

“Nobody has ever been trained to hand part of their job to a machine, and teachers are now expected to be good at it by instinct.”

Every other professional norm in teaching arrived with something behind it. How to mark, how to plan, how to speak to a parent, what to do when a child discloses something — these came with training, with mentoring, with a colleague down the corridor who had done it before. They are not perfect norms, but they exist, and a teacher entering the profession inherits them. Delegation to AI has arrived without any of that. In the space of about two years, teachers have been handed a technology that will write almost anything they ask for, and left to work out entirely alone where the line sits. Is it acceptable to have a first draft of a report comment written for you? A whole lesson? A reply to a difficult parent? Nobody has told them, because nobody has decided, and in the absence of a decision every teacher is quietly making their own — usually at ten at night, usually without discussing it with anyone, and often with a small residue of guilt attached. That guilt is worth examining, because it is not evenly distributed. Teachers feel very little unease about using a textbook someone else wrote, a scheme of work inherited from a predecessor, a worksheet from a colleague, a past paper from an exam board. All of those are delegation. The discomfort attaches specifically to the machine, and mostly because the machine is fast. Speed feels like cheating in a profession that has long used visible effort as a proxy for care. The line that actually matters is not about speed and not about who typed the words. It is about judgement. When a teacher delegates the drafting of something and then reads it, corrects it, and takes responsibility for it, the professional judgement is intact and the work is theirs. When they delegate the drafting and send it unread, the judgement has gone, and it does not matter whether the tool was an AI, a template, or a colleague's old file. The failure mode has always existed. AI just makes it faster to reach. So the question a school or a teacher should be answering is not "how much AI use is acceptable" — an unanswerable question that will change again in six months. It is "which decisions in this job must be made by a human who knows this child, and which do not". That is a question about teaching, and teachers are extremely well equipped to answer it. The test I would apply is this: could you explain to the parent of a child in your class exactly what you delegated and why, without discomfort? If yes, the delegation was sound, whatever tool produced the first draft. If you would rather they didn't ask, that is not a technology problem. It is a signal that a judgement you should have made yourself has quietly been handed over. — the aime team

“"Efficient teaching" is one of the most dangerous phrases in education, and it usually means the opposite of what it says.”

When someone praises an AI tool for making teaching more efficient, it is worth pausing to ask what, exactly, is being measured. In most other industries, efficiency is measured against a well-defined output — units produced, calls handled, tickets closed per hour. The word slid into education through the same door, and it has quietly done a lot of damage on the way in, because the thing teaching actually produces is not efficient by nature and never was. A teacher's output is not the number of worksheets generated, not the number of lessons delivered, not the number of children moved from one grade band to another in a term. Those are activities. The output is a slower, less measurable thing: a child who understands something they did not understand before, in a way that will still be with them in five years. That output is produced by patience, by repetition, by the willingness to notice a specific child on a specific Tuesday and change what you were going to say next. None of that language sounds like efficiency, and it is not supposed to. So when a tool promises to make teaching more efficient, the honest question is which of these two things it is targeting. If it is compressing the administrative overhead around teaching — the second draft of the same email, the fifteenth time you have written the same feedback comment, the reconstruction of a lesson plan you already made once — then "efficient" is exactly the right word, and the tool is doing something valuable. If, on the other hand, it is compressing the teaching itself — trying to shorten the number of examples, shrink the reflection, remove the pauses, produce more lessons per hour of a teacher's time — then the word "efficient" is being used to describe something closer to erosion. The distinction is not academic. The place where a lot of edtech goes wrong is precisely at that boundary: it starts by removing overhead, which is welcome, and then keeps going, and starts to compress the parts of teaching that were never overhead in the first place. Reflection is not overhead. Marking that a teacher reads is not overhead. Time to think about a particular child before a particular lesson is not overhead. It looks like overhead if you are looking at the schedule from the outside; from inside the classroom it is the job. We designed aime to be efficient on the first side of that line and deliberately not on the second. It absorbs the repeating administrative work around a well-taught week — the drafting, the sequencing, the remembering, the standard-issue emails, the routine feedback — so that the finite budget of teacher time and attention can flow towards the parts of teaching that are inefficient by nature and are supposed to be. Not more lessons per hour. More attention per lesson. The test I would put to any tool is not "does it make the teacher faster?" but "does it make them faster at the things they should be fast at, and slower — more attentive, less rushed — at the things they should not?" A tool that only reads the first half of that question is a tool that will slowly turn its users into a version of themselves they did not want to become. A tool that reads both halves is one that could genuinely change what teaching feels like for the rest of a career. — the aime team

“Teachers are judged on the one lesson someone watches, and left alone for the eight hundred nobody sees.”

There is a strange asymmetry at the heart of how the profession is supported. A teacher will teach, at a conservative count, around eight hundred lessons a year. Of those, perhaps two or three will be formally observed. The observation lesson is prepared for over days, weighted in performance reviews, discussed in feedback meetings. The other seven hundred and ninety-seven are taught alone, without a colleague in the room, without a sounding board, without anyone to notice when a particular explanation didn't land or a particular class needs a different opening on Wednesday. This is not a criticism of observation, which does what it does. It is a comment on where teaching's structural support ends. Teachers are lavishly scrutinised at a handful of moments and structurally under-supported the rest of the year. Development budgets go to the visible parts of the job; the unseen bulk is where teachers are most alone, and, not coincidentally, where quiet decline in energy and quality most often begins. Most of what would actually help a teacher across those eight hundred lessons is not another observation, or another training day, or another conference. It is the ordinary daily support a thoughtful head of department would give, if a head of department had time — a hand on the pacing when a class is drifting, a nudge that a topic keeps needing re-teaching and the sequence might be worth changing, a reminder that last month's cohort found this exact explanation confusing, a draft of feedback that already knows what these students have been working on. Small, continuous, quiet interventions across the whole year, not big, occasional, ceremonial ones. This is the gap that a certain kind of AI can genuinely fill, and it is not the gap most edtech aims at. It is not about producing more content or providing more analytics. It is about extending a form of thoughtful colleagueship across the seven hundred and ninety-seven lessons that no colleague was ever going to attend. That is a much less glamorous problem than "AI that teaches", but it is the one that would actually change what a teaching week feels like. We designed aime around this bulk-of-the-work premise rather than around the ceremonial lessons. It works quietly across every lesson a teacher teaches — the routine ones, the awkward ones, the small Wednesday-afternoon ones — carrying the pacing, remembering the class, offering the next-step nudge a wise colleague would if they had the time and were in the room. The one that would be observed still gets observed; everything else stops being taught in silence. The test I would apply is uncomfortable but useful. Look at how a teacher's professional support is distributed across their year. If almost all of it clusters around the handful of visible lessons, and almost none of it reaches the daily bulk of the work, the support architecture is upside-down. The lessons that matter most to the teacher's actual life are the ones that never get a second pair of eyes. That is the problem worth solving, and it is a bigger and quieter problem than the one the sector usually talks about. — the aime team

“A good lesson is much more expensive than the timetable suggests.”

School timetables show a teacher's week as a grid of forty-five or fifty-minute blocks, and there is a widespread assumption — held quietly by administrators, sometimes by parents, occasionally by teachers themselves — that a lesson costs roughly its own length to produce. It does not. Every hour a class spends learning has an hour or two of invisible work behind it, and pretending otherwise is one of the reasons the profession is so tired. Consider what actually goes into a single Tuesday-afternoon lesson worth having. There is the curriculum thinking — where this lesson sits in the arc of the unit, what it builds on, what it prepares. There is the class thinking — who is in the room, who was absent last time, who is still shaky on the prerequisite. There is the resource work — the slides, the worksheet, the exit ticket, the example that will actually land with these students. There is the differentiation — the scaffold for the two who will struggle, the stretch for the one who will finish early. There is the assessment — the low-stakes check, the plan for what to do with what it reveals. There is the marking of last week's work, which is how you know what this week's lesson needs to correct. And there is the small, uncounted work of switching gears from the previous class, the corridor conversation, the printer that jammed. None of this appears on the timetable. All of it costs time. And the teacher pays for it, mostly, in the evenings and at weekends — hours that are structurally invisible to everyone except the family they belong to. The reason this matters is not to solicit sympathy. It is that if you do not account for the real cost of a lesson, you cannot honestly account for what a tool that helps with lessons is worth. When someone says an AI system "saves teachers twenty minutes on planning", they are usually measuring the wrong twenty minutes. The visible planning was never the expensive part; the expensive part was the curriculum thinking, the class thinking, the differentiation, the memory of what was taught last week. A tool that shaves twenty minutes off the visible part while leaving all the invisible work exactly where it was has not helped. It has produced a nicer worksheet and left the teacher's evening where it was. The tools that will actually change teacher workload are the ones that reduce the invisible cost of a lesson. That means holding the curriculum arc so the teacher does not have to reconstruct it each Sunday. It means remembering the class so a lesson takes account of who was absent and what wobbled last week. It means carrying the pacing so the teacher does not have to redo the mental arithmetic every time. When those things are handled, the visible planning gets shorter almost as a side-effect — but the real change is that the hidden hours behind each lesson finally start to shrink. This is what aime is trying to be worth. Not a faster way to write a slide, but a genuine reduction in the invisible cost of a well-taught week. When we talk with teachers about what has actually changed after a term of use, they rarely mention the visible saved minutes. They talk about the Sunday evening that was no longer a project. The test to hold any tool to is not how much of the visible work it removes. It is whether, six weeks in, the invisible work behind a lesson is smaller than it was. That is where the real cost lives, and that is the number that has to move. — the aime team

“The word "personalised" in edtech is almost always pointed at the wrong person.”

For a decade now, "personalised learning" has meant one thing: the software adjusts what the student sees. Different questions for different children, adaptive difficulty, a path through the material that responds to what the learner does. That is a fine thing to build. But it has quietly obscured a more interesting question, which is why we never seem to talk about personalising the tools for the teacher — the one person in the room who is genuinely the same across thirty different classes and thirty different years of experience. A teacher on a Wednesday afternoon in their fifteenth year is not the same professional as a teacher three weeks into their first September. They do not need the same interface, the same defaults, the same amount of scaffolding, or the same tone. And yet almost every piece of software they use behaves as if they were interchangeable — a generic dashboard that assumes nothing about who is using it, no memory of what this particular teacher tends to do, no accommodation for the class in front of them right now. The result is that "personalisation" has become a feature aimed at the child and a burden left on the teacher. The child gets a smoother experience. The teacher gets one more system to configure, one more set of defaults to override every time they open it, one more tool that treats them the same on their fiftieth lesson as on their first. There is a better way to think about this, and it is not more student-personalisation. It is treating the teacher's tool as something that learns the teacher — that notices which subjects they teach, which year groups, which pacing they prefer, which explanations they tend to reach for, which children they've asked about before — and shows up to Monday morning already adjusted to who they are, not asking them to adjust it. That is not a nice-to-have. It is the difference between a tool that costs a teacher energy and a tool that returns some. We built aime around that inversion. The Teaching Memory it keeps is not a profile of children being optimised at; it is a working knowledge of the teacher's own practice, held so their next lesson starts from what they've already established rather than from zero. Personalisation, in the sense that actually helps a school, is a property of the teacher's tool — not of the student's screen. The test I would apply is simple. After a term of use, does the tool know the teacher any better than it did on day one? If the answer is no, "personalised" is just marketing. If the answer is yes, you have a tool that is actually working with the professional, not around them. — the aime team

“You should not have to become a prompt engineer to get help.”

There is a quiet assumption buried in most of the AI tools arriving in schools, and once you see it you cannot unsee it. They all wait to be asked. They present the teacher with a blank box and a blinking cursor, and the quality of what comes back depends entirely on the quality of the request the teacher can find the time and the skill to type. We have built a generation of tools that are powerful in inverse proportion to how busy you are — which, for a teacher, is exactly the wrong way round. Think about what the blank box actually demands. It asks the teacher to stop, to articulate precisely what they need, to know the trick of phrasing that gets a good answer, to iterate when the first attempt disappoints, and to do all of this in the ninety seconds between one class and the next. The tool calls this flexibility. The teacher experiences it as one more thing to learn and one more thing to manage. We have taken the person with the least spare capacity in the building and asked them to become a part-time prompt engineer. The language of "empowerment" hides the cost. Handing someone a blank, infinitely capable box is only empowering if they have the time to wield it. Otherwise it is just a more sophisticated way of leaving the work exactly where it always was: on the teacher. aime was built to refuse that bargain. It does not wait to be asked. It already knows the teacher's timetable and the curriculum they are teaching, so it works ahead of them — preparing next week's lessons, differentiating them, sequencing them — and hands back something finished before the teacher has had to type a single prompt. The default state is not an empty box. It is a draft, already made, waiting to be judged. That difference sounds small and is not. A tool that waits to be prompted puts the first move on the busiest person in the room. A tool that works proactively takes the first move itself, and leaves the teacher to do what they are uniquely good at: looking at what has been prepared and deciding, with their knowledge of these particular children, what to keep and what to change. One design asks the teacher to start from nothing. The other asks them to start from something, and to make it theirs. None of this removes the teacher's control — it relocates their effort. The judgement stays entirely with them; what disappears is the blank page and the burden of having to conjure the request. Because the truth the blank box forgets is that a teacher's scarcest resource is not creativity or capability. It is time and attention. A tool worth having spends those for them, not from them. The test is easy to apply. Open the tool on a Monday morning. Does it greet the teacher with an empty box and wait — or has it already done something useful with what it knows? The first is a tool built for a demo. The second is a tool built for a teacher's actual week. — the aime team

“The marking is never actually finished.”

