a builder's codex
codex · operators · Sherwin Wu · ins_top-performers-benefit-disproportionately

AI tools widen the spread between top and bottom performers — invest in top performers

By Sherwin Wu · Head of Engineering, OpenAI API and Developer Platform · 2026-04-28 · podcast · Sherwin Wu — Codex inside OpenAI, engineers as managers — Lenny's Podcast

Tier A · TL;DR
AI tools widen the spread between top and bottom performers — invest in top performers

Claim

The standard management orthodoxy is "raise the floor" — bring the bottom performers up. AI tools invert this: top performers benefit disproportionately, the spread widens, and management leverage now comes from investing in the top, not the floor. At OpenAI, heavy Codex users open 70% more PRs and the gap is widening.

Mechanism

AI tools amplify whatever judgment, taste, and ambition the user brings. A top performer with Codex can run 10–20 agent threads in parallel; a bottom performer with the same tool struggles to ship a single one because they don't know what to ask. The tool isn't the differentiator; the user's pre-existing skill is. As tools improve, the multiplier widens.

Conditions

Holds when:

Fails when:

Evidence

"Codex really empowers top performers to be a lot more productive... you see a broader spread in team productivity."

Heavy Codex users at OpenAI open 70% more PRs than average users. The spread is widening with each model release.

"Spend more time with top performers, not bottom performers."

— Sherwin Wu on Lenny's Podcast, 2026-04-28

Signals

Counter-evidence

Asha Sharma's "polymath builder" thesis and Anton Osika's generalist hiring both argue for distributed capability across the team rather than concentration in top performers. The two views can coexist: hire generalists with depth, then accept that some of those generalists will compound faster than others. Don't artificially flatten the spread.

Cross-references

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