We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
Woodward's playbook for Labs teams
On Lenny's Podcast, Google Labs VP Josh Woodward said good ideas "very rarely" come from design sprints. They tend to start as side projects by small groups of people who are obsessed with something . His team keeps a list of things that are "almost possible" and acts when one becomes possible. The technology still has to meet a real user problem, ideally one painful enough that people will pay .
- Early PMF signal. Watch people's eyes when you show them prototypes. Daily or monthly active users and day-7 retention don't help this early. Fall in love with the problem, not the product, because it often takes "three, four, five pivots" . Teams usually know a project is failing before their leaders do. He cited a Gemini PM who recommended not launching a feature after early data came back weak, and he thanked her publicly .
- Lab design. In large companies, labs usually fizzle within three to four years or never get their work commercialized . His advice: give the lab real independence, and decide up front whether its job is to graduate ideas into existing products or to create new categories . Zero-to-one work looks for those early reactions. Scaling from 10 to 100 means watching conversion funnels and cost per acquisition .
- Teams and pace. New Labs project teams have dropped from five to seven people to two or three . He still warns that the "everybody's a builder" idea is overhyped, because specialties matter. He tells new PMs to vibe-code "in service of becoming an exceptional extraordinary PM" . Planning runs on a rolling six months, and the best teams reach a meaningful milestone in 50–100 days .
Designing products for agents
Hiten Shah argues every product will get a second interface built for agents. Building one will become as normal as building a website or an API . Opendoor's Morgan Brown said the company recently made its home-selling and financing tools return JSON that agents can act on . Aakash Gupta makes a related point about building agents: Anthropic changed only the harness around the same model and moved its Terminal-Bench 2.0 score by up to 6 points. That is more than the gap between the top models .
AI PM pay and skills
Gupta cites Ankit Shukla's figures: AI PM roles pay 15–20% more than traditional PM roles in the US and Europe, and 30–50% more in India. The US median is $195K across 12,397 postings . Across 12,500 job descriptions, the top skill was choosing the right use case for AI. It ranked above RAG, agents and prompt engineering. In one example, Claude proposed a $420-a-month AWS stack, and two questions brought the shipped cost down to $110 .
Harder PM interviews
A laid-off PM with five-plus years of experience was rejected after a recruiter screen. The stated reasons were vague metrics and too little detail on customer research . Another commenter said recruiters now ask technical, product and behavioral questions, record the calls or use AI note-takers, and share the notes with hiring managers . One commenter's explanation: "Fewer open jobs, more people applying" . A practical tip from the thread: prepare short versions of each story covering the decision, your contribution and the result, then offer more detail . A candidate for Microsoft's CoreAI PM internship says one of three final rounds is "AI Fluency (Vibe Coding)." They understand it to involve building and testing an AI agent .
Wireframes versus AI prototypes
Practitioners describe building interactive prototypes in Claude Code from screenshots . Others use AI-generated HTML wireframes to compare options before building . The pushback: high-fidelity prototypes narrow exploration and make work look further along than it is. In one case, executives asked two months later why a product built from a Claude mockup hadn't sold yet, when it wasn't even half done . One commenter's advice is to make wireframes only as detailed as needed and treat them as a way to communicate, not a contract .
How AI is changing the PM job
A senior PM at a big tech company has used AI for over a year to write PRDs and do competitive research. They say AI can't get another team to deliver, so the role isn't going away soon . Another PM says AI makes them 10x faster, but they're now expected to work at "11x" . A third argues AI should improve discovery, not speed up feature-factory work .
Briefly
- Sachin Rekhi published an AI transformation playbook for product leaders. It covers building a "Compounding OS" so one person's AI gains spread to the whole team, an AI-fluency rubric, and rebuilding the product operating model from first principles instead of bolting AI onto the old one .
- John Cutler's TBM 444 warns about the "Case-by-Caser": a person who keeps demanding specific examples until a systemic problem shrinks to individual blame .
- On tracking competitors, one founder suggests logging only moves that would change your pricing, positioning, channel or roadmap decisions. They also suggest building your competitor list from your last ten lost deals .