Ask a teacher what they are doing on a Sunday evening and, more often than they would like, the answer is marking. Not because they undervalue their weekends, but because the pile does not respect them. A set of thirty essays is not thirty minutes of work; it is thirty separate acts of attention, each one asking the teacher to hold a child's thinking in their head, notice where it went astray, and find the words that will move it forward without crushing it. Done well, it is one of the most valuable things a teacher does. Done at eleven at night, for the fourth hour running, it is where good teachers quietly burn out. Here is what makes marking such a cruel tax. The part that matters — the judgement about what this particular child needs to hear next — is genuinely irreplaceable. But it is wrapped in a great deal of work that is not: writing the same comment about topic sentences for the fifteenth time, re-explaining a misconception the whole class shares, formatting feedback, transcribing marks, remembering what you said to this student three weeks ago so today's comment builds on it rather than contradicts it. The teacher's expertise and the teacher's drudgery are tangled together, and most of the evening goes to the drudgery. Most tools that promise to help with marking make one of two mistakes. They either try to replace the judgement — handing back an automated grade the teacher cannot trust and the student cannot learn from — or they add another system to feed, another place to type, another thing to manage. aime takes the other path. It does the part that is repeatable and leaves the part that is human. It drafts feedback grounded in the actual success criteria of the task, flags the misconception the class keeps making, and remembers what this teacher told this student last time so the next comment continues the conversation instead of restarting it. What comes back is not a verdict but a draft — specific, on-curriculum, and entirely the teacher's to sharpen, soften, or rewrite before a word of it reaches a child. The point is not to mark faster for its own sake. It is to give the teacher back the hours that were going to transcription and repetition, so the judgement — the one thing only they can do — gets a rested mind instead of a depleted one. A teacher who is not exhausted by nine o'clock writes better feedback at eight. That is the whole aim: not to take marking away from teachers, but to take away everything about it that was never really teaching. The test I would put to any tool here is simple. After a teacher uses it, is the feedback the student receives more personal, or less? If the machine has flattened thirty children into one generic response, it has failed, however much time it saved. If it has cleared away the repetition so the teacher's own voice reaches each child more clearly, it has done exactly what it should. — the aime team

“Before we ask what AI can teach a child, we should ask what it must never do.”

Most conversations about AI in education start with capability: what can it teach, how fast, how cheaply. I want to start somewhere less flattering and more important. Before capability comes a prior question — the one a parent asks first: what happens to my child in the room with this thing? Putting AI in front of children is not the same as putting it in front of adults, and the difference is not sentimental. Children are more trusting of a confident voice, more prone to forming attachments to a responsive one, less able to spot when they are being handed something inappropriate or simply wrong. And their data — what they struggle with, how they behave, what they type late at night — is among the most sensitive there is. A great deal of edtech treats all of this as an afterthought, to be handled in a privacy policy nobody reads. So safeguarding is not a compliance box to clear on the way to the interesting features. It is an architecture decision, and it has to be made at the beginning, because you cannot bolt child safety onto a system designed to maximise engagement. aime is built on a choice that shapes everything else: the AI works for the teacher, not directly on the child. It does the planning, the differentiation, the preparation — and hands the result to a professional adult, who decides what reaches the classroom. There is no open chatbot for a nine-year-old to confide in, no engagement loop tuned to keep a child talking. The human who knows the child stays between the child and the machine. That is not a limitation we apologise for. It is the point. And because the AI works through the teacher rather than talking directly to the child, aime simply never generates the stream of intimate, real-time data a child-facing chatbot lives on — the late-night confidences, the behavioural exhaust, the engagement signals. There is far less sensitive data in the first place, because the architecture was never built to collect it. The safest data is the data you never took. None of this shows up in a feature comparison. It shows up years later, in the absence of a harm — which is exactly why it is so easy to skip and so important not to. The right test for an AI meant for children is not how much it can do. It is how carefully someone thought about what it must never do, and whether they built for that first. We did. Everything else aime does sits on top of that decision. — the aime team

“A confident wrong answer is the most expensive thing in a classroom.”

There is a particular failure mode that should worry anyone putting AI in front of teachers, and it is not the one people usually name. The fear we talk about is that the machine will not know the answer. The real danger is that it will produce a wrong one — fluently, confidently, formatted like a proper lesson — and hand it to a teacher who has no time to check. A generic chatbot is optimised, at its core, to produce text that sounds right. Most of the time, sounding right and being right travel together. But in a classroom the gap between them is where the damage lives: a worked example with a subtle error, a "learning objective" that is really just a topic, an explanation pitched two years above the students in the room, a sequence that teaches the hard idea before the one it depends on. None of these look wrong. They look like a lesson. And a tired teacher at nine in the evening, handed something that looks like a lesson, will reasonably assume it is one. This is why trust, not capability, is the real barrier to AI in education. Teachers are not withholding trust because they are behind the times. They are withholding it because they have been burned by tools that were confident and hollow — and because they know something the demo does not: a bad lesson does not fail quietly. It fails in front of thirty children, and the teacher is the one who has to repair it. We built aime to earn that trust structurally, rather than to ask for it. Every lesson it produces is grounded in the curriculum and in learning science — sequenced the way understanding is actually built, pitched to the level of the class, checked against how children learn rather than against what is statistically likely to sound like a lesson. It is not generating plausible text and hoping. It is building from a model of how learning works. And the last safeguard is the teacher. aime hands back a finished, editable lesson, not a verdict — every part of it theirs to keep, change, or discard. The judgement that tells a teacher "this example won't land with my Year 8s" is the one thing we will never try to automate, because it is the thing that makes them a teacher. The test I would put to any of these systems is simple. Not "can it generate a lesson?" — they all can. But: when it is wrong, how easily can the teacher tell? A tool that makes its reasoning legible and its output the teacher's to overrule is one you can trust in a classroom. A tool that hands over confident text and calls it done is a liability wearing the costume of help. — the aime team

“Most of the world's classrooms don't teach in the language AI was built for.”

The AI education story, as it is usually told, happens in English. The demos are in English. The benchmarks are in English. The frontier models think most fluently in English, because that is where their training data runs deepest. And so an impression forms, quietly and almost without argument: that the future of teaching with AI is an English-language future the rest of the world will eventually catch up to. Walk into a classroom in Cape Town, or a rural province of the Philippines, and that story falls apart in the first thirty seconds. The lesson is not in English, or not only. The curriculum is national, specific, and often nothing like the ones these tools were shaped around. The teacher there does not need an AI that can translate an American lesson plan into passable local phrasing. She needs one that understands the syllabus she is actually accountable to, in the language she actually teaches in, pitched at the level her students are actually working at. Most tools treat this as an edge case — a localisation problem to be solved later, once the "real" product is built. That gets the world exactly backwards. The majority of teachers on earth are the edge case only if you are standing in the wrong place. Stand in one of those classrooms and it is the English-first, Western-curriculum tool that looks parochial. aime was built curriculum-first, not language-first, and that distinction is the whole game. It starts from the national curriculum a teacher is teaching, in the medium of instruction that curriculum is delivered in, and grows the lesson from there — not a translation laid over a foreign template, but a lesson built from the local syllabus. Whatever a country teaches, aime teaches to it: every national curriculum, in every country we serve, is a first-class citizen — not a supported few with the rest left as an afterthought. And because the classrooms that most need this are often the ones with the least reliable connectivity, none of it depends on a live link to a distant server. The support arrives in the language and the curriculum of the room it is standing in, whether or not the signal does. The question I would put to any tool arriving in a country is not "does it support multiple languages?" — that is a feature list. It is: does it start from this country's curriculum, or from someone else's, and hope for the best? A tool that begins with the teacher's actual syllabus is building for the classroom. A tool that begins with English and translates outward is building for the demo. — the aime team

“The best substitute in the world still doesn't know your class.”

Every teacher knows the particular quiet of a room the morning a substitute walks in. The lesson may go fine. The substitute may be sharper than the regular teacher, better read, quicker on their feet. And still something is missing that no amount of talent replaces: they do not know that the girl by the window has stopped raising her hand since March, that this class needs fractions taught through money and not pizza, that the boy who looks defiant is really just lost and will meet you halfway if you never say his name in anger. The substitute begins at zero. They will begin at zero again tomorrow. I have been thinking about substitutes because most of the AI arriving in classrooms right now is, structurally, a permanent one. Brilliant every morning. Amnesiac every morning. You explain your class to it, and it helps, and then the session closes and it forgets you completely — your curriculum, your constraints, the seventeen small adjustments you made last week to reach one child. Tomorrow you will explain it all again. We have built tools with astonishing capability and no memory, and then wondered why teachers describe them as more work than relief. You are not being helped by a colleague. You are re-onboarding a stranger, every single day. This matters more than it first appears, because the models themselves are becoming a commodity. Everyone in this industry rents intelligence from the same handful of labs. The raw capability that felt like a moat two years ago is now a utility anyone can plug into. Which means the question of what actually makes an education tool valuable has quietly moved. It is no longer how clever is the model. It is what does the tool remember about this teacher, this room, these children — that no one else holds. That is the thing we decided, early and deliberately, to build aime around. From the first worksheet a teacher generates, aime is writing to a persistent teaching memory — quietly, in the background, without asking them to fill in a profile or configure a thing. It remembers the curriculum they teach and the level they teach it at. It remembers that last month's lesson on photosynthesis needed to be pitched lower, and pitches the next one accordingly. It accumulates. The tenth lesson a teacher makes with aime is better than the first not because the model got smarter, but because aime finally knows them. The relationship compounds the way a real teaching partnership does — the way a substitute never can, because a substitute leaves. I want to be precise about who this is for, because it is easy to imagine memory as a luxury feature for the already well-served. The opposite is true. The teacher who most needs a partner that remembers is the one carrying six classes with no aide, no planning period worth the name, no colleague down the hall with time to spare. For them, a tool that forgets is just one more thing demanding to be re-explained. A tool that remembers is the first thing in years that meets them where they left off. And because aime holds that memory on infrastructure we own and run — including offline, in the classrooms the cloud has never reached — that continuity does not depend on a teacher having the connectivity to rebuild it from scratch each day. There is a test I would offer any school leader weighing one of these tools, and it has nothing to do with the demo. Ask it something on a Monday. Then come back on Friday and ask it something that depends on Monday's answer. If it stares back at you like you have never met, you have bought a very expensive substitute — capable, tireless, and permanently new to your building. If it remembers, you have bought something rarer, and something teaching has always run on: a partner that gets to know the class, and stays. The intelligence in these tools will keep getting cheaper and more evenly distributed. What will not commoditize is memory — the slow, specific, irreplaceable knowledge of particular children in a particular room. That was always the teacher's real expertise. A tool worth putting beside them should be built to hold it, and to hand it back. — the aime team

“Adoption Was the Easy Part”

Something quietly historic happened this year, and it arrived without ceremony. In June, Microsoft published its 2026 AI in Education report — one of the largest surveys of its kind — and the headline was stark: 88% of educators and 92% of students have now used AI for their work. The debate we were having eighteen months ago, about whether artificial intelligence belonged in the classroom at all, is over. It's already there — in almost every staffroom, on almost every desk. Matt Jubelirer, who leads education marketing at Microsoft, put it plainly: "Educators around the world are embracing AI as a classroom ally, and they're now asking not if, but how to make the most of it." He's right. The question has moved from whether to how — and how is a far harder question than the adoption numbers make it look. Because the same report carries a second finding that drew far less attention. Only 53% of educators have had any formal training in the tools they are already using. Two-thirds of them — 66% — want regular training and aren't getting it. More than four in ten name academic integrity as an active worry. Teachers adopted AI the way people adopt anything handed to them under pressure: quickly, quietly, and mostly alone. Adoption, it turns out, was the easy part. Support is the part we skipped. This is the gap that worries me, because I've watched who absorbs it. When a technology spreads faster than the understanding around it, the people at the front carry the difference as stress. Teachers become unpaid pilots of tools nobody trained them to fly — asked to police accuracy, integrity, and trust on top of everything they already hold. That isn't empowerment. It's a new kind of workload wearing an optimistic name. I've heard the hopeful version of this moment too, and I want to be fair to it. MacKenzie Price, who co-founded Alpha School, framed it well this summer: "AI is taking over repetitive administrative work so teachers can spend more time doing what only humans can do — building relationships, inspiring confidence, coaching students through challenges." I believe that vision completely. But I'd add one uncomfortable line to it: reinvention is a promise, not a product. It only comes true if the tools are actually built to deliver it. That distinction is the whole reason aime exists — and it's worth being precise about what we mean. Most AI tools wait to be asked. They sit behind a blank prompt and hand the real work back to the teacher: one more thing to learn, manage, and second-guess. We built aime to work the other way around. It starts from the teacher's actual timetable and curriculum, does the routine planning before anyone asks, and hands back finished, editable lessons rather than a conversation. Every suggestion is the teacher's to keep, change, or discard, because judgement is the one part of teaching we will never try to automate. And it runs whether or not the connection does — because the teachers who most need their time back are often the ones the cloud forgot. None of that shows up in an adoption statistic. A school can be "using AI" and still send its teachers home more tired, not less. So here is the test I'd hold all of us to — aime first of all. Not "how many teachers are using AI?" That question is answered. But "did the teacher end the week more themselves, or less?" Adoption is a number. Support is a promise. And the promise is the only part that was ever going to matter. — the aime team

“On Giving Time Back”

I keep returning to a simple test for any tool that claims to help teachers: at the end of the week, does the teacher have more of themselves, or less? Most technology fails it quietly. It promises to save time and then asks for time — to set it up, to prompt it, to check its work, to fight it when it's confident and wrong. The teacher becomes the tool's assistant. That is the opposite of the arrangement we should be building. The premise I care about is narrow and, I think, non-negotiable: technology should give teachers time back, not take more of it away. That means the system does the proactive work — starting from the timetable, not from a blank prompt. It means every suggestion is the teacher's to keep, edit, or throw out, because judgement is the part of teaching we should never automate. And it means the measure of success isn't how impressive the AI looks, but how much lighter the person feels. The future of the classroom doesn't belong to machines, and it never belonged to overworked humans doing it all alone. It belongs to what the two can do together — when the machine takes the load, and the teacher gets to teach. — the aime team

“The future belongs neither to humans nor machines alone.”

