I listen to 8-10 hours of AI podcasts weekly.

Here's what the smartest people in the field are saying about jobs and the market.

Consistent themes from leaders at Anthropic, Google, Meta, and top startups:

The biggest hiring gap isn't model builders — it's evaluation and deployment engineers. Everyone can train a model. Few can test it properly and get it to production reliably.

Domain expertise is becoming MORE valuable, not less. As models commoditize, the differentiator is understanding the problem domain deeply.

AI safety hiring will 5x in the next 3 years. Every major company is building safety teams from scratch.

Remote work is permanent for AI roles. The talent is too scarce to restrict by geography.

The "AI engineer" role is splitting into 5-6 specialized roles: LLM application engineer, ML platform engineer, AI evaluation engineer, AI agent developer, and domain-specific ML engineer.

Junior roles are getting harder to land (AI tools handle junior-level tasks). Senior and specialized roles are getting easier to land (demand massively exceeds supply).

The meta-insight: the people building AI are telling you what they need. Listen. If every AI leader says "we need evaluation engineers" — become an evaluation engineer.

The market tells you exactly what to learn. You just have to pay attention.