coco
5.7 KiB
title, source, author, published, created, description, tags
| title | source | author | published | created | description | tags | ||
|---|---|---|---|---|---|---|---|---|
| Chief Product Officer in a Box: Introducing The AI PM OS | https://x.com/nurijanian/status/2033111195121865207 |
|
2026-03-15 | 2026-04-26 | I’ve been tracking Cursor use for PM work across Reddit, X, and a few video sources. The same handful of patterns keep showing up.Dennis Yan... |
|
I’ve been tracking Cursor use for PM work across Reddit, X, and a few video sources. The same handful of patterns keep showing up.
Dennis Yang’s workflow at Chime: PRDs written in markdown inside Cursor, published to Confluence via an MCP server connection, with Jira epics auto-generated from the spec. Weekly status reports drafted in a few minutes from the same source files. He described it as reducing his documentation overhead significantly, which is believable: if your spec is already in the same environment as your AI, you stop copy-pasting between tabs.
Zevi Arnovitz at Meta: No technical background. He runs Claude for planning, Gemini for UI generation, and then has both models review each other’s output before shipping. His engineers now ask him to teach them his process. He built the peer review loop from scratch through trial and error, and the output is good enough that his team follows his lead.
https://www.youtube.com/watch?v=1em64iUFt3U
Alan Wright: He queried PostHog directly inside Cursor, had the AI diagnose what the data showed, produce a markdown summary of the finding, push it to Notion, and open a Linear ticket from it. Something that used to take hours now takes a few minutes. He wrote it up on his site. Most PMs reading his post probably thought: “I’d love to do that, but I don’t know where to start.”
The scale of what’s happening: This spring, Y Combinator named “Cursor for Product Management” as a Request for Startups category, meaning they’re actively looking to fund companies building in this space.
The part that none of the videos mention
None of these write-ups mention what it actually takes to get there.
Yang, Arnovitz, and Wright all built their setups from scratch, through trial and error, over weeks.
Each spent real time on the plumbing before they could do any of the interesting work.
Watching a 45-minute podcast episode and then going off to build what Dennis built takes days, weeks — if you know what you’re doing.
Cursor opens with an empty file and no direction for PM work; you have to know what context to load, which workflow to follow, and what prompts to use at each step.
What many forward-thinking PMs are still missing is not even where to start with Cursor or Claude Code — it is how to get to a working system fast enough.
AI PM OS: the system, already built
AI PM OS started as my own attempt to solve this.
It’s a pre-configured operating system for PM work inside Cursor, mapping directly to what Yang, Arnovitz, and Wright each had to build by hand, enriched by all the good PM juice that I had inside my Notion-based Complete PM Bundle.
The setup takes about five minutes:
Run /start**.** The AI asks you questions about your company, product, team, goals, and constraints and fills in five context files as you answer. Every response after that draws on those files. This is what Dennis Yang was describing when he talked about the AI that “knows your product.” In PM OS, that knowledge lives in COMPANY.md, PRODUCTS.md, GOALS.md, TEAM.md, and CONSTRAINTS.md. You fill them once, and every workflow uses them automatically.
190+ skills auto-trigger. You don’t have to remember to prompt or even bring your own prompt library. The skills activate based on what you’re doing. This is the prompt library that I know others maintain as individual files, already pre-installed and wired up.
7 structured workflows. These cover the PM work that actually repeats: strategy, research, decisions, stakeholder communication, opportunity framing, assumption testing, and meeting prep. Each one walks you through the work in the right sequence, rather than leaving you to design the process from scratch.
The Lenny Podcast MCP & my favourite external skills are pre-installed. The knowledge source most PMs reference is already connected. You can query specific episodes, pull frameworks, and reference content without leaving your workflow.
You also have the Shaping skills and the Frontend-Slides skills pre-installed ready to use. I love them, and I can’t live without them now!
The difference between PM OS and what others have built is that you can have it today - you don’t need months to experiment, you don’t need to populate it with frameworks, techniques, skills, examples and all that good stuff I’ve been sharing for years.
You just open Cursor, run /start, and you’re in.