Claim
As code generation gets cheap, the bottleneck moves to "deciding what to write." Hire engineers who can take user feedback and ship a feature end-to-end without PM input. PM and engineering job descriptions are merging; designers and PMs both write code; the durable scarce skill is taste.
Mechanism
The previous PM-engineer split assumed code was expensive and product judgment was specialized. Both have inverted. Code is now table stakes; product judgment — knowing which feature to write, which user pain to chase, which polish to prioritize — is the gate. An engineer with taste outputs more user value per week than a PM-plus-engineer pair where the PM is gating.
Conditions
Holds when:
- The engineering team is genuinely close to users (Anthropic's "Twitter to ship" path; not all orgs have this).
- Product principles are written down (see related card) so engineers can decide from them.
Fails when:
- The work is integration-heavy with hard cross-team dependencies. Coordination work doesn't disappear because an engineer has taste.
- The audience requires careful upstream research (regulated industries, enterprise procurement). PMs still own that.
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
Anthropic's product org is ~30-40 PMs across Research PM, Cloud Developer Platform, Cloud Code, Enterprise, and Growth — small relative to engineering. The pattern is reinforced at OpenAI by Sherwin Wu (95% Codex authorship; ICs becoming tech leads of agent fleets).
Signals
- Engineers regularly read user feedback unprompted.
- PR descriptions reference user pain, not just technical change.
- New PM hires skew senior and ambiguous-problem-focused, not feature-shipping.
Counter-evidence
Anthropic and OpenAI hire from a deep talent pool with high prior selection on taste. Most companies do not have this hiring funnel. Camille Fournier's platform-engineering work argues the opposite for infrastructure teams: there, manager / specialist split still wins because the work is coordination-dense. Treat this as an applied pattern for product-shipping teams, not a universal hiring rule.
Cross-references
- Replace PRDs with weekly metrics readouts plus a written team-principles doc — the operating pattern that enables engineer-led shipping
- Taste is the scarce skill in an AI-native team — the broader thesis on what compounds in the AI era