Coding agents discover routing with conditional progress guarantees
Source check: Source checked. Not independent scientific confirmation.
Finding: Human-directed coding agents synthesized Core-Graph-Anchored Routing, combining recoverable motion commitments with throughput heuristics. Eight paired simulated runs on a calibrated warehouse layout showed 58.9% higher throughput than the proprietary baseline. The source supplies a conditional safety and finite-progress proof; no physical fleet deployment is established.
Technology: Codex 5.6 Sol and Codex Astra coding agents, with Fable model assistance
Family Codex: profile · models page
Publication: Preprint · Test context: retrospective or benchmark. Catalog categories, not an evidence grade.
- AI contribution
- Coding agents proposed, implemented and diagnosed the new routing architecture from objectives and constraints, using simulation feedback. Models operate during discovery, not in the compiled planner.
- Test and result
- Six simulated warehouse layouts, eight paired two-hour runs per layout; one calibrated baseline and five uncalibrated baseline comparisons. An additional synthetic layout was evaluated without changing the policy.
- Limitation / still unknown
- Human steering and simulation-specific selection limit transfer. The theorem assumes valid core-graph separation/connectivity/vacancies, finite motions, recoverable initialization, service coverage and correctly implemented runtime checks. Some storage nodes require extra vacancies beyond the simplified theorem. We did not reproduce the proof or execution; no deployment claim.
Evidence sources and annotation context
Evidence context (descriptions follow the cited sources; test type is a catalog classification):
Evidence sources: https://arxiv.org/html/2610.25946v1 · https://arxiv.org/abs/2610.25946
Source checked · editorial details
The source page was checked against this entry. Independent confirmation is not assessed by this label.
Source type: paper. Paper: not linked in this catalog; this does not mean no paper exists.
Review date: 21 Nov 2026. Review scope: Actual v1 formulation, coding-agent Methods, warehouse Results, limitations and Appendix A proof read. Current history lists v1 October 19 and v2 October 20; represented result is already in v1, without an absolute-priority claim..
No follow-up recorded; this is not a claim that no follow-up exists. Follow-up review scope and date are not recorded. No source-supported relationships recorded.
Editorial assessment: Impact 2/5 · Validation 2/5 · Novelty 2/5. Impact and novelty are editorial judgments; validation follows the annotated test type where one exists. Not an evidence grade. Scoring criteria.