It belongs to what they can achieve together. This is the sentence we end on, because it is the sentence we began with. The future of education is not a competition between human teachers and intelligent machines. It is a partnership in which each does what they are uniquely capable of, in service of an outcome neither could produce alone. The machine brings scale, speed, consistency, tireless availability, and the capacity to hold vast amounts of context at once. The human brings judgement, relationship, presence, moral seriousness, and the irreducible quality of being a person in the same room as another person at a moment of learning. These are not overlapping capabilities. They are complementary, and the most powerful educational configurations are the ones that combine them deliberately. aime exists to make this combination practical at scale. The companion is built to be the machine half of the partnership — capable, restrained, curriculum-aware, teacher-aware, designed to disappear into the work it supports. The teacher remains the human half — central, irreplaceable, supported in ways that allow their work to be more present, more precise and more sustainable than it would otherwise be. This is not a neutral vision. There are other futures available. There is a future in which machines are positioned to replace teachers, and education becomes a thinner, cheaper, lonelier version of itself. There is a future in which schools refuse to engage with the new capability at all, and watch the gap between what is possible and what is delivered widen each year. Neither future is good for students. Neither future is good for the adults who teach them. Both are increasingly easy to imagine. The future we are building toward is different. It is one in which teachers are at the centre of how AI enters education, in which the technology is shaped by educational purpose rather than by what is technically novel, and in which the result is a profession that is stronger, more capable and more sustainable than it has been in living memory. This is not a small ambition. It will not be realised by aime alone, by any one school, by any one system. It will be realised by the slow accumulation of better daily decisions about how to use this new capability in service of the oldest goal in education: helping every learner become more than they were, in the company of a teacher who believed they could. Together. That is the only word that matters.

“Every learner deserves support.”

Every teacher deserves support. Educational AI should help deliver both. The dual deserving is easy to say and hard to operationalise. Most educational technology has tried to support one side at the expense of the other. Direct-to-student tutoring platforms support learners while bypassing teachers, sometimes undermining them. Teacher-facing platforms support teachers in narrow ways that do not always translate into learner experience. Neither side gets the consistent support the sentence implies. aime is built around the conviction that the two are inseparable and need to be supported in the same motion. The way to support every learner well is to support every teacher well. The two are not parallel projects. They are the same project, viewed from different angles. For learners, this means the support arrives through their teacher. A student who has a well-prepared, well-rested, well-mentored teacher gets better support than any direct-to-student platform could provide, because the teacher knows them. aime's contribution to the learner is the strength of the teacher in front of them — and indirectly, through the resources, feedback and personalisation that the teacher can now provide because the companion is absorbing the surrounding load. For teachers, this means the support is meaningful and ongoing. The companion is not a tool the teacher uses occasionally to produce content for students. It is a continuous professional resource that helps them plan, mark, differentiate, communicate, grow, and stay in the profession longer. The teacher's professional life is treated as something to be supported in its own right, not as an instrument for delivering support to students. The dual support model has implications for how the product is designed. We do not let one side dominate the other. We do not optimise for student engagement at the cost of teacher overload. We do not optimise for teacher efficiency at the cost of student experience. Every design decision asks both questions. Both have to be answered well, or the decision does not ship. The deeper claim is that the false choice between supporting students and supporting teachers has held the sector back. The two are the same thing, and the educational AI that takes this seriously will be the one that produces durable improvement. Every learner deserves support. Every teacher deserves support. Both, together. aime is built for both, together.

“Artificial intelligence can accelerate learning.”

Only humans can inspire it. The distinction between acceleration and inspiration is the cleanest way to describe the division of labour between AI and teachers in the next era of education. AI is genuinely good at acceleration. It can shorten the time between question and answer, between draft and final, between practice and feedback, between current understanding and next step. These accelerations are real, valuable, and worth pursuing. aime is built to deliver them. But acceleration without inspiration produces a faster version of compliance, not a richer version of learning. A student who is accelerated through content they do not care about is just bored more efficiently. A student who is inspired to care about the content learns it whether the system is fast or slow. The first job is inspiration. Acceleration is what makes the inspired learning go further. Inspiration is a human achievement. It comes from the teacher who loves the subject visibly enough that students wonder what they are missing. It comes from the moment of being seen by an adult who believes the student is capable of more. It comes from the encounter with an idea, mediated by a person who knows why it matters, that opens a door the student did not know was there. None of this is producible by an algorithm. All of it is producible by teachers, and many teachers do it every day, despite the conditions in which they work. aime's contribution to inspiration is indirect. We cannot generate it; we can protect the conditions in which it happens. A teacher who is not exhausted is more able to inspire. A teacher who is well-prepared is more able to inspire. A teacher who is still in love with their subject in year fifteen is more able to inspire. A teacher who feels respected and supported by the surrounding system is more able to inspire. The companion's job is to make all of these more likely, and to do so reliably across the messy reality of the school year. The two functions, working well together, produce something neither could produce alone. Accelerated learning, in directions that have been inspired by humans. The student moves faster than they would have, in a direction they care about, with the support of an adult who believes in them and the assistance of technology that does its job quietly. Accelerate the work. Inspire the worker. aime and teachers, doing different things in service of the same outcome.

“The schools that thrive in the age of AI will not necessarily be those with the most technology.”

They will be those with the clearest educational purpose. In any moment of technological change, the temptation is to assume that the schools with the most access to the new tools will pull ahead. The historical evidence is the opposite. The schools that pull ahead are those that have the clearest idea of what they are trying to achieve educationally, and that use whatever technology is available in disciplined service of that purpose. Tools without purpose produce activity. Purpose with tools produces results. This will be true of the AI era as much as it was true of every previous wave. The schools that decide their purpose first — what kind of education they are trying to provide, what kind of human they are trying to help young people become, what kind of professional life they are trying to offer their staff — will use AI well. The schools that adopt AI because everyone else is, without that prior clarity, will produce expensive activity and little change. aime is built to be most useful to the first kind of school. The companion serves purpose rather than imposing one. It does not arrive with a view about what kind of education the school should be providing. It arrives with the capability to support whatever educational vision the school has already articulated, and to help that vision be implemented with more depth and consistency than would otherwise be possible. For a school whose purpose is academic excellence with strong pastoral care, aime supports the rigorous curriculum, the precise feedback, the data-informed practice — and also the teacher bandwidth that pastoral care actually requires. For a school whose purpose is inclusion and reaching every learner, aime supports the differentiation, the multilingual adaptation, the careful tracking of individual progress. For a school whose purpose is preparation for a complex future, aime supports the inquiry-rich curriculum, the assessment that values reasoning over recall, the cross-curricular connections that make the world legible. The technology is plastic. The purpose is the structure that gives it shape. System leaders considering AI investment should ask the prior question first. What is our educational purpose, and what would it look like to pursue it with greater depth? The answer to that question should govern the technology decision, not the other way around. aime's job is to support whatever answer the school arrives at, and to make the answer easier to live up to. Purpose first. Technology in service. Always.

“The future of education isn't about replacing expertise.”

It's about making expertise accessible. Expertise has always been the scarce input in education. There are not enough subject specialists, curriculum designers, assessment experts, learning scientists or experienced mentors to go around. The system has tried to handle this scarcity through batching — gather expertise in central places, dispense it through training events, hope it cascades through the layers below. The cascade has always been lossy. aime makes a different bet. Rather than batching expertise into events, we are distributing it into daily work. The companion makes the equivalent of a curriculum specialist, an assessment expert, a learning scientist and an experienced mentor continuously available in every classroom that wants them — not as replacement for the human experts in the system, but as on-demand support that was previously impossible to provide. The replacement framing misunderstands what is happening. The curriculum lead in a school still leads. Their job becomes richer, not redundant — they shape what aime's curriculum awareness looks like for their school, they intervene where the companion's suggestions are not quite right, they focus their personal time on the structural and human work that the companion cannot do. The expert is not replaced. The expert is freed to be more strategic. The same applies to mentors. The early-career teacher with a mentor still has a mentor. The conversations between them get richer because the daily flow of professional support — the planning help, the feedback craft, the in-flow learning — is happening already. The mentor can focus on the relational and developmental work that only a human can do, rather than on the basic survival support that used to consume the mentoring relationship. Assessment experts, learning scientists, subject specialists — the pattern is the same. The companion absorbs the routine application of their expertise, releasing the experts themselves to do the deeper and more bespoke work that justifies their training. This is what accessible expertise looks like at scale. The early-career teacher in the rural school has access to the same baseline of expert support as the teacher in the famous research-school. The mentor in a small system can support more colleagues because the daily layer is handled. The specialist in a low-resource context has tools that previously only the well-funded systems could afford. Expertise widely distributed is one of the great democratic projects of this century. aime is one attempt to take it seriously.

“Teachers shouldn't have to choose between innovation and wellbeing.”

Both should be possible. For too long, the implicit deal in education has been that innovation comes at the cost of wellbeing. The teacher who wanted to try a new approach, integrate a new tool, redesign a unit or change their feedback practice paid for the innovation with their evenings and weekends. The system rewarded innovation rhetorically and punished it operationally. Unsurprisingly, many teachers concluded that wellbeing required playing it safe. aime is built to dissolve this false choice. The companion makes innovation cheap, in the sense that matters most — cheap in teacher time. A new approach can be drafted, tested, refined and embedded without the cost being borne entirely by the teacher's personal hours. A concrete example. A teacher wants to shift their assessment practice toward more frequent, lower-stakes formative checks rather than infrequent summative tests. Historically, this kind of shift would require the teacher to design new instruments, build a new marking workflow, change how they record evidence, and adjust how they communicate with students and parents. Each of these is a multi-evening project, and the cumulative cost makes the shift impractical for most teachers under normal load. With aime beside them, the same shift becomes a series of small, supported decisions. The companion drafts the formative checks, suggests the marking workflow that minimises overhead, helps redesign the recording, and adjusts the student-facing and parent-facing language. The teacher's role is to decide, refine and adopt. The hours required collapse from many evenings to a couple of focused planning sessions, all within working hours. The same pattern applies across the range of innovations a teacher might want to attempt. New curriculum sequence. Different homework approach. Revised feedback practice. New approach to differentiation. New way of engaging parents. In each case, the companion takes the production cost of innovation onto itself and lets the teacher focus on the judgement. The cumulative effect is a profession in which innovation is no longer a luxury for teachers who happen to have unusual amounts of personal time. It becomes part of normal practice — the natural way the work evolves, supported rather than punished by the surrounding system. Wellbeing protected. Innovation accelerated. The two are not in tension when the support around the teacher is built deliberately. aime's job is to make sure they are not.

“There is a difference between generating content and generating learning.”

The distinction matters. It is now trivially easy to generate content. A model will produce a lesson plan, a worksheet, a quiz, a feedback comment, a parent letter, a unit overview — fluently, instantly, at any volume. The supply of plausibly educational content has, almost overnight, become effectively infinite. The supply of learning has not. Learning is what happens, slowly, in the minds of students, when the right material meets the right teacher in the right sequence at the right time. Content is the raw input. Learning is the output, and the relationship between them is mediated by everything we know about pedagogy, curriculum, relationship and the specifics of the individual child. aime is built around this distinction with religious seriousness. The companion does generate content — first drafts of lessons, sets of practice questions, feedback comments, parent communications. But every generation is filtered through the question: is this likely to produce learning, in this context, with this teacher, for this class? Generation that does not pass this filter is not surfaced. The filter has many layers. Curriculum awareness, so the content fits the sequence. Pedagogical grounding, so the structure respects what we know about cognitive load, retrieval and feedback. Teacher awareness, so the suggestion fits the professional who will deliver it. Student awareness, so the pitch matches the actual class. None of these is dramatic; together, they are the difference between useful suggestions and impressive nonsense. The risk in the broader market is that schools, overwhelmed by the volume of generated content now available, will mistake quantity for value. A platform that produces a hundred lesson plans an hour is not a hundred times better than one that produces one carefully-grounded plan an hour. Often it is worse, because the volume creates a new kind of cognitive load on teachers who now have to triage what is worth using. aime's positioning is the opposite. We produce less, more carefully. Every suggestion is meant to be usable, with light editing, by a teacher who trusts that the underlying work has been done — curriculum check, pedagogical grounding, contextual fit. The teacher's time goes to the choices that matter, not to filtering noise. Content is cheap. Learning is precious. The distinction has always existed and it has never been more important to act on.

“The best educational leaders ask:”

"What do teachers need?" before asking: "What should teachers do?" The order of these two questions is one of the most diagnostic things about an educational leadership team. Leaders who ask "what should teachers do?" first treat teachers as instruments of strategy — the variable to be adjusted in pursuit of the goal. Leaders who ask "what do teachers need?" first treat teachers as the strategy — the people whose growth, capability and energy the school is fundamentally built around. aime is built to be most useful in the second kind of school. The companion's full value is realised when it is paired with a leadership culture that takes teacher need seriously and acts on what it learns. In those schools, the time and capability aime returns to teachers gets invested back into the things teachers said they needed — planning depth, mentoring, professional learning, manageable workload, room to be excellent. In schools where leadership defaults to the second question first, aime still helps, but the gains are smaller. The hour returned to the teacher gets absorbed by the next initiative on the leadership roadmap. The capability built by daily companion support gets directed at compliance tasks rather than pedagogical depth. The companion delivers, but the system around it does not metabolise the delivery into something larger. We share this honestly with school leaders considering aime. The companion is a serious accelerant for schools that are clear about what their teachers need. It is a useful but limited tool for schools whose primary orientation is to extract more from teachers. The product will not change the orientation. The orientation has to change the product's place in the school. The good news is that the orientation is itself shiftable. Leadership teams that start to ask "what do teachers need?" — and then act on the answers — discover that the question reshapes everything downstream. Curriculum decisions, assessment routines, professional learning, communication with parents, even the design of leadership meetings. The school becomes a different kind of place, and aime's contribution within it becomes much larger. The question is not a slogan. It is a discipline. Ask it first. Listen carefully. Act on what you hear. The rest of the strategy follows from there. What do teachers need? It remains the most underused question in educational leadership. The best leaders ask it relentlessly. aime is built to be useful to them.

“Technology should help schools become more human.”

