A dedicated publishing area for weekly AI signal: commentary, public opinions, and practical analysis for professionals making decisions.
March 7, 2026
Sequoia's Julien Bek says the next trillion-dollar company sells outcomes, not tools. The insight holds. The framing is a pitch memo wearing analysis.
March 5, 2026
GPT-5.4's coding and tool-use improvements are real. The claim that it 'matches professionals' is doing more heavy lifting than the model itself.
March 5, 2026
London's structural advantages for AI are real. The question nobody's asking is why those same advantages haven't fixed productivity since 2008.
March 5, 2026
OpenAI's harness engineering post contains a genuine insight about documentation as infrastructure. It is also, unmistakably, a piece of marketing.
March 3, 2026
A thorough survey of how top AI agents are actually built makes the case that the harness is the product. The data underneath tells a more complicated story.
March 3, 2026
A CEO's essay on identity and AI adaptation sounds right — until you notice the seventeen jobs lost get one paragraph and the freedom of letting go gets six.
March 3, 2026
Code review is drowning under AI-generated output. The proposed fix — specs, agents, and layered verification — diagnoses a real problem then overshoots the landing.
March 3, 2026
A viral thread says AI agents can work while you sleep, turning idle hours into output. Coal doesn't hallucinate. Spare bedrooms don't need code review.
March 3, 2026
Agent frameworks filled the space between good-enough models and missing infrastructure. Now the labs are building down and automation platforms are building up.
March 3, 2026
AI became a power story, not a technology story. Framing it as Act Two implies an Act Three. There is no evidence that's coming.
March 3, 2026
A viral guide to agentic engineering buries useful advice on context management inside an ascension narrative that contradicts its own thesis.
March 3, 2026
A weekend guide to Perplexity Computer tests every prompt and none of the failure modes. The most useful thing a tool review can tell you is where it breaks.
Subscribe to the AI Primer Briefing for weekly commentary on what is happening in AI and what it means for your work.