AI Benchmarks
The Benchmark Mirage: Why Coding Agents Fail on Private Code
Real-SWE tests AI coding agents on private enterprise repos, revealing that top models solve under 40% of real-world production tasks.
Topic
Coding Agents
A collection of essays to explore, one idea at a time.
Essays
Newest first
AI
The Short Leash Era: Taming AI Coding Agents in 2026
Developers are moving away from fully autonomous AI to 'short leash' workflows and tools like ctx to manage coding agents and prevent architectural drift.