The AI and Quantum Paper

The design sweeps that found both machines' operating points were driven by computational methods, documented and deposited so the search itself can be rerun.

The methods behind the design points

A fusion design point is the output of a large search over plasma, geometry, and field parameters. The AI and quantum paper documents the computational methods Kronos used to run those sweeps and to locate the breeder and burner operating points. Publishing the method, not just the winning configuration, is what lets a reader ask whether the search was fair and whether the reported point is really where the design lives.

Why the search matters as much as the answer

Any optimizer can be steered, consciously or not, toward a flattering result. The defense against that is to expose the objective, the parameter ranges, and the code. When the search is public, a skeptic can widen a range, change a constraint, and see whether the design point survives. That is a stronger form of honesty than reporting a single number.

Reproducibility discipline

• The deposit follows a cold-reproduction runbook across multiple environments; see the cold-reproduction runbook.
• Results are checked at two tiers — exact byte match where possible, numerical tolerance where not; see two-tier reproduction.
• The methods feed the public 3D model and the design tooling used elsewhere in the program.

The AI and quantum paper is where the open-science posture reaches the tooling itself: not only the results are public, but the machinery that produced them. The deposit and DOI (AI and quantum) are on the DOI map.

If the search cannot be rerun, the design point is a claim about our code rather than about the machine. So we release the search.