I'm building AgentProof around a question that keeps coming up in AI-assisted development: Who independently checks whether the result matches what was actually requested? The direction: Clear requirements. Human approval. Independent verification. Evidence-backed defects. Still early. Building and testing one piece at a time. #BuildInPublic #AICoding
@ganesh2412 The useful part is the audit trail: who signed off, not just who typed the prompt.
That's an important distinction! 😄 An AI agent can write the code, generate tests and report success. But who approved the requirement, who verified the evidence, and who accepted the remaining risks? That's the accountability I'm trying to build into AgentProof. Not just a green checkmark, but a traceable record connecting requirements, approvals, test results and defects. Because when something breaks in production, “the AI said it was done” isn't a very useful audit trail! 😂🐛 Appreciate you highlighting this!
A principle I'm adopting while building AgentProof: AI can suggest missing acceptance criteria. It should not silently turn those suggestions into approved requirements. That's a product decision. The verifier should identify uncertainty, not invent certainty. #SoftwareTesting #ProductDevelopment
