a builder's codex
codex · operators · Anton Osika · ins_generalists-over-specialists-ai-native

In an AI-native team, hire generalists with one deep dimension, not specialists

By Anton Osika · Co-founder and CEO, Lovable · 2026-04-28 · podcast · Anton Osika on Lovable, the last piece of software — Lenny's Podcast

Tier A · TL;DR
In an AI-native team, hire generalists with one deep dimension, not specialists

Claim

For a small AI-native product team, optimize hires for the breadth of skills a single person can hold, not for depth in one. Each hire should know architecture, design, product taste, and user research, then go deep on one of them. Lovable shipped to 10M ARR in two months with 15 people on this model — 12 of 18 write code part-time; the other 6 cover product, design, and ops, all shipping.

Mechanism

AI agents collapse the cost of execution in any single domain. The bottleneck shifts to whoever can decide what to build, design it, instrument it, talk to the user, and ship it without expensive coordination overhead. A team of specialists pays a high "handoff tax" between roles; a team of generalists with deep dimensions absorbs the handoffs internally and ships faster. The generalist has the cognitive surface area to use AI agents in any domain, where a specialist would need to bring in a colleague.

Conditions

Holds when:

Fails when:

Evidence

"If I'm putting together a product team today, I would really obsess about getting as many skill sets as possible for each person I hire."

Lovable structure: 12 of 18 write code part-time, the rest do product, design, ops — everyone shipping. 4M ARR in 4 weeks, 10M ARR in 2 months on this team shape.

— Anton Osika on Lenny's Podcast, 2026-04-28

Signals

Counter-evidence

Camille Fournier's platform-engineering work argues for specialist depth on infrastructure teams; the coordination cost is paid back in reliability. Asha Sharma's "polymath builder" frame agrees with Anton on consumer/product teams but warns that mid-size orgs need both: generalists for the loop, specialists for the spine. The pattern transfers cleanly to small AI-native product teams; less cleanly to scaled infrastructure.

Cross-references

Open the interactive view → View original source → Markdown source →