EmbeddingGemma 2: an open, lightweight multimodal embedding model
deepmind.google·6d ago
AI Engineering
Hacker News·Aug 16·5 min read
ProofRun – a local verification receipt for AI coding agents
TL;DR
ProofRun is a cryptographic tool that verifies which tests were executed on your code, ensuring transparency in AI coding agents' claims about test results.
✦ Why It Matters
Implement ProofRun to verify test claims from AI coding agents and enhance code reliability today.
Key Takeaways
- Use ProofRun to bind test results to specific code states.
- Ensure accountability for AI coding agents with clear test statuses.
- Run tests offline without relying on AI for verification.
- Integrate ProofRun into your CI pipeline for reliable results.
How It Works
ProofRun operates by running tests in a subprocess and capturing their exit codes, which are then linked to a unique fingerprint of the code state. This fingerprint includes the current git commit and a hash of any uncommitted changes.
If any code is modified, the test results are automatically marked as STALE, ensuring that developers are always aware of the validity of the test results.