Your coding agent doesn't always follow your rules
summary
Claims a concrete enforcement layer called Vector Harness that hooks into Claude hooks and runs deterministic checks whenever an agent reports completion. It uses a config file of test cases, calls back automatically, reports pass/fail, and retries the agent until checks pass. The prototype runs across models, lets cheaper models be used behind stronger guards, and is public as Vector Harness so teams can adopt the pattern rather than rebuild bespoke enforcement.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 1:11Deterministic verification layers
- 2:57Trust in coding agents
- 3:57Model capability vs reliability
- 5:23Enforcement patterns in companies
- 6:47Universal verification contract
- 7:53Industry verification systems
- 8:36TLDR and conclusion
related talks
16mBeyond the IDE: The Future of Developing Software With Agents
32mBuilding My Own Coding Agent
30mBeyond the Hype: Using AI Coding Agents Today
30mAgents Building Agents
32mEscape the Try-Again Loop: Verifying AI-Generated Code
22mAgents FC: Harness Agents Into a Winning Team
15mDevelop at Idea Velocity
23mBeyond Tests What to Verify in AI Generated Code