NASA’s new dark energy space telescope can also detect killer asteroids
technologyreview.com·2h ago
Cloud & Infrastructure
The New Stack·Jun 26·5 min read
Your engineering org needs an AI slop registry
TL;DR
AI coding tools can lead to systematic errors in code that appear correct but are subtly flawed. To address this, teams are encouraged to create a structured AI slop registry to track and manage these issues.
✦ Why It Matters
Engineers can implement an AI slop registry to systematically track and reduce errors in AI-generated code.
Key Takeaways
- Create an invariant catalog to enforce coding standards automatically.
- Shift focus from code review to verifying intent before coding.
- Use historical review comments to identify potential invariants.
- Separate code generation from verification to improve quality control.
How It Works
Aviator's verification system uses two layers: user criteria for specific PRs and invariant criteria for universal rules. This separation allows for consistent enforcement of coding standards without relying on human judgment.