Addy Osmani: give your coding agent ways to check its own work
addyosmani · x · 2026-10-05
Addy Osmani outlines how to get high-quality output from coding agents by equipping them with self-verification. His recommendations: write end-to-end tests simulating real user flows as ground truth; use property-based testing to specify what must never happen and auto-generate thousands of adversarial cases; when replacing a system, diff old vs new on random inputs; and keep tests fast and deterministic—unreliable loops teach the agent to retry rather than actually fix the problem. As he puts it: examples say what should happen, properties say what must not.
More from coding & agent
- The 'Reverse Prompting Loop': A 3-Step Trick to Unlock Your AI Agent's Untapped 99% — Roger_M_Taylor · 2026-10-05
- Swarms Ships v16 'OverClock': 114 Commits Add DecisionModel, MCPDeployer, and TreeOfThoughts — KyeGomezB · 2026-10-05
- Developer Gets Proactive Agent Progress Updates in His Earpiece While Doing Housework via Dot — gregmushen · 2026-10-05
- Open Instinct: MIT-licensed personal-agent policy engine with allow/ask/deny outcomes — maritime_sh · 2026-10-05
- Skills vs MCP vs RAG vs Memory: a 4-part framework for agent knowledge — MaryamMiradi · 2026-10-05
- Model-written tests rejected a known-correct solution 77% of the time in agent pipeline test — deadatreides1 · 2026-10-05