Not less. The risk of any new technology in a school is that it crowds out the human texture that makes the place work. The risk grows with the capability of the technology. A weak tool can be ignored. A strong tool, badly deployed, can change the rhythms of a school in ways that are hard to reverse. aime is built with deliberate attention to this risk. The companion is designed to make schools more human, not less, by clearing the administrative weight that has been displacing human moments from the day. Consider where the human texture of a school actually lives. It lives at the door at the start of a lesson, in the conversation in the corridor between periods, in the lunchtime supervision that turns into a meaningful chat, in the after-school club where a different kind of relationship is possible, in the staffroom where colleagues compare notes on what is working, in the parent meeting where a difficult truth is shared with care. These moments are the connective tissue of a school. They are also the first things to be sacrificed when teachers run out of time. The most damaging effect of teacher overload is not the marking that gets done at midnight. It is the conversations that do not happen because there is no spare attention for them. The student who needed a word and did not get one. The colleague who needed a quick check-in and was waved past. The parent who sensed the teacher was rushing and chose not to share what they came to say. aime's contribution to the humanity of schools is to give back the attention budget that these moments require. The teacher whose week is twenty percent less full of administrative work has more bandwidth for the door, the corridor, the lunchtime and the staffroom. Multiplied across a staff, the school feels different — warmer, more responsive, less harried. We are explicit about this in how we work with schools. The hour saved is not meant to be filled with another initiative. It is meant to land somewhere that increases the human texture of the place. The schools that use aime best are the ones that protect this principle deliberately. The technology serves the humanity of the school. If it ever stops doing that, the technology should change. aime's design discipline is to make sure it never does.

“Educational AI should be curriculum-aware. Context-aware. Teacher-aware. Student-aware.”

Otherwise it risks becoming noise. This is the design specification for serious educational AI, and it is exacting. Each of the four awarenesses is non-trivial. Together, they describe the difference between a tool that is genuinely useful in a classroom and a generic content engine that happens to be pointed at education. Curriculum awareness means the companion knows the sequence, the prerequisites, the connections and the destination of every unit it touches. It does not generate a lesson in isolation. It generates a lesson that fits where the class actually is in the longer arc of their education. aime's curriculum maps are built by educators for each system we serve, not extracted from general training data. Context awareness means the companion understands the realities of this school — the timetable, the cohort, the term-time pressures, the school's policies on assessment and behaviour, the wider community the school serves. A suggestion that ignores these realities is not a helpful suggestion. aime is built to absorb context from the systems the school already uses, so the support arrives appropriately shaped. Teacher awareness means the companion learns how this teacher works — the style of feedback they prefer, the resources they have used before, the strengths they bring, the topics they are growing into. Over time, the support becomes calibrated to the individual professional, not to a generic teacher. aime does this through the natural flow of use, not through extra configuration burden on the teacher. Student awareness means the companion holds a working model of each student in the class — what they have understood, what they have struggled with, what their next step looks like, what context matters for them. This is the part of teaching that experienced teachers carry in their heads for thirty students at once, and that newer teachers find hardest. aime makes this carrying easier by holding the model alongside the teacher, ready to inform any decision the teacher is making. Without all four awarenesses, an AI tool in a school becomes noise. It produces output that is technically fluent and educationally wrong — a lesson at the wrong level, a worksheet at the wrong pace, a feedback comment for a generic student rather than this one. Noise of this kind is worse than silence, because it consumes the teacher's attention to correct. Build the awarenesses. Earn the right to be heard. aime is built to that bar.

“The future of professional learning may be continuous rather than episodic.”

Supported rather than scheduled. The episodic model of professional learning — booked sessions, external trainers, large group formats, calendar-managed — was a sensible response to the constraint that expertise was expensive to bring to teachers. That constraint has loosened. The technology now exists to make professional learning continuous, contextual and individual. The question is whether the profession will reshape its assumptions to match. aime is one of the instruments through which this reshaping is happening. The companion's daily presence in the teacher's work creates an unbroken thread of learning opportunity. Every lesson planned is a chance to deepen pedagogical content knowledge. Every piece of work marked is a chance to refine feedback craft. Every parent conversation is a chance to grow communication skill. The development is in the doing. This does not eliminate the need for episodic professional learning entirely. There remain things that are best done in groups, face to face, away from the daily grind — the deep dive into a new curriculum, the shared analysis of practice across a faculty, the sustained engagement with a piece of research, the human connection across schools and regions. These have their place and aime is no substitute for them. What changes is the balance. The episodic sessions, freed from the burden of delivering all the development, can do what they are best at — building shared understanding, generating community, surfacing the hard questions. The daily work, supported by the companion, does the steady accumulation of small improvements that, over a career, become professional mastery. This shift has implications for how schools and systems design their professional learning offer. The expensive whole-staff session on a topic only a third of teachers needed becomes harder to justify when the companion can deliver that support to the relevant third in flow. The CPD calendar designed around what is convenient for trainers becomes harder to defend when continuous support is available. The accountability conversations around professional learning shift from attendance at sessions to evidence of practice change over time. aime is built to support this shift. We do not see ourselves as the entire professional learning system for any school. We see ourselves as the daily layer underneath the human layer, supporting the kind of continuous, contextual growth that has always been the secret of the best teachers and that is now becoming available to everyone. Continuous beats episodic. Supported beats scheduled. The future is already being built this way.

“A well-supported teacher can transform hundreds of lives.”

Supporting teachers is one of the highest-impact investments any education system can make. The arithmetic is striking. A primary teacher might teach thirty children a year for thirty years — nine hundred children, each spending a full year in their care, in a period of life when they are most receptive to being shaped. A secondary teacher might teach five hundred students a year across their career, with shorter exposure but at a developmentally pivotal stage. The leverage is enormous, and largely under-appreciated in policy discussions that often treat teachers as interchangeable units. This leverage works in both directions. A teacher who has been supported well produces hundreds of better educated, better encouraged, better launched young adults over a career. A teacher who has been ground down by an unsupportive system produces a thinner version of the same outcome — or, increasingly often, leaves the profession altogether, and the system inherits the cost of training a replacement who will face the same conditions. aime exists because we believe the support-the-teacher lever is among the highest-impact, most under-invested available to education systems. The technology that has reached the profession in the past two decades has, on the whole, made teaching harder rather than easier. The opportunity is to reverse that pattern at scale. What does well-supported actually look like? It looks like the early-career teacher who survives their first three years with their idealism intact, because they had a companion that helped them plan, mark and respond to parents without drowning. It looks like the mid-career teacher who is still growing in their fifteenth year, because the daily flow of teaching includes ongoing pedagogical support rather than three CPD days a year. It looks like the senior teacher who stays in the classroom rather than moving into administration, because the classroom is still rewarding. System leaders making decisions about where to invest should weigh the teacher-support lever seriously against the more visible alternatives. New buildings, new devices, new accountability frameworks — each has its place. But the highest expected return, by some distance, is in the human and infrastructural support that allows good teachers to be more capable and to stay longer. Hundreds of lives per teacher. The maths argues for itself. Build the support, and the impact compounds across generations of children whose names will never appear in the procurement memo.

“Curiosity remains one of humanity's greatest superpowers.”

No algorithm can replace it. Curiosity is the first verb of learning. It is what makes a child pull at a thread, a student linger after class with one more question, a teacher try a new approach after twenty years in the job. Without it, education becomes compliance — content delivered, attendance recorded, outputs measured, nothing alive. With it, the same content produces something different in every learner. aime is built with deep respect for curiosity and deliberate caution about the ways in which technology can suffocate it. A poorly designed AI tool can suffocate curiosity in at least three ways. It can answer questions too fast, training students out of the productive struggle that builds understanding. It can over-personalise content, removing the surprising encounter with material the student would not have chosen. It can replace the human conversation that turns a momentary spark into a sustained inquiry. The companion is designed to avoid all three. When a teacher uses aime to support a student, the suggestions emphasise productive struggle over instant answers. They are built around the questions a great teacher would ask — the ones that hand the cognitive work back to the learner rather than doing it for them. The companion's role is to make the teacher more able to sustain this kind of questioning, not to bypass it with shortcut explanations. We are also deliberate about what we will not build. We do not build direct-to-student tutoring agents that aim to answer every question the student has. We do not build platforms that personalise content so narrowly that students never encounter the productive friction of an unexpected text. We do not build engagement loops that hook attention without producing learning. These design choices cost us features, and we are at peace with that cost. On the teacher side, aime is designed to protect teacher curiosity too. The companion does not flatten the teacher's intellectual life into template completion. It surfaces interesting connections, suggests alternative approaches, and brings the kind of provocative reading or example that might re-spark a teacher's interest in a topic they have taught many times. The professional curiosity that sustains a long career is itself a resource to be looked after. Algorithms can support curiosity. They cannot generate it. The generation remains a human achievement, and our job is to protect the conditions under which it keeps happening.

“The greatest educational breakthroughs are often invisible.”

They're the moments when teachers have what they need, exactly when they need it. The history of education is full of dramatic-sounding innovations that produced very little, and quiet ones that produced a great deal. The dramatic ones — new buildings, new devices, new curricula imposed from above — tend to make headlines and budgets. The quiet ones — strong mentoring, coherent curriculum, reliable cover, a colleague who listens — tend to make the actual difference and are rarely celebrated. aime aims to belong to the second category. Our best moments are the ones that produce no headline. The teacher whose planning was strong because the companion surfaced the right starter at 6pm on Sunday. The teacher whose marking turned around in two days because the companion drafted the comments and held the consistency check. The teacher who had the right differentiated worksheet ready for the student who joined the class halfway through the unit. The teacher who walked into a lesson on a topic outside their specialism feeling prepared rather than exposed. None of these moments is filmable. None of them produces a quote for a brochure. Each of them, multiplied across a teaching career, changes the texture of the profession and the experience of the students inside it. This is why we are suspicious of demonstration culture in educational technology. A product that demos well is not the same as a product that works well in the actual rhythm of a school week. The demo is choreographed. The week is messy. The features that survive contact with the mess are different from the features that look most impressive in a forty-minute pitch. aime is built to be evaluated by the mess. Does the companion help on the Tuesday when three things have already gone wrong by 9am? Does it help in November when energy is low and the term feels long? Does it help in the second year, after the novelty has faded? Those are the conditions under which real value either accumulates or evaporates. The invisible breakthrough is the most valuable kind. The teacher who, at the end of a hard term, says "I don't know how I would have got through that without it" — and means it not as a tagline but as the literal truth of their week. That is what we are building for. Invisible is fine. Useful is everything.

“Educational technology often asks:”

"How can teachers use this?" Better question: "How can this support teachers?" The two questions sound similar and produce very different products. The first puts the teacher in service of the technology — the teacher's job becomes to find ways to integrate the new tool into their practice, to justify the procurement, to make the platform's metrics move. The second puts the technology in service of the teacher — its job is to fit into the work the teacher is already trying to do, to make that work easier, to disappear when it is not needed. aime is built around the second question. Every design conversation begins with what the teacher is actually trying to do — plan a lesson, mark a set of essays, prepare for a parent meeting, recover from a difficult class — and asks how the companion can support that work. The capabilities of the underlying model are downstream of that question, not upstream. This sounds obvious. In practice it is not how most educational technology is built. The more common pattern is that a capability becomes available, a product team imagines uses for it, and a sales motion is built around persuading schools that they need it. The teacher is asked to find a place for the new thing in their week. The new thing rarely earns its keep. The reframed question forces honesty. If we cannot describe how a feature supports a teacher in work the teacher actually wanted to do, the feature does not ship. If we can only describe how a feature might create new work for the teacher, the feature definitely does not ship. The discipline rules out a great deal of what would otherwise have been built. It also reshapes how aime is sold and deployed. We do not ask schools to invent uses for the companion. We arrive with a clear account of the work it supports — planning, marking, differentiation, communication, professional growth — and we are explicit about the work it does not support, so that schools are not misled into expecting magic in places where the human work is irreducible. The cultural shift this represents is large but overdue. The whole industry would benefit from inverting the question. Stop asking how teachers can use technology. Start asking how technology can support teachers. The products that result will be different, and the profession will be the better for it. Support the teacher. The rest follows.

“The future of education is personalised.”

But personalisation without teachers is not education. The dream of personalised learning is old. It predates AI by decades — every adaptive textbook, programmed instruction system and personalised learning platform of the past sixty years has, in various ways, promised it. The dream has resurfaced with new force now that AI can plausibly tailor content to individual learners at scale. It is worth stating clearly what personalisation can and cannot do. It can match content to the student's current level. It can adjust pacing. It can offer variation in modality. It can provide instant feedback on closed-form tasks. It can free students from waiting for the rest of the class. These are real benefits, and aime is designed to support them. What it cannot do is constitute an education on its own. Education is not a series of personalised content deliveries. It is the formation of a person — their habits of mind, their relationships with ideas, their sense of themselves as a learner, their understanding of what they owe to others. None of this happens through content alone, however well-matched. It happens through sustained relationship with adults who model it, expect it, and notice when it is faltering. aime's view of personalisation is therefore strict. The companion supports the teacher in personalising the parts of learning that should be personalised — the practice tasks, the scaffolds, the next-step questions, the reading at the right level. It does not attempt to personalise the parts that depend on human relationship — the discussion that surfaces a student's thinking, the encouragement that lands because the teacher knows the student, the high expectation that holds even when the student wants to give up. This is why aime refuses to be a direct-to-student tutoring platform that bypasses the teacher. The technology would be capable of such a product. The educational logic is wrong. A student personalised to entirely by an algorithm is being prepared for a thinner version of life than a student personalised to by an algorithm and a teacher together. The right configuration is teacher plus AI, with the AI handling personalisation of content and the teacher handling personalisation of relationship and expectation. aime is built for that configuration and we will continue to build for it. Personalisation, yes. Personalisation without teachers, no.

“AI should help teachers spend less time preparing and more time engaging.”

