The product manager's identity crisis is the story nobody is writing about, even though it's happening at every tech company right now. Engineering found its AI-native workflow months ago. PM is still figuring out whether to use Claude to write better PRDs or to become something else entirely. Meanwhile, a Chinese model just showed up that has Vercel's CEO doing a double-take. It's a Monday.
01
Product managers are splitting into two species
Madhu Guru put a sharp frame on something a lot of builders are feeling but not naming: AI gave engineers a clear power-up (coding agents that multiply individual output), but nobody told PMs what their equivalent is. The result is a fork. One type of PM uses AI to produce more of the same: faster PRDs, more strategy decks, cleaner specs. More output, not more judgment. The other type uses AI to reach across the product lifecycle and do things that used to require a full team. ---
Why it matters: If you're a PM still optimizing for documents-per-sprint, you're training your company to see you as a content generator. The PMs who figure out how to ship, prototype, and validate with AI rather than just write about it are going to make the first group look very expensive.
A Chinese coding model just surprised Vercel's CEO
Vercel CEO Guillermo Rauch called GLM-5.2, a model from the Chinese AI lab ZhipuAI, "genuinely impressive, almost shocking" at coding. Rauch doesn't hand out compliments like that. He said it "changes things." ---
Why it matters: The American lab assumption, that coding dominance belongs to OpenAI, Anthropic, and Google, keeps getting tested. If GLM-5.2 holds up under scrutiny, developers suddenly have a credible open alternative that doesn't come with US pricing or API waitlists. That's worth watching this week.
An OpenAI engineer hints the Codex App was built on mediocre front-end models
Thibault Sottiaux, who worked on the Codex App at OpenAI, posted that the team built it with models that were "okayish at front-end" and that meaningfully better front-end capabilities are coming. His framing: "That day will be something." ---
Why it matters: This is a useful calibration if you've been disappointed by AI-generated UI code. What you've seen so far isn't the ceiling. If OpenAI's own engineers are calling the current models mediocre at front-end and promising a step change, the current tools are a floor, not a finished product.
Replit CEO on the internet as accidental AI training data
Replit CEO Amjad Masad posted a two-sentence reflection: "We posted for twenty years, thinking we were talking to each other. Then the transformer came online, and the network read what we'd written, and became itself." A piece of writing worth sitting with. No product announcement, no call to action. Just a true thing, stated cleanly. ---
A developer trolled Gmail and Outlook with an agent-addressed bug report
Nan Yu posted a one-liner directed at developers at Outlook and Gmail, telling them they could point their agents at his tweet and "they will fix it for you." No visible context for what bug he was reporting. This is too thin to do much with, but the framing is real: people are already half-seriously imagining bug reports written for AI agents to read and act on, not humans. That's a small cultural shift worth noticing.