Ex-Gemini PM says fewer than 75 people worldwide are qualified for model PM roles
aakashgupta · x · 2026-10-21
Madhu Guru, who led product for Gemini and is now a senior AI director at Meta, estimates fewer than 75 PMs globally have the depth for AI-native product work. He lists five core skills: vision and opinionated UX design, model intuition, going from pixels to evals to hill climbing, agentic flows (tools/context/safety guardrails), and deep user understanding — a rare but learnable skillset built by making things.
Aakash Gupta validated this against 46 open model PM postings: "pixels→evals→hill climbing" appears in 76% of JDs, UX vision 61%, model intuition 61%, agentic flows 59%, user understanding 50%. Only 7 of 46 require production experience — employers mainly want side projects. Median posted base pay is $345K (range $144.6K–$490K). His starter advice: pick a task your favorite model fails at, log failure modes as a seed eval set, write a behavior spec, then have Claude Code build and iterate a product that passes those evals.
More from coding & agent
- Hamel Husain: public benchmark quality is abysmal — broken graders, sloppy harnesses — HamelHusain · 2026-10-21
- Dev jokes Claude is never blocked from making mistakes, only from fixing them — cjimti · 2026-10-21
- Compile Your LLM Workflow Into Small Model Weights: A LoRA Hypothesis Worth Testing — arpit_bhayani · 2026-10-21
- Vercel Drives goes GA: developer says it replicates the Durable Objects programming model — steren · 2026-10-21
- S-Agent 2.0 uses a general coding agent for spatial reasoning, topping five benchmarks — ropedia-ai · 2026-10-21
- WebMCP goes live in ChatGPT Desktop as OpenAI pushes it as default for MCP Apps — prd_008 · 2026-10-21