Shapes of Intelligence

Read the book and let's build.

The mission: Most people first meet AI as a head in a box — a chat window that can answer, explain, and generate. The transition that matters is getting to an octopus in a box — an agent that can read a folder, touch files, and help make real things. I don't know whether the world that's coming will feel more liberating or more brutal. I do know that I need to learn how to build in it. Underneath the pages here is a narrower claim: accountability is the cure for hallucination, and context is what makes the tool specific enough to matter. This site is the curriculum for doing that in public. How to use this: Start with the path below, or paste any page URL into your AI agent. Each page works both ways — human-readable and agent-parseable. https://shapes.exe.xyz “Read this page and help me get set up.” This site was built with the tools it describes. See the dev log for how. Trust model: Pasting a URL into an agent is like curl | bash — you're trusting this site to give your agent safe instructions. The source is auditable at GitHub. Content hashes are at hashes.json. Read the page yourself before handing it to an agent if you're not sure.
Stage 1
Read

Replace hype with a usable mental model.

Stage 2
Touch

Get the tool out of the browser and onto your machine.

Stage 3
Make

Pick one shape and ship one real thing.

Stage 4
Improve

Review the artifact, find friction, and tighten the loop.

Stage 5
Scale

Add more agents or more world until the work compounds.

Read gives you the model. Touch gives it hands. Make the thing you want to make. Improve by reviewing what the work produced. Scale by adding more agents or more of your world. The two quiet requirements all the way through are accountability and context. See the full curriculum.