TB: Help Claude Help Us (causalinf)
Core Thesis
Domain expertise is the multiplier that transforms AI from a curiosity into a force multiplier. The most dangerous use of AI is not the expert with a rocket on their back — it is the novice using AI to fake expertise they haven't built, on problems they can't evaluate.
Key Takeaways
Expertise × AI = rocket
- "When you know your domain, the AI agent is like a rocket strapped to your back."
- The expert uses AI to execute faster, explore more, and automate the tedious — while evaluating output from a position of genuine understanding.
The thin ice problem
- "The thinnest of ice really comes when you don't know the domain very well and you're using AI to teach it to you during the actual coding of the project itself."
- AI in the hands of a non-expert is doubly dangerous: it produces plausible-looking output and the user cannot evaluate it.
- This creates a false sense of progress — velocity without competence.
The implication
- The premium on domain expertise is increasing, not decreasing, with AI.
- Learning your domain before using AI to operate in it is not optional — it determines whether AI amplifies or misleads.
Mental Models
- Compounding Returns — domain knowledge compounds over time; AI makes that compounding dramatically faster for those who have it
- Second-Order Thinking — the second-order effect of AI-without-expertise is institutionalizing confident ignorance
See also
- Agentic Engineering (Willison) — the practitioner's view of agentic accountability
- Tyler Cowen on AI and Academia — the institutional version of this failure mode
Source
Causal Inference (Scott Cunningham), Substack, 2026 Original article