March 3, 2026
Clout vs code: who broke the app?
My take on vibe coding for PMs
Stop the vibe-code flex: PMs told to quit coding for clout; engineers rage-clean the mess
TLDR: Daniel McKinnon says product managers shouldn’t chase clout by shipping code at big-company scale, and should fix priorities or build better AI tests instead. Comments split: some hail technical PMs, others roast “vibe coding” and metric gaming, with engineers complaining they’re left to mop up the mess.
Product managers (PMs) at big companies, put down the keyboard. Daniel D. McKinnon’s spicy memo says stop “vibe coding” for clout and start fixing prioritization—the actual job. His take: PMs hacking features move at junior speed for senior money, rack up tech debt, and burn time from top engineers reviewing pet changes. He still wants PMs to code, but for showing ideas, understanding systems, running real experiments, and, yes, “fun”—and points them toward building AI tests called evals (Build EVALS!!!).
The comments lit up. One camp cheers the rise of technical PMs: aurareturn calls them “the future most valuable.” Another camp is savage: sublinear wonders “how bad” a PM’s communication must be if code is required to explain. Engineers pile on: 650 blasts metric gaming—lines of code and ticket-count bragging—saying they’re stuck reviewing mystery diffs and mopping up production messes (classic Goodhart’s law: when a measure becomes a target, it stops measuring). Meanwhile, Ronsenshi clowns the constant “Fun!!!!!” hype, meme-ifying PMs as “LinkedIn flex first, impact later.” Ambicapter side-eyes the whole brawl and says the evals guide is the real story. Verdict: vibes are cute; impact wins. Readers laughed, winced, and bookmarked the evals link for receipts. Seriously.
Key Points
- •The author advises PMs not to spend time landing production code at large-company (Meta) scale.
- •He argues PMs should fix prioritization systems instead of bypassing them to ship features.
- •PM-authored production diffs can be inefficient, create technical debt, and risk breaking production.
- •Coding can be useful for PMs to communicate ideas, understand systems, run realistic experiments, and leverage unique expertise.
- •For AI work, the author recommends PMs focus on building evaluation suites for generative AI.