All publications

Code Enhancement System and Method Using Artificial Intelligence

2026·US patent application
generatemulti-model variantsmeasurebuild · benchmarkvalidatetests · equivalenceselect & shipevidence attachedoutcomes feed the next round (claim 4)the claimed measure–search–validate loop

Figure 1: The claimed enhancement loop. Candidate variants are generated (a), built and benchmarked against a baseline with correctness validation (b), and the selected winner is emitted with its evidence (c).

The patent application covers the system design behind Artemis: a code enhancement pipeline in which candidate program variants are generated by multiple AI models, evaluated by building and benchmarking on target hardware against a statistical baseline, validated for correctness, and only then surfaced as proposed changes.

The claims center on the closed loop — generation, measurement, validation, selection, and learning from outcomes — rather than on any single model, which is what keeps the method durable as underlying models improve.

Key results

  • US patent application 19/299,938
  • Claims cover the measure–search–validate loop, not a specific model
  • Filed 2026