Hill-climbing Skills: How to Improve Agents Without Touching the Model
Official Schedule Context
- Date/time: 2026-06-29 · 4:30pm-5:30pm
- Track/room: Workshops Day 1 · Track 4
- Speaker(s): Shubhankar Srivastava
- Session type/status: workshop · confirmed
Official Description
Agent Capability is now highly dependent on the markdown files read at runtime -- skills.This
workshop treats skills as a first-class optimization surface. We borrow the concept of autoresearch
(from Karpathy) and apply it to the skills your agents already read. You'll see how we at
Browserbase did the same for browser agents, enabling our customers to scale the coverage of their
browser agents while improving performance(2x faster runs) and optimizing for token spend(upto 10x
cheaper).You'll leave with a working http://SKILL.md you generated through an auto-research loop,
and a mental model for when skill optimization beats fine-tuning or prompt engineering.
Related YouTube Video
No related AI Engineer channel video found yet.
Transcript Status
No official session recording transcript was found by exact title match on the AI Engineer YouTube channel during this run.
People
Notes
- Pending transcript synthesis when an official recording or confirmed matching video is available.