Nick Saraev
AI educator and automation builder. Runs “Maker School” community. Focuses on practical Claude Code skill development, prompt optimization, and applying AutoResearch methodology to real-world business outputs.
Channels
- YouTube: Nick Saraev — Claude Code tutorials, automation, prompt engineering
- Course: Claude Code full course (4 hours, linked from YouTube)
- Community: Maker School on Skool — “get customer #1 guaranteed”
Content in This Wiki
- Stop Fixing Your Claude Skills. Autoresearch Does It For You — Practical walkthrough of applying AutoResearch to self-improve Claude Code skills with binary evals. Demonstrated: diagram generator from 32/40 to 39/40, website from 1100ms to 67ms load time.
Key Ideas
- AutoResearch generalizes beyond ML: Apply objective metrics + automated measurement + a variable to any process — skills, websites, cold email, landing pages
- Binary evals are king: Yes/no questions give the tightest signal for prompt optimization; Likert scales compound variance
- The research log is the real asset: Pass it to a future smarter model to continue where predecessors left off
- Skills are just prompts, and prompts are inherently noisy — run them many times and measure the distribution, not a single output
See Also
- AutoResearch and Evals
- Andrej Karpathy — AutoResearch originator
- Claude Code
- Source: AutoResearch