This week the AI internet gave us a delayed model roasted into oblivion, fresh bubble-pop doomposting, and the usual chatbot refusal comedy. A field guide to the cope.
AI would rather invent a plausible-sounding lie than admit it has no idea. This week the internet had enough. A wry autopsy of machine confidence.
The prompt was about not needing an installer. The agent responded by deleting the entire project. A love letter to backups, and a warning about terminal agents.
A good automation runs while you sleep and doesn’t re-do work it already did. Here’s how to build set-and-forget loops in Cowork that stay cheap.
Every message re-sends your entire growing transcript. Subagents let the expensive reading happen somewhere else and hand you back just the answer.
The difference between a $5 workflow and a $500 one is usually context you keep re-sending. Here’s how to build recurring Claude Code systems that stay cheap.
Being contrarian in 2026 might mean being bullish. Here’s the strongest version of the bull case — Ives’s “supercycle,” the Jevons paradox, and why the capex-vs-revenue gap may be a timing problem.
IonQ trades near 100x sales. Rigetti above 800x. Insiders have sold ~$931M net. Here’s what the credible voices — Huang, Aaronson — say about whether the tech is close.
When Michael Burry deregisters his fund to short Nvidia full-time and Ray Dalio starts citing 1929, the “AI bubble” talk has stopped being fringe. The voices and the data.
Three of the biggest IPOs in history are landing in one quarter. The valuations are staggering. So are the operating losses. Not financial advice — just eyes open.