About
Works on My Agent is about the engineering evidence.
Coding agents can produce impressive results quickly. Works on My Agent is interested in what happens after that: whether the resulting software builds, survives review, fits the architecture, remains understandable, and is something we would actually maintain.
Real projects, not isolated demos
The experiments follow real applications as they evolve. Prompts, repository changes, tests, failures, corrections, and review findings are part of the evidence rather than details hidden behind the final result.
Series, not verdicts
Tools are explored over time, not judged from a single benchmark or toy task. Codex: Zero to Hero is the first series; future series may explore other coding agents, tools, comparisons, and engineering approaches as real questions emerge.
Human engineering judgment
Generated code and passing tests are evidence, not the finish line. Review also considers architecture, maintainability, scope, trade-offs, and whether the resulting change is something we would actually merge and maintain.
For questions, corrections, experiment ideas, or related discussion:but@worksonmyagent.com.