This is the simplest possible statement of aime's purpose, and it remains the right one. The engagement — being present with students, in the moment of learning — is the irreducible core of the job. Everything else is, in some sense, preparation for it or aftermath of it. The arithmetic of the profession has, over decades, pushed too much of the week into the preparation and aftermath, and squeezed the engagement. aime exists to reverse that arithmetic. The companion takes on preparation work in two ways. The first is by handling the parts that are genuinely standardisable — the formatting, the differentiation, the resource adaptation, the curriculum alignment, the basic structure of a unit plan. These are tasks where the teacher's professional value-add is small, and yet they have historically consumed disproportionate time. aime absorbs them. The second is by accelerating the parts that are not standardisable but that follow patterns the teacher could not personally master across every topic and every class. The pedagogical content knowledge for a topic outside the teacher's specialism. The misconception map for a concept the teacher has not taught for three years. The sequence of questions that builds toward the desired understanding. aime brings these patterns to the teacher's planning, fast. What remains, on the preparation side, is the work that is genuinely the teacher's — the choices about pitch, the decisions about which student needs what, the judgement calls about how to open and how to close, the small acts of personalisation that make a lesson belong to this teacher with this class. These choices benefit from being made by a teacher who is not exhausted by the standardisable work that came before. On the engagement side, the time saved compounds. A teacher who walks into a lesson having spent twenty minutes planning rather than two hours is fresher, more flexible, more able to adapt to the room. They are more present at the door at the start. They are more available for the conversation after the bell. They are more likely to notice the student who needs noticing. This is what the future of the profession looks like at its best. Less time alone at a screen. More time with young people. aime's job is to shift the balance, week by week, in that direction. Less preparation. More engagement. Same teacher, more present.

“The next generation of educational systems should understand curriculum.”

Not simply store content. There is an important difference between a system that holds a pile of resources and a system that knows what the resources mean. The first is a filing cabinet. The second is a colleague. Most educational technology to date has been a filing cabinet — a place where content lives, organised by tags, searchable by keyword, but ultimately inert. aime is built to be the second. Understanding curriculum means the companion knows that this lesson on quadratic functions sits inside a sequence that started with linear relationships and will lead, in three weeks, to roots and discriminants. It knows that the prerequisite for today's work is comfort with rearranging equations, and that the typical sticking point for students at this stage is the visual link between the algebra and the parabola. It knows that the school's scheme of work allocates four periods to this concept, and that the assessment in eight weeks will require students to apply this in an unfamiliar context. This level of understanding is what allows aime to make suggestions that are actually useful. When a teacher asks for a resource, the companion does not return everything tagged "quadratics". It returns the resource that fits this point in this unit for this class, pitched at the right level, building on what was taught last lesson and feeding into what is coming next. The understanding is built deliberately, in partnership with teachers and subject specialists, across every curriculum aime supports. It is not extracted from a model's general knowledge of the world. It is the result of careful curriculum mapping — what comes before, what comes after, what the connecting threads are, what the typical learner trajectory looks like, where the common misconceptions sit. This is unglamorous work. It is also the reason aime's suggestions feel different from generic content generation. A content generator can produce a lesson plan that mentions quadratics. A curriculum-aware companion produces a lesson plan that fits where the class actually is. For schools, this distinction is the difference between buying a smarter filing cabinet and gaining a thoughtful department colleague who happens to be available on demand. The first saves a little time. The second changes the texture of professional practice. Store content if you must. Understand curriculum if you can. Build the second.

“When teachers thrive, students thrive.”

Simple. Powerful. True. This is the sentence we return to when we are tempted by complexity. Every product decision at aime has, eventually, to answer to it. Will this help a teacher thrive? If yes, it earns its place. If no, it is overhead. Teacher thriving is not a vague aspiration. It has components that can be supported deliberately. It includes sustainable workload, the experience of growing professionally year over year, a sense of mastery in the craft, supportive colleagues, time for life outside the job, and the feeling at the end of most days that the work mattered. aime works on the components it can touch. The workload component is the most obvious — the companion absorbs the small administrative weight that drains evenings and weekends, returning hours to the teacher's actual life. The professional growth component is supported through the daily in-flow learning the companion enables: every lesson planned, every piece of work marked, every unit reviewed becomes an opportunity to get a little better. The mastery component is supported through the depth of curriculum and pedagogical content knowledge the companion brings into reach. The components aime cannot directly touch — colleagues, leadership culture, school climate — we work to support indirectly. Strong aime deployments are paired with conversations about how the returned time gets invested back into the human dimensions of professional life. The hour the companion gives back is not meant to be filled with new admin. It is meant to land somewhere meaningful: more planning depth, more mentoring time, more family dinners. The student side of the equation is, in our experience, downstream and reliable. Teachers who are not exhausted are warmer in the classroom. Teachers who are not behind on marking give feedback faster. Teachers who have planned with care teach with conviction. Teachers who are still in love with the job after fifteen years bring something to the room that no replacement teacher, however well-prepared, can substitute for. The sentence is simple because the underlying causal chain is genuinely simple — even if it has been ignored in policy debates that prefer more complicated stories. Healthy teachers produce healthy classrooms produce healthy learning. aime's contribution is to make teacher health a little easier to sustain across a long career. When teachers thrive, students thrive. We build to that, and we measure against it.

“The purpose of educational AI is not efficiency alone.”

It is effectiveness. Efficiency and effectiveness are easy to confuse and very different. Efficiency is doing things faster or cheaper. Effectiveness is doing the right things, well, in ways that produce the outcome that matters. Education has plenty of efficient practices that are not effective and plenty of effective practices that are not efficient. The job of educational AI is to lean toward effectiveness, not just speed. aime is built on this distinction. We could make the companion more efficient in shallow ways — auto-generate everything, fill every dashboard, push every notification — and the metrics would look impressive. We have deliberately chosen not to. The product is shaped by the question: what helps teachers teach more effectively, and what is just busy-work dressed up as productivity? Effectiveness in teaching has a particular grain. It is the lesson that is well-pitched for this class, not just any class. It is the feedback that the student can act on, not just feedback that exists. It is the unit that produces durable understanding, not just coverage. It is the assessment that informs the next teaching decision, not just the next data submission. aime is tuned to support these things, even when they are slower than the alternative. A small example. aime will sometimes suggest that the teacher slow down a sequence rather than speed it up — because the evidence on this concept suggests that more time on the prerequisite will pay off later. A pure efficiency tool would never make that suggestion. An effectiveness-oriented companion does. Another. aime's marking support encourages the teacher to write fewer, better comments rather than more comments faster. The evidence on feedback is clear that quality beats quantity. Efficiency would push for more output. Effectiveness pushes for better output. This orientation has implications for how aime is evaluated. We track effectiveness measures — student progress, teacher practice change, professional retention — alongside the efficiency measures that are easier to capture. We are willing to be slower on the latter to be stronger on the former. The deeper point is that education has spent decades trying to optimise for efficiency in ways that have left effectiveness flat. The next generation of tools should resist this. The point is not to do the wrong things faster. It is to do the right things at all. Effectiveness first. Efficiency in service of it. Never the other way around.

“Teaching remains one of the most intellectually demanding professions in society.”

AI should recognise that. Not underestimate it. There is a strain of educational technology that treats teaching as if it were a relatively simple, largely scriptable activity — as if a model that could generate a passable lesson plan had captured most of what the job involved. This view is wrong, and the products built on it are correspondingly weak. Teaching is the simultaneous management of subject knowledge, pedagogical content knowledge, classroom dynamics, individual learner needs, curriculum coherence, assessment evidence, behavioural patterns, emotional context and timing — all in real time, across thirty children, for five hours a day, year after year. There are very few professional activities in society that demand more. aime is built with this recognition at its centre. We do not assume the teacher is the part of the system that needs simplifying. We assume the teacher is the most sophisticated component in the room, and the support around them needs to rise to that level. This shows up in how the companion behaves. It does not condescend. It does not present the obvious as if it were a revelation. It does not bury the teacher in suggestions, knowing that a teacher's attention budget is finite and precious. It assumes the teacher knows their craft and offers help in the register of a knowledgeable colleague rather than a tutorial system. It also shows up in what the companion will and will not do. aime will draft, suggest, surface, summarise and remember. It will not pretend to make the judgements that depend on being in the room — which student to call on, when to push, when to back off, when to abandon the plan because something more important has just happened. Those calls are honoured as the expert work they are. When educational AI underestimates teaching, the result is products that feel patronising to good teachers and dangerous in the hands of weak ones. The good teachers will not use them. The weaker teachers will be pushed further away from the deep practice they need to develop. aime's bet is the opposite. Respect the demand of the work. Build support that meets it. Trust the teacher to make the calls that matter. Over time, you produce more capable teachers, not less. Underestimate the job and you produce neither. The work is hard. The support should be honest about that.

“The strongest schools are learning organisations.”

Not technology organisations. A learning organisation is one whose members get better, individually and collectively, over time. The structures, routines and culture of the place are bent toward that goal. The technology in the building is in service of it, not a substitute for it. The strongest schools we know are not the ones with the most impressive device estates. They are the ones with rich professional conversations in the staffroom, strong mentoring structures for early-career teachers, honest assessment of what is working and what is not, and a leadership team that protects the conditions under which adults can learn. aime is built to strengthen the learning organisation, not to bypass it. The companion is most powerful in schools that already have a culture of professional inquiry — because in those schools, the time and capability aime returns to teachers is invested back into that inquiry. The unit gets rewritten. The mentoring conversation goes deeper. The faculty meeting discusses pedagogy rather than logistics. This is why aime deployments include support for the human side of professional learning, not just the technical onboarding. We help schools think about how to use the time the companion gives back. We surface patterns across the staff that can become the basis for shared inquiry. We make it easier for heads of department to see what is working in colleagues' classrooms and to spread it. Where schools treat aime purely as a technology procurement — install it, train people on the buttons, walk away — the impact is modest. Where they treat it as part of building a stronger learning organisation, the impact compounds year over year. The implication for system leaders is that the question is not "what technology should we adopt?" It is "what kind of organisation do we want to be, and what technology will help us become that?" The answers are different for different schools, and the same technology can produce very different outcomes depending on the organisational soil it lands in. aime's strongest partnerships are with schools that are clear about the kind of learning organisation they are trying to build. The companion is then a serious accelerant. Without that clarity, even the best tool is just another platform. Be the learning organisation first. The technology serves that, or it serves nothing.

“We often measure educational technology by adoption.”

Perhaps we should measure it by teacher capability. Adoption is a flattering metric because it is easy to move. Licences sold, accounts created, logins recorded. None of these tell us whether the teachers in the system are any better at their work than they were before. A platform can be adopted by every teacher in the country and still leave the profession exactly where it found it. The measurement that matters is teacher capability — the actual quality, range and confidence of professional practice in the system. It is harder to measure, because capability is multidimensional and slow-moving. But it is the only metric that connects to anything that matters downstream for students. aime is built to be evaluated by capability metrics, even though they are more demanding for us than adoption metrics would be. We track whether teachers using the companion plan more coherent units over time, write more useful feedback comments, run more precise formative assessment, differentiate more confidently, and stay in the profession longer. We commission independent evaluation rather than relying on internal usage data. We publish what we find, including when the findings are uncomfortable. This discipline has shaped the product. A feature that drives adoption but does not improve capability — a flashy dashboard, an engaging gamified flow, a notification stream that gets the teacher to open the app — is a feature that distracts from the real work. We have killed several such features after watching the capability data fail to move. It has also shaped how we work with schools. Successful aime deployments are not measured by the percentage of staff who have logged in. They are measured by the changes in practice across the staffroom — the conversations about pedagogy that get richer, the new teachers who hit professional milestones faster, the experienced teachers who stay engaged for longer. The broader call here is for the sector to grow up about its metrics. The next generation of educational technology should be procured, evaluated and renewed on the basis of what it does to teacher capability over time, not on the basis of how many people opened the app last week. Adoption is the vanity metric. Capability is the one that matters. Build, buy and evaluate accordingly.

“The most important educational innovation may not be AI itself.”

It may be what AI allows teachers to focus on. This is the unfashionable view inside an industry that loves to talk about the technology. But it is the view aime is built around. The technology is the means. The end is what becomes possible for teachers when the technology is doing its job quietly in the background. What becomes possible? Consider what teachers have always wanted to do more of, and rarely had time for. More noticing. The early-career English teacher who finally has the bandwidth to spot which two students have been quiet for three lessons in a row and to find a moment for them after class. The veteran maths teacher who can see, because the data is now legible at a glance, that the cohort's grasp of place value is shakier than it looks on the surface — and can quietly re-teach it before it becomes a problem. More planning. Not in volume, but in depth. The teacher who can spend the saved hour on Sunday actually thinking about how to open Monday's lesson on the partition of India in a way that holds the room — rather than scrambling to assemble the basic resource pack. More feedback that lands. The teacher who can write a comment that names the specific move the student made, the specific next step that will unlock the next level, and the specific way to practise it — because they are not racing to finish marking before midnight. More relationship. The teacher at the door at the start of the lesson, greeting students by name, asking about the football match, noticing the new haircut, registering the flat affect that means a private conversation is needed later. This is the unglamorous work that produces the conditions in which learning happens. More professional growth. The teacher who has time to read the article, watch the colleague teach, reflect on a difficult lesson, try a new approach. The continuous improvement that everyone agrees is essential and few teachers have ever had the time to sustain. aime's contribution to education will ultimately not be measured by what it does. It will be measured by what teachers, freed from the surrounding load, are able to do. The innovation is in the focus, not the feature. Build for the focus.

“Imagine if every teacher had immediate access to:”

Lesson support. Assessment support. Curriculum support. Differentiation support. That future is closer than many realise. In fact, for the schools already working with aime, it is no longer a future at all. It is Tuesday morning. Lesson support means a teacher sitting down to plan can ask for a starter that activates the relevant prior knowledge, a sequence of explanations that respects cognitive load, a check for understanding that genuinely checks, and a plenary that consolidates rather than rehashes — and have a strong first draft in minutes rather than the hour it would otherwise have taken. The teacher still shapes, edits, and decides. The blank page is gone. Assessment support means the teacher can ask for success criteria that translate the learning intention into student-facing language, a mark scheme that captures the distinctions that matter, exemplar responses at multiple levels, and a feedback bank grounded in the most useful next steps. The companion does not replace the teacher's judgement on any individual piece. It removes the friction around the parts that should be standard. Curriculum support means the teacher always knows where this lesson sits in the wider arc — what came before, what comes next, which threads run through the year, which prerequisites should already be secure. The companion holds the curriculum map in working memory so the teacher does not have to, and surfaces the connections that make individual lessons part of a coherent education. Differentiation support means every resource arrives in multiple versions, pitched at the spread of the actual class. The student with English as an additional language gets a version with the scaffolds they need. The student working two years ahead gets a genuinely stretching extension, not a longer version of the same task. The student who missed last week gets a bridge back in. None of this is hypothetical. None of it requires a research project to evaluate. It is the daily experience of teachers working with aime today. The future people are still imagining for the profession is already arriving for the schools that have decided not to wait. The only question left is how quickly the rest of the system can catch up — and how to make sure the catching up happens in a way that strengthens teachers rather than bypassing them. Closer than many realise. Already here for those who chose it.

