How Figma’s PMs use AI in their daily work

How Figma’s PMs use AI in their daily work

Lenny’s Newsletter presents growth product manager Ezra Mechaber describing how Figma PMs use AI in their work. Mechaber says he and fellow PMs turned his personal Claude Code setup into PM OS, a shared repo of 31 skills that the description says every Figma PM uses. The episode preview describes the repo as used by 60+ PMs. The examples range from a “chief of staff” setup and a skill Mechaber runs to manage Slack, to bug bashing with Codex. The description also highlights AutoXP, called an automated growth PM, which it says found a “huge win” on its first experiment and runs experiments with one human checkpoint. It offers practical examples of how a team might organize AI-assisted work; the provided page does not establish how broadly the workflows improve results or what the experiment measured. The episode description frames PM OS as a repo, not a product, and points to how the skills were built and refined. That makes the account about a shared working setup as well as individual tools. What was the first AutoXP experiment, and what did its human checkpoint review?

Publication: f6c4760f-d5e2-4b1d-8bda-387b7cca2ff4

attribution — uncertain: The page text is the host's episode description, written by Claire Vo in the first person on her How I AI show. The draft never names Vo. It also says 'Mechaber says' he and fellow PMs built PM OS, but the evidence is Vo's narration, not a quote from Mechaber. The claims list correctly says 'the episode description says.' Smallest repair: credit Claire Vo's How I AI episode on Lenny's Newsletter, and change 'Mechaber says' to 'the episode description says.'

In this episode, Ezra walks me through his favorite skills

summary — uncertain: The summary is faithful and appropriately hedged. However, central terms are left unexplained for a reader without specialist background: repo, skills (in the Claude Code sense), Claude Code, Codex, bug bashing, growth PM, and human checkpoint. Without these, it is hard to grasp what PM OS actually is. Per EM style, these should keep their exact wording and be explained through term-library tooltips, adding entries where they are missing. The title and the question are neutral and follow from the evidence.

turned his personal Claude Code setup into PM OS, a shared repo of 31 skills that every PM at Figma now uses

date — pass: The page shows Nov 02, 2026. This matches originalPublished (2026-11-02, from the publisher page) and the feed date.

Claire Vo Nov 02, 2026

support — pass: Each claim is grounded in the description. The 'huge win' is attributed rather than asserted as fact. The limitations correctly note there is no transcript, the experiment is unspecified, and adoption is only loosely described. The implied value ('practical examples of how a team might organize AI-assisted work') is modest and conditional. The account of building and refining the skills is supported by the chapter list. No invented benefits or costs.

How AutoXP runs experiments with one human checkpoint

media — pass: The selected media is the verified original article page, the direct source of the evidence. The verified YouTube video is a reasonable alternative, but the article choice is appropriate.