Shape Intelligence is an AI lab and product studio. The gap between a promising model and a tool someone reaches for every day is where most of the work lives, so that's where we spend our time.
Live site: https://keithagroves.github.io/Shape/
The parts of intelligence without clean answers yet — reasoning, memory, and how systems stay reliable once they leave the benchmark. Everything here is chosen because a product downstream needs the answer.
- Reasoning under uncertainty — getting models to know what they don't know, and to act sensibly when the ground is shifting.
- Long-horizon memory — systems that carry context across days and tasks instead of forgetting the moment the window fills.
- Evaluation that means something — measuring the behavior you actually care about, not the proxy that happens to be easy to score.
- Human-model interaction — the design questions that decide whether a capable model becomes a tool people trust.
Research earns its keep when it ships. Everything we learn feeds directly into tools people put in front of real work.
- Platform — the core models and APIs teams build on, tuned for reliability over demos.
- Applications — end-to-end products for the workflows where good AI changes the day-to-day.
- Partnerships — we take on a small number of hard problems with teams who own the domain.
Small team. Long attention span.
- Research and product at one table — no handoff between the people who study the models and the people who ship them. Same room, same roadmap.
- Reliability is the feature — a model that's right 99% of the time and honest about the other 1% beats one that's impressive and unpredictable.
- Slow is smooth — we'd rather ship one thing that holds up than ten that impress for a week.
That's the whole idea behind the name: taking raw capability and giving it a shape someone can hold.
This is a Next.js app.
npm install
npm run devOpen http://localhost:3000 — the main page lives in app/page.tsx. Pushes to main deploy to GitHub Pages.