---
id: ins_taste-as-scarce-skill
operator: Cat Wu
operator_role: Head of Product, Claude Code + Co-work, Anthropic
source_url: https://www.youtube.com/watch?v=PplmzlgE0kg
source_type: podcast
source_title: How Anthropic's product team moves faster than anyone else — Lenny's Podcast
source_date: 2026-04-27
captured_date: 2026-05-01
domain: [product, ai-native, leadership]
lifecycle: [strategy-bets, hiring-team-design]
maturity: applied
artifact_class: framework
score: { originality: 3, specificity: 3, evidence: 4, transferability: 5, source: 5 }
tier: B
related: [ins_engineers-with-product-taste]
raw_ref: raw/podcasts/cat-wu--anthropic-product-team--2026-04-27.md
---

# Taste is the scarce skill in an AI-native team

## Claim
With code generation cheap, the new high-leverage job is deciding what to write, taste. Taste comes from any background; an engineering background helps in the short term because it gives you a sense of how hard something *should* be, which feeds prioritization. The premium on taste rises the more capable models become.

## Mechanism
When the cost of producing artifacts collapses, the constraint shifts to selection: which artifact, for whom, in what order. That selection work is taste, accumulated judgment about audience, craft, market, and tradeoffs. It is not separable from execution but is increasingly the larger share of the value.

## Conditions
Holds when the team operates in domains where outputs are infinite-but-rankable (copy variants, design directions, feature options). Fails in domains where outputs are infinite-and-unrankable (open-ended research where nobody knows what good looks like), there, taste is no advantage because no one has it.

## Evidence
> "As code becomes much cheaper to write, the thing that becomes more valuable is deciding what to write."

· Cat Wu on Lenny's Podcast, 2026-04-27

Krithika Shankarraman echoes this directly: "Taste = differentiator in the AI era... companies that distinguish themselves show their craft." Jenny Wen at Anthropic Design adds the heretical caveat that AI taste will get good, humans hold the *decision* seat, not the unique-creative-judgment seat.

## Signals
- Hiring loops weight portfolio and prior judgment over credentials.
- Senior leaders dogfood the product personally to keep their taste calibrated.
- The team can articulate, in writing, what good looks like for each output class.

## Counter-evidence
"Taste" is easily a vibes word. Without a written quality bar (Cat's team-principles, Krithika's brand consistency, April Dunford's positioning), "we hire for taste" becomes "we hire people who agree with the loudest senior person." Codify the bar before claiming taste as a differentiator.

## Cross-references
- `ins_engineers-with-product-taste`, applied to a specific role