The working memory you were never issued.

For a long time, having a non-standard cognitive profile meant carrying a workaround. Whatever you couldn't do reliably, you built a coping strategy for. The cost was the workaround. The thing you traded for it was usually invisible. Now that AI can do the workaround, the thing you traded for it becomes load-bearing.

This isn't a story about dyslexia specifically. But dyslexia is the cleanest example I have, because I've lived it.

A cognitive profile is a trade. Whatever scored low usually came with something that scored high. Working memory recall scored low on my university assessment - the reason my friends wouldn't let me on their pub quiz team. Pattern recognition, spatial reasoning, and big-picture thinking scored sharply. That's a trade, not a deficit list. A shape, not a verdict.

For a long time the cost showed up daily and the upside showed up rarely. Holding a thread for the duration of a conversation, surfacing the right reference at the right moment, keeping a long task on rails - the everyday administrative weight of work selected against me. The bigger pattern thinking - the things I'm actually built for - needed the right kind of room to land.

What's changed is AI. A good AI setup can do exactly what dyslexic working memory can't: hold things still while you think, surface the thread you dropped twenty minutes ago, take a half-formed idea you can almost see and give it structure. It's the working memory you were never issued, running quietly in the background. Which frees the brain to do what it's actually built for - the creative, connective, zoomed-out thinking AI itself can't do.

The deeper claim isn't "AI helps dyslexics." It's that in an age of AI, every cognitive profile becomes more itself. The AI absorbs the administrative friction; what's left is the thing each person was always meant to be doing. For some people that's deep focused execution. For others it's pattern-matching across domains. For others again it's the slow patient analytic work that doesn't scale on raw speed.

This is what I'm building toward in the work itself. The personal AI operating system isn't a generic productivity layer bolted onto everyone's life identically. It's cognitive infrastructure shaped to a specific person's profile - what they want to keep doing themselves, what they want held for them, what they want quietly handled in the background. Two operators of the same business with different cognitive profiles end up running different versions of the same plugin. That's the point.

The future I'm betting on isn't "AI for the average user." It's AI as personal cognitive infrastructure - the working memory you were never issued, plus the executive function, plus the long-context recall, plus whatever else your particular profile traded for the thing you're actually good at. AI fills the synaptic gaps so the upside of every kind of thinking can finally show up unhindered.