The workbench
Think by making.
Small experiments that run in your browser. Pick one. It opens here, on the bench.
Bench 01 / Beneath the interface
How software runs.
Scheduling, compiled code, and search, one small view at a time.
Bench 02 / Many hands
Building together.
People and AI change one piece of work. Watch the work stay whole.
Bench 03 / Things in motion
A ball and an arm.
A cricket delivery and a robot arm. Each one moves by a small rule.
Field notes
Field notes
How the experiments on this workbench work, and what each model leaves out.
How the workbench works
- Every experiment runs in your browser. Nothing goes to a server.
- No accounts. No AI calls.
- Scripted experiments give the same result each time.
- Each experiment has its own address. Copy the link to share it.
What each model leaves out
- The event-loop explorer
- This teaching model leaves out rendering and requestAnimationFrame. The check runs the same program in a worker in your browser. It never runs code as text.
- A little WebAssembly
- Clang compiled the C source. A small interpreter on this page runs the same bytes, and your browser's engine checks the result.
- Find a way
- This toy uses a made-up grid. It does not show how the ship-routing prototype calculates routes.
- A shared idea
- This is an independent teaching example. It does not use Mindalike code, workflows, or services. The AI text is scripted.
- Who adapts?
- The real gitring asks an AI model about close plans. Here, fixed rules decide. gitring uses the same rules when the AI fails.
- Main moves
- The steps follow gitring's merge order and redo policy. The agents, the files, and the test results are scripted.
- Line and length
- A teaching model with simple physics and made-up constants. It is not ball tracking, and it does not apply the laws of cricket.
- Reach
- A flat arm with two links and no joint limits. Real arms have more joints, limits, and weight.
- What this shows
- Every experiment runs in your browser. Each one is small enough to read.
- Honest note
- Scripted experiments give the same result each time.
The experiments run locally in your browser.
Small enough to inspect.