Niranjan Kulkarni@niranjan_k_

AI coding agents can implement software changes quickly, but implementation alone does not prove that a change is correct. Developers still need to inspect repositories, identify affected areas, run tests, interpret failures, and communicate fixes back to the coding agent. This creates a fragmented development loop. RepoMind solves this by acting as a verification layer around AI-assisted software development. A developer starts with a natural-language change request. RepoMind analyzes the controlled repository, identifies likely affected files, assesses risk, determines the relevant test scope, and generates a structured implementation task. RepoMind then independently verifies the implementation by running the repository's real build and test commands. It captures command output, exit codes, execution time, and the final verification result. If verification fails, RepoMind converts the failure into actionable FEEDBACK.md containing the failing test, expected and received values, source location, suggested correction, and raw output. The corrected implementation can then be verified again, creating a closed loop: Request → Analyze → Impact → Task → Implement → Verify → Feedback → Fix → Verify Our demonstration intentionally starts with a regression where POST /tasks returns HTTP 200 while the tests require HTTP 201. RepoMind detects the failure, produces targeted feedback, the implementation is corrected, and verification ultimately reaches a passing result with the build successful and all 12 tests passing. The target users are developers and engineering teams using AI coding agents who want stronger evidence that generated changes actually work. RepoMind is deliberately lightweight, using Next.js, Node.js, Git, and filesystem-based state without requiring additional infrastructure. What makes RepoMind different is its focus on evidence-driven completion: an AI-generated implementation is not treated as complete merely because code was produced. It must survive the repository's actual build and tests and produce a verifiable result.

🌎 Online Hackathon | 💻 Online Build: Sept 26 - Oct 5, 2026 A virtual, global hackathon. Build from anywhere in the world. Dark Factory is the official hackathon of WeAreDevelopers World Congress North America, presented by BAND, but you don't need to be in San Jose to take part. For one week, you build a software factory in BAND Desktop: a band of coding agents that plans work, implements it and checks its own results. 🎟️ Bonus for sign-ups: you are eligible for a free ticket to WeAreDevelopers World Congress North America in San Jose (Sep 23–25). If you do go, stop by the BAND booth: a surprise is waiting for you.

💻 Online Hackathon | ⏳ Build: 48 hours | 📅 September 25–27, 2026 The official IBM Bob 2.0 hackathon. Build what's next in AI-assisted development, alongside the developers and experts shaping it. ⭐ Get help from experts throughout the event 🤝 Join a team or build solo 💰 Prize pool: $12,000 🧑🏻‍💻 Register before the Kick-Off Stream to start building with Bob