“The future classroom should adapt to learners.”

Not force learners to adapt to systems. The current default still runs the other way. Students are sorted into year groups by birthday, taught the same content at the same pace, assessed against the same milestones at the same times, and judged by how well they conform to a structure that was designed for administrative convenience more than for learning. The students who happen to fit thrive. The students who do not, struggle, often invisibly. True adaptation has been impossible at scale because it requires more attention than a single teacher with thirty students can possibly distribute. aime exists to make that attention possible — not by replacing the teacher's judgement, but by giving them the capacity to act on the differences they already see. The companion helps the teacher hold thirty individual learning trajectories in mind at once. It remembers which student needs the additional scaffold, which is ready for the extension, which has been absent and needs a bridge back into the current unit, which has flagged a wellbeing concern that should shape how they are approached this week. The teacher does not have to carry all of this in working memory alone. Practically, this shows up in the resources that arrive on the teacher's desk already differentiated, the marking comments that recognise the specific student rather than treating them as a generic case, the seating plan that accounts for the friendships and frictions in the room, the parent communication that lands with appropriate context. Crucially, the adaptation goes both ways. aime helps the system adapt to the learner — but it also helps the teacher hold the learner accountable to the curriculum's high expectations. Adaptation does not mean lower standards. It means meeting the student where they are so that they can be carried, by stages, to where the curriculum needs them to go. The longer arc of this work is system change. As more classrooms become genuinely adaptive, the assumptions of the year-group, same-pace, same-test model will start to look increasingly arbitrary. The schools and systems that adapt early will find themselves better at the central educational task: helping each individual learner make as much progress as they are capable of. The classroom that adapts is the classroom that teaches everyone. aime is built for that classroom.

“A teacher's confidence has a direct impact on student learning.”

We don't talk about this enough. The evidence on teacher self-efficacy is unambiguous and largely ignored. Teachers who believe they can teach a topic well, teach it better. Teachers who feel underprepared, however hard they work, transmit some of that uncertainty to their students. Confidence is not a soft trait. It is an instructional variable that shows up in attainment data. We do not talk about this enough because the policy levers for confidence are subtle. You cannot mandate confidence. You can only create the conditions under which it grows: strong subject knowledge, clear curriculum, reliable support, manageable workload, and the experience of being trusted as a professional. aime is built around those conditions. The companion grows subject knowledge in the moment of use. A teacher preparing a lesson on a topic at the edge of their expertise has aime beside them — explaining the underlying concepts at their depth of need, surfacing the typical misconceptions, suggesting analogies and worked examples. Over a term, the teacher's grasp of that topic strengthens without any formal study having taken place. Their confidence grows because their actual capability has grown. It clarifies curriculum. aime holds the unit in coherent shape, shows the teacher where this lesson sits in the wider arc, and makes the connections to prior and future learning explicit. The teacher walks into the room knowing not just what they are teaching but why it matters and how it fits. That clarity is felt by the students. It provides reliable support. The early-career teacher who used to dread the question they could not answer now has a trusted second opinion at hand. They are more willing to take pedagogical risks because they have a safety net. Confidence grows in the trying. It reduces the workload that drains confidence faster than anything else. A teacher who is exhausted by Thursday cannot bring confident energy to Friday's lessons. aime's quiet absorption of the surrounding load buys back the energy that confidence requires. This is the under-discussed mechanism through which AI can improve learning. Not by teaching the students directly. By making the people who teach them feel — and be — more capable. Confident teachers produce confident learners. Build for the confidence and the learning follows.

“Many teachers don't need another tool.”

They need a trusted companion. A tool is something you pick up to do a specific job. A companion is something that walks alongside you through the whole work. The distinction matters in a profession where the work is continuous, contextual and emotionally demanding, and where the cost of yet another tool has long since exceeded the benefit. aime is built as a companion, not a tool. The difference shows up in dozens of design choices. A tool is configured. A companion learns. aime does not require the teacher to set up dashboards, configure preferences or maintain a profile. It picks up the teacher's style from the work they do — the way they write feedback, the curriculum they teach, the resources they prefer, the students they are responsible for — and adapts accordingly. A tool is summoned for a task and then closed. A companion is present across the week. aime is there in the planning on Sunday, the lesson on Tuesday, the marking on Thursday, the report on Friday. The continuity is part of what makes it useful — it remembers, it carries context, it does not require the teacher to brief it from scratch each time. A tool is impersonal. A companion is trusted. Trust is built by being reliable, honest about uncertainty, restrained where restraint is appropriate, and consistent over time. aime is deliberately conservative about claims, transparent about how suggestions are generated, and willing to say "I don't know" when the situation calls for it. We would rather earn slow trust than borrow fast confidence. Crucially, a companion does not compete with the teacher's other relationships. aime is not trying to replace the mentor, the head of department, the staffroom colleague or the spouse who listens at the end of a hard day. It is trying to take enough off the teacher's plate that those human relationships have more room to breathe. The teachers we work with describe aime in companion language. "It just gets it." "It saved me on Tuesday." "It feels like having a really good TA who actually understands the curriculum." That language is not accidental. It is the result of years of deliberate design choices that prioritised trust over capability, presence over features, and quiet usefulness over flashy demonstrations. Another tool is the last thing teachers need. A trusted companion may be one of the first.

“The future of assessment isn't more testing.”

It's better understanding. The system has been testing students more than any previous generation in history. The marginal return on additional testing has long since flattened. More tests do not produce more learning; they produce more data, much of which is collected, analysed and never acted upon. The pile of evidence grows. The understanding of any individual student does not. The future of assessment is not in the volume of tests. It is in the depth and timeliness of the understanding the assessments produce. aime is built to make this shift practical. The companion is less interested in adding new assessments and more interested in extracting more meaning from the assessment activity that is already happening. Every piece of student work — a homework, an essay, a problem set, an exit ticket, a class discussion — contains diagnostic information about what the student knows, what they don't, and what they should do next. Most of that information is currently lost because there is no time to surface it. aime surfaces it. As the teacher marks, the companion is quietly building a model of each student's understanding, flagging misconceptions, noticing patterns across pieces, and proposing the next teaching move that the evidence supports. The teacher's marking time produces not just a grade but an actionable understanding. The same approach applies at class and cohort level. The companion can show the teacher, at a glance, which concepts the class has secured and which are still fragile, which students are progressing and which are stalling, where the curriculum is working and where it needs adjustment. None of this requires additional testing. It requires better attention to the assessment evidence that already exists. The educational point is to move from assessment as accountability to assessment as understanding. Tests have a place — they certify, they motivate, they create useful endpoints. But the day-to-day engine of learning is the small, continuous, formative loop: see what the student knows, decide what to teach next, watch what happens. aime is built to make that loop fast, precise and sustainable across a full timetable. Less testing. More understanding. Same evidence, more value, all in service of the next teaching decision.

“Professional development shouldn't happen three times a year.”

Support should be available every day. The traditional model of teacher professional development is a relic of an era when expertise was expensive to deliver and so had to be batched. Twice or three times a year, teachers gather in a hall, listen to an expert, complete a workshop, and return to their classrooms with a folder and the best of intentions. By the end of the following week, the daily pressures of the job have absorbed most of what was learned. This model persists not because anyone defends it on educational grounds but because the alternative — continuous, contextual, individualised support — has been impossible to deliver. Until now. aime is, among other things, a continuous professional development system. The companion does not wait for the next twilight session. It supports the teacher in the moment of need: when they are planning the lesson, marking the work, drafting the feedback, preparing the parent conversation. The development is embedded in the work, not bolted on as an extra. When a teacher is preparing a unit on a topic outside their core specialism, aime quietly surfaces the pedagogical content knowledge that an experienced colleague would bring — the typical misconceptions, the analogies that work, the worked examples that build the right schema. The teacher does not have to wait for a CPD session on this topic to be scheduled. The development happens now, in the planning. When a teacher is marking, aime suggests feedback patterns grounded in the evidence on what actually changes performance. Over months, the teacher's feedback improves without any formal training event having taken place. The development is in the repetition. When a teacher is reviewing a difficult lesson, aime can act as a coaching partner — asking the questions a good mentor would ask, surfacing the patterns in the data, offering hypotheses to test next time. The reflective practice that everyone agrees is valuable but few teachers have time for becomes routine. None of this replaces the human dimensions of professional growth — the colleague who watches a lesson, the mentor who knows the teacher's history, the network that gives perspective. Those remain essential. What aime adds is the daily layer, the in-flow support, the constant low-level investment in capability that the old model could never sustain. Development as a daily practice, not an annual event. That is the change.

“Educational equity isn't just about devices.”

It's about access to expertise. AI may become one of the most powerful equity tools ever created. The conversation about educational equity has long focused on inputs that are easy to count — buildings, devices, staffing ratios, hours of instruction. These matter. But they do not capture the deepest inequity in the system, which is the uneven distribution of expertise. A student in a well-resourced school is taught by teachers who, in turn, are supported by experienced heads of department, generous planning time, strong subject networks, and a shared culture of expert practice. A student in an under-resourced school may be taught by equally committed teachers who lack any of those surrounding supports. The difference shows up not on day one but over years — in feedback quality, curriculum coherence, the depth of the unit, the precision of the question asked. aime is one attempt to narrow this gap at scale. The companion brings curriculum specialist, assessment expert, learning science researcher and experienced mentor into every classroom that adopts it, regardless of postcode. The early-career teacher in a small rural school gets the same baseline of support as the head of department in a famous city school. This will not equalise everything. Buildings still matter. Class sizes still matter. The particular human in front of the children still matters most of all. But the uneven distribution of expertise — the part that has historically been hardest to fix because experts cannot be replicated quickly — is now genuinely addressable. We take this responsibility seriously. aime is priced to be reachable by systems that have historically been priced out of premium educational tools. Our curriculum coverage extends deliberately into the regions and subjects where commercial products have not bothered. Our partnerships with system leaders in low-resource contexts are core to the roadmap, not afterthoughts. There is a version of the AI future in which the new technology widens the gap — where well-resourced schools use it brilliantly and others fall further behind. There is another version in which it narrows the gap — where the support that was previously rationed by geography and budget becomes available wherever there is a teacher and a connection. aime is built for the second version. Equity through expertise, distributed at scale. That is the prize worth chasing.

“Great teachers adapt constantly.”

Perhaps the greatest value of AI is helping adaptation happen faster. The hidden craft of teaching is not the plan. It is the constant in-flight re-planning that happens when the plan meets the room. The opening question lands flat, so the teacher reroutes. Three students grasp the concept faster than expected, so the extension task moves forward. A misconception surfaces that was not predicted, so the next ten minutes are improvised around it. By the end of a good lesson, the teacher has made dozens of small adaptations, most of them invisible to anyone who was not watching closely. These adaptations are the heart of effective teaching. They are also the part of the job that is hardest to support, because they happen too fast for any traditional intervention to help. By the time a teacher could log in, search for an alternative resource, and return to the lesson, the moment has passed. aime is built to make adaptation faster. The companion holds the alternative explanations, the differentiated tasks, the extension problems, the scaffolds, the diagnostic questions — already aligned to the curriculum, already pitched at the class — in a form the teacher can call on in seconds rather than minutes. When the original plan needs to bend, the bending is supported. The same principle applies between lessons. A teacher finishes period three knowing that something did not work. aime helps them adjust the next iteration before period six — a new starter, a different sequence, a revised model answer. The adaptation that would have waited until next year's version of the unit happens this week. And it applies across longer arcs. A teacher reviewing a term's assessment data sees patterns that suggest the unit needs structural change. aime drafts the revised unit plan, redistributes the time, and proposes the new sequence. The improvement cycle that used to depend on a free weekend can now happen on a Tuesday afternoon. Adaptation is a teaching superpower. It has always been throttled by the speed at which a single human can re-plan under pressure. aime's contribution is to take some of that pressure off — to make the adaptation that great teachers have always done possible at the pace the classroom actually demands. Faster adjustment. Better teaching. Same teacher, more capable.

“The conversation about AI in education is often dominated by fear.”

A better conversation might be: How do we help teachers become even more effective? Fear is understandable. The pace of change is genuine, the unknowns are real, and the early examples of AI misuse in classrooms have made headlines. But fear is a poor compass for policy. Fear leads to bans, blanket restrictions and defensive postures that delay the harder, more useful conversation: how to channel the new capability toward the things education has always tried to do. aime exists to be a serious participant in that better conversation. Our starting position is that AI is now part of the educational landscape, whether systems welcome it or not, and that the responsible question is not whether to engage but how to engage with intent. The intent that matters is teacher effectiveness. Not student usage of chatbots. Not novelty in lesson design. Not headline-friendly projects. Just: are the teachers in this system more effective at teaching this year than they were last year, because of how AI has been brought into their work? That question forces the right design choices. It rules out the tools that put the model in front of the student and the teacher off to one side. It rules in the tools that put the model behind the teacher, amplifying their planning, their feedback, their differentiation, their professional growth. It rules out vanity adoption. It rules in patient integration. It also reframes the policy debate. Instead of asking how to prevent students from using AI to cheat, we can ask how to design assessments that reflect the world students will actually live in — where AI is a co-worker, and the skill is in directing it well. Instead of asking how to ban tools, we can ask how to train teachers to use the good ones with confidence. Instead of treating AI as a foreign object to be quarantined, we can treat it as a capability to be domesticated into the service of teaching. aime is not a neutral participant in this conversation. We have a view, and we build to it. Our view is that the future of education is brighter, not bleaker, if teachers are at the centre of how AI enters the profession. Fear is the wrong frame. Effectiveness is the right one. Build from there.

