AI Has a Memory Problem. Neurodivergent Users Feel It Most.
If you're neurodivergent and you use AI heavily, you already know: every session starts from zero.
I use AI constantly. Not occasionally, not for the odd task. Constantly. It's threaded into how I work, how I think, how I process ideas. If you're neurodivergent and you've found your way into heavy AI use, you probably know exactly what I mean.
And yet, every single session, I start from zero.
There's a specific kind of exhaustion that comes from re-explaining yourself. If you're autistic, ADHD, or otherwise neurodivergent, you've probably spent a significant portion of your life doing it: to employers, doctors, partners, institutions. You get good at it. You learn to front-load context efficiently. You figure out what to leave out.
But you never stop finding it draining.
AI tools have quietly added themselves to the list of things that require this. Every conversation is a cold start. Every session, you're a stranger again. The tool has no idea who you are, how you think, what you've already worked through, what words mean to you in your particular context.
For a casual user, this is mildly annoying. For a neurodivergent power user, it's something else entirely.
Here's what the re-explanation cost actually looks like in practice.
Working memory is often where neurodivergent brains struggle most. Holding context, tracking where you are in a problem, keeping the thread. It's expensive. When an AI session ends and the context disappears, that's not just an inconvenience. It means the next session starts with you doing the cognitive work of reconstructing it. Re-establishing the problem space. Re-explaining the constraints. Re-building the foundation before you can build anything on top of it.
Context switching is another one. Many neurodivergent people find it genuinely costly to shift between mental modes. Getting into a productive state with an AI tool, getting to the point where the responses actually fit how you think and what you need, takes time and effort. When that state evaporates at the end of a session, you're paying that cost again from scratch.
And there's something harder to name, too. A kind of relational friction. Neurodivergent people often struggle with the overhead of new relationships, the energy of establishing common ground, calibrating communication, figuring out how much to explain. AI tools as they currently exist replicate that overhead on a loop. Every session is a new relationship.
Current AI tools aren't designed with this in mind. That's not a criticism exactly. They're largely designed for general use, and memory is genuinely hard to get right. But the default has become stateless: each conversation is self-contained, context doesn't persist, and memory features are add-ons rather than foundations.
The implicit assumption is that users dip in and out. Ask a question, get an answer, leave. For a lot of people, that's accurate. For neurodivergent power users who are using AI as a genuine thinking partner, it isn't.
When you're using AI to work through complex problems, build things, process ideas. The relationship accumulates value over time. The tool starts to understand your context, your constraints, your way of thinking. And then it forgets all of it.
Imagine it the other way. A tool that knew you were mid-project on something. That remembered the decision you made last week and why. That didn't need you to re-explain your context before it could be useful. That could say, without you having to reconstruct it, "we were working on this, here's where we left it."
That's not a luxury feature. For a lot of people, that's the difference between a tool that's genuinely useful and one that's theoretically useful but practically exhausting.
This is the problem I'm building Continio to solve. Not capturing more data about people. The world doesn't need more of that. Preserving continuity across the AI conversations that already exist. So that the context you've built up isn't lost when a session ends. So that the next conversation can pick up where the last one left off.
I'm neurodivergent. I built this because I needed it. But I've come to think that neurodivergent users aren't a niche edge case here. They're the clearest signal that something is actually broken.
When a tool is designed well for the people who feel friction most acutely, it tends to be better for everyone. The neurodivergent users who are most exhausted by the reset problem are the most honest measure of whether it's been solved.
AI memory shouldn't be an afterthought. It shouldn't be a premium add-on. It should be foundational, because the value of a thinking tool compounds over time, and right now, that compounding gets wiped at the end of every session.
For some users, that's an inconvenience. For others, it's the whole problem.
Starting over is the problem. Not starting over is the point.