The $500K Engineer and the $250K Token Bill “If your $500K engineer isn’t burning at least $250K in tokens, something is wrong.” That quote, posted by Sunny Madra on X earlier this week, has been living rent-free in my head. My first reaction was to roll my eyes. My second reaction was to realize I…
Evals First. Everything Else Second. Most engineers I know treat evals like cleanup work. Something you wire up after the “real” engineering is done, right before you push to production and hope nobody notices the weird edge cases. I have done this. You have probably done this too. It is the wrong order of operations,…
Context Is the Moat. Not the Model. Most teams building AI products are optimizing the wrong thing. They debate model size, chase benchmark scores, and spend weeks on fine-tuning runs. Then they ship something that feels hollow. Generic. Like a customer support bot that clearly has no idea who it’s talking to. I’ve seen this…
The Study Nobody Is Reading Correctly Anthropic just published what they’re calling the largest qualitative study of AI users ever conducted. 81,000 people. One week. And most of the coverage I’ve seen misses the point entirely. Everyone is focused on the number. 81,000 is a big number, so people write about the number. But the…
The Hardest Skill in AI Engineering Is Knowing When to Stop I’ve been sitting with a uncomfortable observation for a few months now. We’ve crossed some kind of threshold in AI tooling where the answer to “can an agent do this?” is almost always yes. The models are capable enough, the frameworks are mature enough,…
Karpathy Just Scored Your Job. Software Engineers Got an 8.5. Here’s What That Actually Means. Andrej Karpathy built a tool that scraped all 342 occupations from the Bureau of Labor Statistics, fed each one to an LLM with a detailed scoring rubric, and produced an AI exposure score from 0 to 10. Software developers landed…
AI Is Not Your Search Engine. It’s Your Thinking Partner. But Only If You Let It. There’s a loneliness problem in AI development that nobody wants to admit exists. Not the philosophical kind. The practical, 11pm-debugging-an-agentic-workflow kind. Your teammates are asleep. Slack is quiet. Stack Overflow has never seen your exact edge case. You’re staring…
Karpathy Just Scored Every Job in America. I’m Not Sure I Like What I See. There’s a particular kind of clarity that lands like a gut punch. Andrej Karpathy, one of the founding architects of modern deep learning and a former Tesla AI director, just released an open-source project that scores all 342 Bureau of…
The Pattern That Actually Makes AI Agents Get Better Over Time Most AI-assisted development workflows have a hidden flaw. The model forgets. You fix a bug, the session ends, and next time you spin up a new context, the agent walks straight into the same class of mistake. You’re not building anything. You’re just running…
Assumption Debt: The Silent Killer of AI Products Most AI products don’t fail because the model was bad. They fail because the team spent six months optimizing for the wrong thing, and by the time anyone noticed, the cost of changing course felt unbearable. I’ve watched this happen repeatedly, and I’ve done it myself. The…
