← All projects

Drag to orbit, ctrl and scroll to zoom, W A S D to flyHow it works and the numbers

How it works

  1. Write it, break it

    You write the requirements as a model, and the harness breaks every decision in the code every way it knows, in one compile; each break must fail a test.

  2. One graph

    Every requirement, step, method, branch, break and test becomes a node in one graph, and that is the graph flying above, drawn with its names removed.

  3. Checked reasoning

    Claude writes its reasoning as a chain of steps, each citing what it stands on, and a citation that is not there fails the run.

  4. A proposer

    A graph neural network proposes which requirement owns new code, as a question and never a verdict, and it has to beat plain rules.

  5. Self-improving

    Self-improvement is on by default: every pass records its own stats, checks them against its own rules and names what to improve next.

The numbers

Each one says what it was measured on.

On facts

From our knowledge graph. Not measured on code.

99%

accurate when the graph can back the answer; it abstains otherwise.

99.03% of 35,951 blind derivations, on facts, from our knowledge graph.

616 of 616

top-tier facts right against outside truth (Wikidata), on a frozen population of 2,714.

On real code

One feature of a game, the one in the graph above.

96.5%

of 1,104 breaks of one feature caught by a test, leaving out those proven to change nothing (23), those with a written reason (20) and those with no verdict yet (52).

  • 974 caught
  • 35 uncaught, unexplained
  • 95 left out

The 130 red points in the graph are those 95 and those 35.

294 of 317

statements held by a test, up from 31 of 82. Held means breaking it fails a test.

216×

less to read to learn who owns a method: 775 characters at the median against 167,665. This is the size of what is read; no agent was timed.

10

requirements nobody had written, found by walking the use cases, beside the designers' 17.

  • 17 the designers wrote
  • 10 found by the walk

Where it comes from

31 of 63

ARC3 lineage: its best agent reached 31 of 63 games where random reached 14.

What it costs, and its limits

Costs

  • 23 minutes for a full pass of 1,104 breaks.

Limits

  • Prototype 0.1.
  • Built for large, complicated systems. To build something simple, do not use it. It has not been measured on large systems yet.
  • A green result on a thin model means little.
  • 95% accuracy on code is a target, not a measurement.
  • On Unreal's SemVer.cs the C# adapter saw 35 of 68 statements, and its report did not say so. Fixed, with planted tests.