Most of what’s written about AI agents sounds great in theory — until you try to make them work in production. The seductive ideas (multi-agent orchestration, RAG, prompt stacking) often collapse under real-world constraints. Why? Because they optimize for the wrong thing. In this talk, Nik Pash shares hard-won lessons from building large-scale coding agents at Cline — what failed, what survived, and why the next leap forward won’t come from clever scaffolds, but from evals and environments that truly measure and improve reasoning. Attendees will walk away with a clearer sense of what actually drives progress — and what’s just noise.

https://www.linkedin.com/in/nikpash

AI Engineer

Frontier AI models have outgrown complex scaffolding like RAG and tool-calling frameworks. Model capability now matters more than clever engineering tricks. The real bottleneck is creating quality benchmarks and RL environments from real-world coding data. Cline built an automated RL environment factory that converts user tasks into training environments in under 20 minutes. They're launching ClineBench, an open-source benchmark built from actual engineering work, to give the community a substrate for improving models beyond synthetic coding puzzles.

Hard Won Lessons from Building Effective AI Coding Agents – Nik Pash, Cline