Weekly
The Sunday review
Every Sunday: the week's posts in one place, what connects them, and a few public reads I found worth your time. By email or RSS. Get it by email or via RSS.
2026
- Weekly Review — Oct 12–18, 2026
What sits outside the frame: unread messages, a series closer, how wings lift, ads in AI answers, and the human an agent must ask.
- Weekly Review — Oct 5–11, 2026
Who carries the consequence: one-way doors, one-way ratings, a swarm's summary, GLM 5.3's reasoning default, agentic adoption, and ice.
- Weekly Review — Sep 28–Oct 4, 2026
Commitment and alternatives, an AI-run brand canvas, an 88-hour agent run, invisible effort and the blue sky: a week about what others can and can't see.
- Weekly Review — Sep 21–27, 2026
The gap between minds: the feedback exchange rate, why a fridge only moves heat, when a model becomes an org, and the expert's blind spot.
- Weekly Review — Sep 14–20, 2026
Agent authorization is mostly OAuth, suction cups run on air pressure, empathy at volume is structural, and the agentic web splits in two.
- Weekly Review — Sep 7–13, 2026
The agentic web gets a terms layer in the 402, the prompt-to-context-to-memory arc built on AWS, which of your work is actually verifiable, and one post about Germany.
- Weekly Review — Aug 31–Sep 6, 2026
A pipeline that grades its own homework, a gate blind to its own defect, an agent paid to lean, and the asymmetry of waiting.
- Weekly Review — Aug 24–30, 2026
Whether your agent should trust what it reads, with ads as a prompt-injection channel, and how it finds what to read, with MCP as the discovery layer.
- Weekly Review — Aug 17-23, 2026
Amara's Law and the AI value already here, the second web only AI crawlers see (now with ads), and why AI coding is still single-player.
- Weekly Review — Aug 10–16, 2026
Goodhart's Law and the goal you specify, why big neural networks stop getting stuck, AI moving into the database, what Anthropic's watermark really signals, and GuardDuty's SOC investigation agent.
