top10.dev

Your real economic health isn't GDP or unemployment. It's whether your salary keeps pace with inflation, housing, and healthcare. Everything else is noise. → top10.dev

This essay hit 259 points on Hacker News because engineering teams have whispered about this for years. Good salaries, flat wages, shrinking dollars. Finally a name for what we've known.

GDP goes up. You get poorer. Meet the dead economy: headline metrics surge while median income degrades. Vital signs green, patient dying.

The quiet post-MCP pattern winning in production: 3-7 hand-curated tools + a code interpreter, not a 20-tool MCP server burning 10k tokens per turn. MCP is right for IDEs and desktop assistants. Wrong for the agent you ship. #ai #devtools top10.dev/story/post-mcp…

Your car is a data broker on wheels. The surveillance is relentless. Regulation is the only fix. → top10.dev

Here's the trap: all 25 failed equally. So there's no competitive advantage in privacy. No manufacturer has incentive to improve.

Mozilla tested 25 car brands. All 25 failed privacy. The worst category Mozilla has ever audited—worse than dating apps, mental-health apps, or smart speakers.

Your phone might be watching. Your car is definitely monetizing what it sees. That data feeds directly into your insurance calculation. → top10.dev

Here's what separates cars from phones: they don't just spy. They monetize what they spy on. Data gets sold to actuaries that price your risk.

Your car is more invasive than your phone. We obsess over Apple. The FTC just fined GM for selling driving data to insurers.

AI's code speed gains get eaten by review cycles. Licensing tightens, interpretability has caveats, C's undefined behavior persists. The friction's catching up. → top10.dev

Real innovation isn't models per week. It's integration velocity. If you ship faster than devs can absorb, you create noise, not progress. → top10.dev

Labs measure success by ship velocity. They don't see the cascade: fragmenting stacks, unmaintained code, unaudited security. When there's no time to evaluate, nobody maintains.

'Can we have the day off?' just hit 1,095 on HN. It's a joke, but devs aren't laughing—they're drowning in model releases.

An essay titled 'Can we have the day off?' hit 798 points on #hackernews this week. The comments are the real story: senior devs report shipping more, with smaller teams, fuller calendars. The #ai productivity dividend never reached labor — it got absorbed as more roadmap. #softwareengineering [link]

Anthropic's profit is amortization, not proof-of-model. The hard question: what revenue structure works at $1T capex scale? → top10.dev/item/simon-wil…

Building a startup on API tokens? Don't model on today's pricing. Your defensibility has to survive at $1, $10, or $100 per token. If it doesn't, you were never profitable.

$1 trillion per year. That's the token revenue AI labs need just to break even on $5-10T in capex. Most investors aren't pricing in the actual math.