Three systems, one problem.

The same failure, in three places.

Look at the three systems together and something becomes obvious. Each one attacks a different point in the same chain.

AILattice is about what goes in — AI systems read the web badly, and most sites give them nothing useful to work with.

Intellistasis is about what happens in the middle — an answer built on real analysis rather than a plausible guess.

AI Warlock is about what comes out — because one model, alone, will tell you it’s certain whether it is or not.

It started with what AI couldn’t read.

Around 2019, Bryan noticed something that wasn’t getting much attention: AI systems were reading the web badly. Content that made sense to a person arrived at a model as a mess — structure lost, meaning inferred, context guessed at.

The obvious fix was to make the content easier to read. The more useful realisation was that it wouldn’t have been enough. A model that reads a page perfectly still only summarises the page. What was missing wasn’t access to information — it was analysis of it. Nobody had done the work underneath, so AI was producing answers that sounded right rather than answers the data supported.

That became Intellistasis: the analysis layer that runs before the AI answers. Building it exposed the problem sitting above it — websites had no way to tell AI systems what they actually were — which became AILattice. AI Warlock came out of the same line of work.

Not a believer. Not a sceptic.

Nobody builds three systems at three different points of the same problem by accident. That’s someone who uses these tools enough to know exactly where they break, and would rather fix it than argue about it.

It’s a more useful position than either of the loud ones. AI is genuinely remarkable and genuinely unreliable, and both of those things being true at once is the whole reason this work exists.

Build it and find out.

Bryan is based in Sydney, Australia, and self-taught. The through-line in his work is making things rather than talking about them — not theorising about where technology is heading, but building the thing and seeing whether it holds up.

It’s a slower way to have an opinion, and a more honest one. A system either does what it claims or it doesn’t, and there’s no arguing with it.

More of the same, probably.

The gaps keep appearing faster than anyone can close them, which is an irritating problem to have and a good one to work on.