“Technology should remove friction.”

Not create it. This is a deceptively simple test, and most educational technology fails it. The honest audit of a typical teacher's digital day reveals dozens of small frictions: the login that times out, the file format that does not transfer, the dashboard that takes three clicks to reach the one number that mattered, the integration that broke after the update, the notification that arrived at the wrong moment, the platform that demands attention before it gives anything back. Each friction is small. Cumulatively, they are the reason "technology in schools" so often feels like a tax rather than a gift. aime's design discipline is the opposite. Every feature is interrogated against a single question: does this remove friction from the teacher's day, or add to it? If it adds, even slightly, it does not ship. In practice this means the companion lives where the teacher already works. It does not require a separate destination, a separate login, a separate place to remember. It speaks the language of teaching, not the language of software. It offers help without demanding configuration. It learns the teacher's preferences quietly, in the background, rather than asking them to be set up in advance. It also means we are willing to do unglamorous integration work so the teacher does not have to. aime meets the school's existing systems where they are — the timetable, the assessment platform, the document tools the staff already use — and absorbs the differences between them. The teacher experiences a single coherent companion. The plumbing underneath is our problem, not theirs. Removing friction is harder than adding features. It requires us to say no to things that would be straightforward to build. It requires us to be patient with the unflashy work of making the existing experience smoother rather than the new experience flashier. It requires us to measure success in seconds saved and small irritations removed, rather than in lines of new functionality shipped. The teachers who use aime rarely describe it in terms of what it does. They describe it in terms of what is no longer in their way. That, for us, is the right kind of compliment. The job of good technology in a school is to disappear into the work it makes possible. Less friction. More teaching.

“Students don't remember platforms.”

They remember teachers. Ask any adult to name the people who shaped them. The answers are not interfaces. They are faces. The teacher who said something at the right moment. The teacher who refused to give up. The teacher who saw something in them they had not yet seen in themselves. The teacher whose subject suddenly made sense because of the way they explained it. This is the data that should govern the design of every educational product. Anything that increases the chance of a student having that kind of teacher, in their lifetime, is worth building. Anything that decreases it — by exhausting teachers, distracting them, or pushing them out of the profession — is worth stopping. aime is built with this data in mind. We are not trying to be remembered. We are trying to be the reason a teacher is remembered. The companion stays deliberately in the background of the student experience. Students should not be opening aime, configuring aime, or relating to aime. They should be relating to the human at the front of their classroom, who happens to be better-prepared, better-resourced and less depleted than they would otherwise be — because aime has absorbed the surrounding load. We have made design choices that follow from this. The companion does not have a personality marketed to children. It does not seek engagement metrics from students. It is not a tutor that competes with the teacher for the student's attention. It is an invisible scaffold under the teacher's week, the existence of which most students will never need to know. When the system works, the credit belongs entirely to the teacher. The lesson that landed, the feedback that turned a struggling student around, the unit that finally clicked — these are teacher achievements, not AI achievements. aime's contribution is that the teacher had the time and resources to produce them, week after week, without burning out. In thirty years, the students currently sitting in classrooms will remember the people who taught them. They will not remember the platforms. aime's ambition is to help make sure the people they remember are still in the profession, still loved it, and were able to give them their best. The teacher is the memory. We are just the support.

“A school's greatest asset isn't its devices.”

It's its educators. Always has been. Always will be. Every few years a new piece of hardware is presented as the thing that will transform education. Interactive whiteboards. Tablets for every student. Virtual reality headsets. The transformation never quite arrives, because devices have never been the bottleneck. The bottleneck is the time, energy and expertise of the adults in the building. aime is built around this realism. We do not believe a school's future depends on its device ratio. We believe it depends on whether its educators are supported enough, prepared enough and respected enough to do their work well. The companion is designed to invest in the educator, not to compete for budget with them. We have seen schools spend extraordinary sums on hardware while the teachers using that hardware were under-resourced in the ways that actually mattered — no planning time, no curriculum continuity, no mentoring, no relief from the rising administrative load. The hardware sat unused or under-used because the humans who were supposed to bring it to life were exhausted. aime's value proposition is the inverse. The technology is light, the human investment is heavy. A school that adopts the companion is investing in the capability of its teachers — their planning, their assessment, their feedback, their professional growth. The devices through which the companion is accessed are almost incidental. A laptop the school already owns, a tablet, a phone. The substance is in what the teacher can now do, not in what they hold. This framing matters for how schools make decisions. The question is not "what should we buy?" but "what would make our teachers more capable next term than they are this term?" Sometimes the answer is a piece of technology. More often it is time, support, a clearer curriculum, a stronger mentoring structure — and a companion that quietly amplifies all of those. The asset is and has always been the people. aime's job is to make sure the asset appreciates rather than depreciates over the course of a career — that teachers leave each year stronger than they arrived, supported into a longer and richer professional life. Invest in the educator. The rest is detail.

“The future belongs to teachers who know how to combine:”

Learning science. Curriculum expertise. Human judgement. Artificial intelligence. Each of these on its own is insufficient. Learning science without curriculum expertise produces techniques in search of content. Curriculum expertise without learning science produces coverage without retention. Human judgement without either produces well-meaning improvisation. AI without all three produces fluent nonsense at scale. The teachers who will define the next era of the profession are the ones who hold all four in working tension — and aime is built to be the connective tissue between them. The companion brings learning science to the surface of daily decisions. When a teacher is planning, aime quietly suggests the retrieval routine that the evidence supports, the worked example before the independent practice, the spacing of revisits across the half term. The teacher does not have to remember the literature in the moment. The literature comes to them. The companion is curriculum-aware in a deep sense. It knows what was taught in the previous unit, what is coming next, where the syllabus places this content in the wider arc, and which prerequisites the students should already hold. This is the kind of knowledge that experienced heads of department carry in their heads and that early-career teachers spend years assembling. aime makes it available to everyone on day one. Human judgement remains the teacher's. The companion suggests; the teacher decides. The decision is always informed by the things the model cannot see — the conversation in the corridor, the family circumstance, the energy in the room. aime's restraint at this layer is deliberate and load-bearing. It would be a much weaker product if it tried to overrule the teacher on the calls that depend on being there. Artificial intelligence is the fourth element, and it is the one that makes the other three combinable at speed. Without it, integrating learning science, curriculum and judgement is a Sunday-night job that few teachers have the energy to complete every week. With it, the integration happens continuously, in the background, as part of the natural flow of preparing and teaching a lesson. The future is not the teacher who knows the most about AI. It is the teacher who fluently combines all four — and aime exists to make that combination available to every classroom that wants it.

“AI has created a fascinating paradox.”

Knowledge has become easier to access. But wisdom remains as valuable as ever. This paradox sits at the centre of what schools are now for. For most of educational history, schools earned their place by being the most efficient way to deliver knowledge. That monopoly has ended. A curious teenager with a phone has access to more knowledge in a Tuesday afternoon than a medieval scholar could accumulate in a lifetime. What has not been democratised is wisdom — the capacity to judge what is true, what matters, what to do with what you know, and who to become because of it. Wisdom is not retrieved. It is grown, slowly, in the company of people who have it and are willing to share it. That, more than ever, is what schools and teachers are for. aime is built to protect this distinction. The companion is excellent at the knowledge layer — summarising, explaining, retrieving, generating. Deliberately, it is restrained at the wisdom layer. It does not pretend to make the calls that only a teacher who knows the student can make. It does not weigh in on whether this is the week to push a particular child, or whether the class needs a quieter lesson because of something that happened at lunch. Those judgements stay with the human. This restraint is a design choice, not a limitation. It would be technically straightforward to have aime offer more confident opinions on everything. It would also be educationally wrong. Wisdom transferred without lived context is just confident guessing, and confident guessing in a classroom is dangerous. So the companion supports the wisdom work by making the knowledge work fast. The teacher who would have spent an hour assembling the background facts for a discussion on civic responsibility now has those facts in five minutes — and an hour to think about how to lead the discussion with this particular group of fourteen-year-olds. The cognitive load shifts from gathering to judging. Schools that get this right will be more important than they have ever been. They will be the places where young people learn what knowledge is for — how to use it, question it, weigh it, and live with it. The knowledge itself will increasingly come from everywhere. The wisdom will still come from teachers. aime's job is to make sure the teachers have the time to give it.

“Every minute a teacher spends searching for resources is a minute not spent helping learners.”

Small efficiencies matter. There is a particular kind of invisible work that drains the teaching week: the hunt. The hunt for the worksheet that worked well two years ago. The hunt for the diagram that explained electromagnetic induction in a way the class understood. The hunt for the example essay that hit the mark band the students are aiming for. The hunt for the seating plan that balanced the difficult group. The hunt for the slide that used to make the joke land. Individually, each hunt costs minutes. Collectively, they cost hours every week. And because they are scattered across the day, in the gaps between lessons, they rarely show up in any workload audit. They are simply absorbed into "preparation time", which expands to fill whatever evening is available. aime is designed to end the hunt. The companion maintains a continuously-organised, curriculum-aligned memory of the teacher's resources, the school's shared library, and the wider evidence base of what has worked elsewhere. When the teacher needs the worksheet on quadratic factorisation, they ask. They do not search five folders, three platforms and an old laptop. More importantly, the companion can adapt what it finds. The worksheet that worked two years ago for a different class is offered in a version pitched at this year's group. The diagram is reformatted for the display the teacher is actually using. The example essay is paired with a scaffold for the students who need more structure. This is what small efficiencies look like at scale. A teacher saves three minutes finding the right resource, two minutes adapting it, four minutes preparing the differentiated version. Nine minutes back, per lesson, per teacher. Across a school of forty staff and twenty lessons a week, the numbers become serious — thousands of hours a year returned to the parts of teaching that matter. We are sometimes asked why aime does not focus on the dramatic capabilities — the auto-generated lesson, the AI tutor, the synthetic feedback loop. The honest answer is that the dramatic capabilities are downstream of getting the small things right. A teacher whose week is not consumed by the hunt has the bandwidth for the bigger work. A teacher still hunting at 10pm does not. Small efficiencies, compounded, are the foundation. aime builds them in deliberately, one minute at a time.

“The best educational systems don't simply deliver content.”

They build confidence. In learners. In teachers. In communities. Confidence is the quiet currency of education. It is what makes a student raise their hand when they are unsure, a teacher try a new approach when the old one is tired, a parent walk into a school building they once found intimidating. Without it, every other educational input is degraded. With it, learning compounds. aime treats confidence as an outcome to be designed for, not a happy by-product. The companion is built to grow confidence in three directions at once. For learners, confidence comes from feedback that is timely, specific and forward-looking. aime helps teachers turn around marking faster, write comments that name what the student did well and what to try next, and surface progress in ways the student can see and feel. A student who knows where they are and where they are heading is a confident student, even when the work is hard. For teachers, confidence comes from preparation, knowing the curriculum sits underneath them, and from the experience of being supported in real time. The early-career teacher who has aime beside them while drafting a unit plan is not guessing alone. They are receiving the kind of mentorship that used to depend on which mentor they happened to be assigned. Their confidence grows because the support is consistent. For communities — parents, governors, the wider public — confidence comes from being able to see the work of the school clearly. aime helps schools communicate progress in language that is honest, specific and human. Parents who understand what their child is learning, why it matters, and how they are doing, are partners. Parents who are kept in the dark become anxious, and anxious communities make brittle systems. A confident system is a resilient system. It absorbs setbacks. It tries hard things. It tells the truth when something is not working and adjusts. It does not need to perform success because it is busy producing it. The infrastructure of confidence is unglamorous. It is timely feedback, clear communication, consistent support, honest data. aime's job is to make that infrastructure cheap, reliable and available everywhere — so that confidence is not a luxury good distributed by postcode, but a baseline condition of being in education. Build confidence and the rest follows.

“Educational AI should not start with technology.”

It should start with pedagogy. The question isn't: "What can AI generate?" The question is: "What helps students learn?" The order matters more than it sounds. Start with the technology and you end up retrofitting pedagogy to whatever the tool happens to do well. Start with the pedagogy and you build the tool around what we already know about how learning actually happens. aime starts with pedagogy. The companion is shaped by decades of learning science — retrieval practice, spaced repetition, worked examples, the testing effect, cognitive load theory, the importance of prior knowledge, the role of metacognition, the conditions under which feedback actually changes performance. Every suggestion the companion makes is filtered through this body of evidence before it reaches the teacher. This is the difference between a content generator and an educational companion. A content generator will happily produce a lesson that violates everything we know about how attention works in a classroom of thirty Year 8s on a Friday afternoon. An educational companion will not. It will quietly steer the teacher toward a sequence that respects working memory limits, builds in retrieval, and ends with a check for understanding that actually checks for understanding rather than for compliance. The pedagogical grounding shows up in small ways. When aime drafts a set of practice questions, it spaces them across difficulty rather than clustering them. When it suggests feedback, it focuses on the next step rather than the score. When it proposes a unit sequence, it surfaces the prerequisite checks that experienced teachers know to run and newer teachers often skip. None of this is visible as "AI". It looks like good teaching practice, available a little faster and a little more reliably than the teacher could produce it under time pressure on a Sunday night. The risk in the current market is that schools will adopt tools built by engineers who have never taught and never seriously engaged with the evidence base. Those tools will generate content efficiently and learning poorly. aime is built the other way around — by educators, grounded in research, with the technology serving the pedagogy rather than the reverse. Start with how children learn. Build backward from there. The technology will follow.

