Andrew Ambrosino is an engineer and product leader who helped found the financial infrastructure company Catch before joining OpenAI as a Member of Technical Staff, where he leads the Codex desktop app. He argues that as AI drives software implementation costs to zero, product work inverts, leaving curation and taste as a team’s primary value. This profile collects his notes on AI workflows, design, and strategy to show how technical roles are adapting to frontier models. — Lenny's Podcast, "OpenAI Codex lead on the new shape of product work" (2026).

Visual summary of operating lessons from Andrew Ambrosino.

Part 1: The Inversion of Product Work

  1. On the shifting cost structure: The product process is backwards now. Implementation is no longer the expensive part; as he puts it, "It's dare I say taste." — Reference: YouTube video of Andrew Ambrosino's 2026 interview on the new shape of product work.
  2. On the new bottleneck: When there may be ninety uncoordinated explorations of the same feature, the real work becomes curation: deciding what is good in them, what to fold into other features, and how to frame the result. — Reference: YouTube video of Andrew Ambrosino's 2026 interview on the new shape of product work.
  3. On evaluating quality: With effectively unlimited tokens, the scarce skill is taste: being able to tell signal from noise instead of shipping slop. — Reference: YouTube video of Andrew Ambrosino's 2026 interview on the new shape of product work.

Part 2: Taste and Design Judgment

  1. On the compiler for design: Design is harder to grade than code: "creating a loop where you can train the model on what's good design and what's bad design is just a little bit more tedious and onerous than, you know, does the code compile?" — Reference: Business Insider, "OpenAI's head of Codex says AI still can't get creative design right" (2026).

Part 3: The OpenAI Codex Desktop App

  1. On initial adoption: Nearly 100% of OpenAI employees, not just engineers, use Codex every week. — Reference: Lenny's Podcast, "OpenAI Codex lead on the new shape of product work" (2026).
  2. On creating a home base: The vision for the Codex app is a home base that coordinates work across ChatGPT, Codex and the tools people already use. — Reference: Lenny's Podcast, "OpenAI Codex lead on the new shape of product work" (2026).
  3. On timing the launch: Timing mattered: he believes the Codex app would have failed had it launched in November rather than February. — Reference: Lenny's Podcast, "OpenAI Codex lead on the new shape of product work" (2026).
  4. On Diverse use cases: People are already using the application for video editing, documents, spreadsheets, slide decks, production code, and internal launch-tracking sites. — Big Technology
  5. On User experience toggles: A toggle can change whether the app presents a code-intensive or less technical experience without changing the product's underlying capabilities. — Big Technology

Part 4: Redefining Team Dynamics and Roles

  1. On zone defense: He runs product like zone defense: if two product people are working too closely that is a bad signal. Spread out, cover the gaps, and let "taste makers" guide ideas from inception to product. — Reference: YouTube video of Andrew Ambrosino's 2026 interview on the new shape of product work.
  2. On collapsing roles: Roles overlap more than ever, but the goal is for everyone to have these skills while choosing what to go deep on. He wants engineers who are product-minded, so the team is not reviewing a flood of code for product coherence. — Reference: YouTube video of Andrew Ambrosino's 2026 interview on the new shape of product work.
  3. On the danger of eliminating PMs: Getting rid of the product role is, in his words, "a terrible idea": eliminating the concept of roles risks losing the idea that specialties have knowable best practices. — Reference: YouTube video of Andrew Ambrosino's 2026 interview on the new shape of product work.
  4. On Flexible team structures: Ambrosino uses a “zone defense” model for product management at OpenAI when everyone can build, while cautioning that eliminating roles entirely would be a mistake. — Lenny's Newsletter

Part 5: AI's Capability Thresholds

  1. On building ahead of the models: Some prototypes should wait for the models to catch up. Code that exists is not a reason to ship; it can be an artifact to test against future models. — Reference: YouTube video of Andrew Ambrosino's 2026 interview on the new shape of product work.
  2. On roadmapping with AI: At the current pace, long roadmaps break down. Top-down, year-long planning is "not going to work," so teams need taste makers steering a lot of parallel exploration. — Reference: YouTube video of Andrew Ambrosino's 2026 interview on the new shape of product work.
  3. On Over-refusal in general AI: General knowledge-work tools should avoid unnecessary constraints and refusals that prevent actions coding agents can already perform, while retaining permission checks. — Big Technology

Part 6: Catch and Financial Infrastructure

  1. On founding Catch: Before OpenAI he founded Catch (YC W19), serving as CTO/CPO from 2016 to 2023, building benefits and financial infrastructure for independent workers. — Reference: Andrew Ambrosino, personal site.
  2. On building from your own pain: He found the problem behind Catch while freelancing: "Setting up a benefits plan was a pain. You had to become an expert in the space, and even once you were, executing and getting the stuff you needed was pretty difficult." — Reference: TechCrunch, "Gig workers need health & benefits — Catch is their safety net" (2019).
  3. On a portable safety net: "We believe that everyone should have simple and holistic access to savings, investment and insurance." Catch set aside part of each paycheck for taxes, time off and retirement for people without an employer's benefits. — Reference: Catch press release, "Freelancers Begin Building a Financial Safety Net" (2018).
  4. On holding out for enough capital: Weeks from contingency plans to shut down, Catch got into Y Combinator, then turned down a $3 million offer from a Boston firm because the founders believed the company needed $5 million rather than a shoestring round. — Reference: The Boston Globe, "Catch's founders just wouldn't take no for an answer" (2020).

Part 7: The Limitations of AI in Creative Software

  1. On Explicit intent: "There are a lot of use cases where we can’t necessarily tell what you want" — Big Technology

Part 8: The Future of Workflow and Software

  1. On Tapping into memory: The new ChatGPT experience limits automatic reliance on user history but will access its memory if explicitly prompted. — Big Technology
  2. On The Superapp: Ambrosino described the new application as the start of a superapp meant to handle anything a user can do on a computer. — Big Technology
  3. On Competitive advantage: Moving after competitors can be useful when it lets a team observe how the product category should work before executing its own version. — Big Technology