I spent most of this week coding. Not something I expected to say in 2026. But here I am, decades later, building an app with AI agents, shipping features between meetings, and feeling that same rush I felt as a junior engineer watching my code compile for the first time.
Back to the Terminal — The main essay. Let's stop calling it "vibe coding." Karpathy retired that term in favor of "agentic engineering," and I think that's right. The process is the same: ideation, architecture, design, build, test, validate, document. The speed is radically different. I kick off multiple AI agents before a meeting and come back to code complete and ready to test. The bottleneck flipped from coding to human judgment. AI still can't tell you which design is better for the user, and it still goes down rabbit holes trying the same broken fix six times. The WarGames tic-tac-toe lesson applies: some games you stop playing.
Also in this issue:
- Signals This Week — The race to make AI agents run your computer. Anthropic shipped Dispatch, Background Agents, Computer Use, and Cowork in March alone. OpenAI is pushing Codex. OpenClaw set the bar for how easy it should feel. The gap between early adopter and everyone is shrinking fast.
- The Wire — AI agent security is becoming its own product category (Snyk + Bessemer). Salesforce bakes Agentforce into SMB tiers at no extra cost. The consulting model is breaking, and Liz Henderson wrote it down.
- What I'm Consuming — Stripe's "Minions" (1,300 PRs/week, zero human coding). Comprehension debt: when AI generates code faster than you can review it, you stop understanding why your systems work.
- After Hours — Piranesi by Susanna Clarke. A mystical experience, should you be in a mood to take a break from the stress of modern life.