“Teachers are expected to be:”

Educators. Counsellors. Data analysts. Content creators. Assessment specialists. Technology integrators. No wonder workload remains one of the profession's biggest challenges. The role has expanded without anyone formally agreeing to expand it. Each addition arrived with good intentions — more pastoral support for vulnerable students, more data-informed practice, more differentiated resources, more digital fluency — and each is, in isolation, defensible. The problem is the accumulation. No one in the original job description signed up to do seven jobs at once with the time allocated for one. aime's response to this is not to add an eighth thing. It is to take pieces of the existing seven off the teacher's plate, quietly, where the teacher would prefer support rather than ownership. Content creator? aime drafts the worksheet, the slide deck, the exit ticket, the homework — aligned to the unit, pitched at the right level, ready for the teacher to edit rather than originate. Data analyst? aime watches the patterns across the cohort, flags the student whose engagement has dipped, surfaces the misconception that keeps recurring in last week's quiz, prepares the data view the teacher would otherwise build by hand. Assessment specialist? aime suggests success criteria that match the learning intention, proposes formative checks that fit the lesson flow, drafts feedback comments grounded in what the student actually wrote. Technology integrator? aime is the integration. The teacher does not need to learn three platforms to get the benefit — the companion meets them where they already work. The counselling and the educating remain entirely the teacher's. Those are the parts of the role that are non-delegable, and they are the parts that the rest of the workload has been crowding out. The point of aime is not to reduce the teacher's contribution. It is to redistribute the teacher's time toward the parts of the contribution that only they can make. When teachers describe what they want from the future of the profession, they rarely ask for a smaller role. They ask for a more coherent one — a role where the time spent matches the work that matters. aime is one attempt to bend the role back toward that coherence. Seven jobs cannot be done well by one person. One job, supported well, can be transformative.

“The future classroom will contain more AI.”

But the future classroom will also require more humanity. Empathy. Trust. Belonging. Encouragement. These are not becoming less important. They're becoming more important. There is a paradox at the heart of the current moment. As models become more capable of imitating human conversation, the things that are unmistakably human become more valuable, not less. A student can get a passable explanation from any chatbot. What they cannot get is the teacher who remembers that they were anxious last week, asks how the weekend went, and notices when their shoulders drop because something at home is wrong. aime is built around this paradox. The more we can offload the mechanical and procedural parts of teaching to the companion, the more human bandwidth the teacher has for the irreducibly human parts. We do not see AI and humanity as competing for space in the classroom. We see them as inversely related — the more capable the assistant, the more present the teacher can be. Empathy takes time. It takes the kind of unrushed attention that a teacher cannot give while simultaneously trying to remember which student needs which extension task. Trust takes consistency — being seen, being known, being responded to the same way on a bad day as on a good one. Belonging takes the small rituals that say "you are part of this place" — the greeting at the door, the inside joke, the photograph of last term's trip still on the wall. Encouragement takes specificity — naming the exact thing the student did that was brave, hard, or unusually good. None of this is automatable. None of it should be. What aime does is protect the conditions under which these things can happen. A teacher who is not marking until midnight has more warmth on Tuesday morning. A teacher whose planning is already strong has more attention for the student at the back. A teacher whose admin is under control has more capacity for the parent in distress. The future classroom will be more technological in its infrastructure and more human in its experience. Those two facts are not in tension. They are the same fact, viewed from different angles. Build the support well, and the humanity has room to grow.

“A simple test for any educational technology:”

Does it make teaching easier? Or does it simply create another system teachers must learn? This test has saved aime from shipping a dozen features that demoed well and would have failed in practice. Every time we are tempted by a flashy capability — a new dashboard, a new analytics view, a new configuration screen — we run the question. If the answer is "the teacher would have to learn another thing", we go back to the drawing board. The accumulated cost of educational technology is rarely captured in procurement decisions. A school adds a new platform. It takes thirty minutes of training. The login lives in a new place. There are notifications to manage, settings to configure, integrations to maintain. Multiply that by the dozen platforms a typical teacher now uses, and a meaningful slice of the week is spent on the meta-work of using the tools, not on the work the tools were supposed to enable. aime's design north star is the opposite. The companion should feel like a colleague, not a system. It should arrive where the teacher already is — inside the planning document, inside the marking flow, inside the lesson preparation — rather than asking the teacher to come to it. Practically, this means we have said no to many things. No separate dashboard the teacher has to check. No notification stream competing for attention. No configuration screen for the teacher to maintain. The companion learns the teacher's curriculum, the school's policies and the class's needs in the background, and surfaces help only when it is genuinely useful. The test extends to language. aime speaks in the vocabulary of teaching — learning intentions, success criteria, formative assessment, retrieval practice — not in the vocabulary of software. A teacher should never have to translate from their professional world into the tool's world to get help. The discipline is hard. The temptation to add is constant. But every feature that does not pass the simple test is a feature that quietly steals time from the people the product is meant to serve. Easier or harder. That is the only question that matters.

“What if every teacher had access to a curriculum specialist whenever they needed one?”

What if every teacher had access to an assessment expert? What if every teacher had access to a mentor? This is the promise of educational AI. Not replacement. Support. For most of the profession's history, expertise has been rationed. A school might have one literacy lead, one head of mathematics, one assessment coordinator, one mentor for early-career teachers. These specialists are excellent and overstretched. They cannot be in every classroom at every moment of need. The new teacher who needs a quick read on whether a worksheet is pitched correctly at 9:47pm on a Tuesday is on their own. aime exists to close that gap. The companion holds the equivalent of a curriculum specialist, an assessment lead, a learning science researcher and a mentor — not as a substitute for the human leads in the building, but as an always-on first responder for the small questions that used to go unasked because there was no one to ask. A teacher drafts a unit plan. aime checks it against the curriculum sequence, flags a missing prerequisite from the previous year, and suggests a five-minute diagnostic to use at the start of the first lesson. A teacher marks a set of essays. aime spots that three students have made the same conceptual error and proposes a short re-teach for the next lesson. A teacher writes a difficult parent email. aime suggests a phrasing that lands the message without bruising the relationship. None of this replaces the human specialists. The literacy lead still leads. The mentor still mentors. What changes is that the teacher walks into those conversations better prepared, with sharper questions, having already exhausted the obvious checks. This is what equity inside a school actually looks like. The early-career teacher in the portable classroom at the back of the site gets the same baseline of expert support as the head of department in the office next to the principal. The teacher in a small rural school gets the same companion as the teacher in the well-resourced city school. Support, distributed at scale, is the promise. aime is one attempt to make good on it.

“The most powerful educational technology is still a great teacher.”

Everything else should support that reality. This is the design principle that underwrites every product decision at aime. When we sit down to scope a new feature, we do not ask what the model is now capable of. We ask whether a great teacher would feel more powerful, more confident and more present in their classroom because of it. If the answer is no, the feature does not ship — no matter how technically interesting it might be. Most educational technology over the past two decades has inverted this principle. Products have been built around what was newly possible — interactive whiteboards, learning management systems, video platforms, adaptive engines — and teachers have been asked to bend their practice around them. The result is the modern teaching workload: a stack of tools, each requiring login, configuration and ongoing attention, very few of which were designed from the teacher's chair. aime is designed from the teacher's chair. That means the companion lives where the teacher already works. It speaks the language of curriculum, not the language of databases. It does not ask the teacher to learn a new interface to access support — the support comes to the planning document, the marking pile, the parent message, the lesson the teacher is about to teach in twenty minutes. It also means we are deeply suspicious of anything that puts a screen between teacher and student. The companion's job is to clear the desk, not to occupy it. The most successful aime interactions are often the ones the student never sees: the planning that happened on Sunday evening, the differentiation that was ready when the bell rang, the marking that came back two days earlier than it would have otherwise. Great teachers existed before AI and will exist after it. What AI can do is amplify the conditions in which great teaching happens — by removing the friction that even the best teachers cannot eliminate on their own. We measure ourselves against that standard. Did this teacher have a better week because of us? Did this student get a better explanation, a faster piece of feedback, a more thoughtful question? Did the human at the centre of the classroom feel stronger? If yes, the technology has done its job. If no, it has missed the point entirely.

“One of the greatest misconceptions in education is that information equals learning.”

Students have access to more information than any generation in history. What they need is understanding. That is where teachers remain irreplaceable. A twelve-year-old today can ask a model to explain the French Revolution, summarise photosynthesis, or write an essay on the causes of the First World War. The information is available, accurate enough for most purposes, and instant. So why do they still need school? Because information is not the same thing as understanding, and understanding is not the same thing as wisdom. A student who can recite the causes of the war does not yet grasp how an ordinary person becomes complicit in atrocity. A student who can describe photosynthesis does not yet feel the strangeness of a leaf turning sunlight into sugar. A student who can summarise a poem has not yet been moved by it. These shifts — from information to understanding, from understanding to meaning — happen in the company of a teacher who knows the student well enough to ask the next question, sit with the silence, and notice the moment a confusion turns into a question worth following. aime is built around this distinction. Our companion is not a content generator pretending to be a tutor. It is a teaching assistant that holds the information so the teacher can hold the relationship. When a teacher asks aime to prepare a set of probing questions about a text, aime draws on learning science and curriculum context to suggest sequences that move students from comprehension to inference to evaluation. The teacher then chooses, edits and delivers — because only the teacher knows which student needs to be pushed and which student needs to be heard first. This is the division of labour that works. Models handle the abundance. Teachers handle the meaning. The risk in the current moment is that schools will be tempted to outsource the meaning-making part to whichever tool talks most fluently. aime's view is the opposite. The more capable the models become at handling information, the more precious the human work of teaching becomes — and the more carefully we must protect the conditions under which it can happen. Information has been democratised. Understanding still requires a teacher.

“The quality of a curriculum matters. The quality of teaching matters. But neither matter if teachers are too overwhelmed to deliver them effectively.”

Teacher wellbeing is not a separate issue. It is an educational outcome. For too long, wellbeing has been treated as a pastoral concern — something that lives in staff yoga sessions, mindfulness apps and end-of-term thank-yous. It is, of course, all of those things. But it is also something far more fundamental: the precondition for everything else the system is trying to achieve. A burnt-out teacher cannot deliver a brilliant curriculum. An exhausted teacher cannot run a precise formative assessment cycle. A teacher who has slept four hours, marked until midnight and is dreading the next day's parent meeting cannot bring the warmth, patience and improvisation that turns a lesson plan into a learning experience. This is why aime treats workload as the wellbeing issue. Most of what damages teacher wellbeing is not the children. It is the surrounding administrative weight — the duplicated data entry, the resource hunting, the marking pile that grew while the teacher was running an after-school club, the report comment cycle that eats December. aime is designed to absorb that weight. The companion drafts the report comment from the evidence the teacher has already gathered. It restructures the unit plan when timetable changes force a re-sequence. It pulls the differentiated version of yesterday's task from the resource library so the teacher does not start from a blank page on Sunday night. None of this is glamorous. None of it makes a keynote slide. But it is the actual texture of the work, and it is where wellbeing is won or lost. Schools that have used aime for a full year report a measurable shift in the conversations happening in staffrooms. Less talk about being underwater. More talk about the children. Less Sunday-night dread. More Monday-morning curiosity. Teachers staying in the profession who had quietly decided this would be their last year. The system-level point is straightforward. If we want better student outcomes, we need teachers with the capacity to produce them. Capacity is not a slogan. It is hours, energy and headspace. Treat teacher wellbeing as an educational outcome and the rest of the curriculum conversation becomes possible.

“Artificial intelligence will not replace teachers.”

But teachers using AI may eventually outperform those without access to intelligent support. The future is not humans versus machines. The future is humans working alongside machines. This is the central premise on which aime is built. We are not an automation company dressed in educational language. We are an education company that happens to use AI as the most powerful support tool the profession has ever had access to. The replacement narrative is seductive because it is simple. It allows commentators to predict, policymakers to budget, and headlines to be written. But it misunderstands what teaching actually is. Teaching is not the transmission of information — if it were, textbooks would have replaced teachers a century ago, and YouTube would have finished the job. Teaching is the act of standing with another human being while they struggle to understand something hard, and adjusting in real time to who they are, what they bring, and what they need next. No model does that. No model ever will. What models can do is take the surrounding load off the shoulders of the person who can. aime sits beside the teacher — not in front of the student — generating first drafts of resources, summarising assessment patterns, flagging the student who has gone quiet in the data for three weeks, holding the curriculum map in working memory so the teacher does not have to. The teachers who learn to use this kind of support well will pull ahead. Not because they are working harder. Because they are working with more of their attention free for the parts of the job that only they can do: the noticing, the encouraging, the explaining, the believing. The competitive frame is uncomfortable but accurate. A teacher with a strong intelligent companion has more bandwidth for relationships, more precision in feedback, more depth in planning. Over a year, the difference compounds. Over a career, it becomes generational. The question for every system, school and teacher is no longer whether to engage with AI. It is which kind of AI to engage with — the kind that replaces, or the kind that strengthens. aime exists for the second answer.

“The best teachers I know are not looking for more technology.”

They're looking for more time. Time to plan. Time to think. Time to support students. Time to teach. Perhaps the most important question we should ask about educational AI is not: "What can it do?" But: "What time can it give back?" At aime, this question is the founding measurement we hold every feature against. Before we ship anything — a planning companion, a marking assistant, a differentiation engine — we ask one thing: how many minutes does this return to the teacher's week? Not how clever it is. Not how impressive the demo. Just: time returned. The arithmetic of teaching is brutal. A secondary teacher with five classes might face thirty hours a week of contact time, plus marking, planning, parent communication, data entry, behaviour follow-ups, duty rosters, professional learning and the constant emotional labour of being present for young people. The hours simply do not exist. Something gives — usually sleep, family, or the depth of attention any single student receives. aime was built to put hours back. Not by removing the teacher from any decision that matters, but by absorbing the small repetitive tasks that compound across a week: re-formatting a worksheet for a student with dyslexia, generating three differentiated exit tickets aligned to a learning intention, drafting a parent message that captures the nuance of a tricky conversation, surfacing the marking comment a teacher already wrote three weeks ago for a similar piece of work. Each of these is a few minutes. Multiplied across a term, they become weekends. The teachers we work with don't describe the change in technological terms. They describe it in human ones. "I read with my own daughter again." "I went for a walk on Sunday." "I planned the lesson I actually wanted to teach, not the one I had time for." That is the measurement that matters. Technology in education has spent decades proving what it can do. The next decade will be defined by what it gives back.