Andrew Ambrosino

Sources#

• OpenAI Codex lead on the new shape of product work

Summary#

Andrew Ambrosino leads product and engineering for the Codex desktop app at OpenAI — the surface that grew from an internal developer tool into the app OpenAI reports ~90% of its entire company (not just engineers) uses. A self-described designer → engineer → product manager → founder generalist, he is the primary voice in the June 2026 Lenny's Podcast interview that supplies this wiki's OpenAI-side, practitioner account of the AI-native product shift. His throughline: implementation is no longer the expensive part of building software — taste and curation are (Implementation Abundance Inverts Product Work).

Evidence note. practitioner-opinion — a talk by a frontier-lab product leader, not measured research. His usage figures (90% company-wide, 5M+ weekly actives, 6× growth since January) are first-party OpenAI claims, unverified here.

Career shape (why he's a useful witness)#

Ambrosino embodies the role fluidity he describes. His LinkedIn reads "engineer, designer, product manager, founder"; he says "if you average up all of the things somebody does… that's your role." Before OpenAI he was a startup founder for years in a heavily regulated space ("ended up selling the company for parts"), then worked on AI tools at another startup in a "locked-down regulated industry" — a decade-plus of what felt like "constant failure" before the Codex app lined up "skill set, passion, point in the market." He hates repetition ("sucks to be a founder when you hate repeating yourself") and frames his own job as the ultimate case of building the tool that reshapes his role.

What he articulates (his contributions to this corpus)#

• Implementation Abundance Inverts Product Work — "the implementation is actually not the expensive part anymore… it's dare I say taste"; 90 uncoordinated teams building the same needed feature, so the work becomes curation.
• Polish No Longer Signals Readiness — the medium used to encode process-stage; now a polished prototype can be early exploration, and over-anchoring on it is the trap.
• Why AI Lags at Design — design is hard to grade, sat outside the AI-research flywheel, rewards novelty over pattern-matching, and hides a design↔code abstraction layer.
• Role Averaging, Not Role Elimination — welcomes "not your lane" going away, but warns that eliminating roles eliminates specialties with knowable best practices ("getting rid of the product role is a terrible idea"); "zone defense" for product coverage.
• Build for the Next Model — the Codex app released in February would have "absolutely failed" in November with the same shape; only the models changed. Operator → Atlas → Codex is "fundamentally the same feature" re-released at different intelligence.

Connections#

• OpenAI — employer; leads the Codex desktop-app effort
• Codex — the product he owns; his account is the wiki's richest description of the app (vs. the CLI)
• Implementation Abundance Inverts Product Work — his headline thesis
• Build for the Next Model — his cleanest retrospective ("only the models changed between November and February")
• Role Averaging, Not Role Elimination — he both describes and personifies it
• Dan Carey — the Anthropic-side product voice whose "prototype-over-PRD" view Ambrosino explicitly pushes back on (pick the medium; PRDs aren't dead)
• Cat Wu / Boris Cherny — Anthropic counterparts on role convergence; Ambrosino is the OpenAI-side witness with a sharper caution about over-collapsing roles
• The Three Loops of AI-Native Building — Andrew Ng published a three-loop taxonomy the same week Ambrosino declared "loops are so last week"; compatible once harness-loops (absorbed by capability) are separated from product-loops (structural, and unshortened)
• Context Advantage, Not Taste — supplies what Ambrosino's "taste is the bottleneck" leaves undefined: Ng reads taste as an information asymmetry, which makes the new bottleneck perishable

Sources#

• OpenAI Codex lead on the new shape of product work — Lenny's Podcast, published 2026-06-28 (guest; host Lenny Rachitsky)