How DHH Went from AI Skeptic to 'Fully Pilled' in Three Months

DHH was impressed by AI from day one. He was a ChatGPT user from the first days of the interface. But impressed and converted are different things. For a long time, he used AI as a better Google, a better Stack Overflow, a pair programmer he could bounce problems off. He hated autocomplete in his editor because it interrupted his train of thought every five seconds to produce code he did not like. Not wrong code necessarily, but code that did not match his aesthetics.

The pattern held through his Lex Friedman interview in summer 2025, where he was essentially saying: this is amazing, you would be a fool not to take it seriously, but it is not in my business yet. He gave a similar take in early December. Then the models that dropped in late November, including Opus 4.5, crossed a threshold. The output went from "I don't like what it's making" to something he was willing to merge.

The shift in the last three months has been the most churn in DHH's mental approach to computers in his entire career. More deep questions, more workflow changes in his daily eight hours than in any previous period. He describes the experience as having grown 18 arms and seven more brains, operating 22 screens simultaneously. He is still doing the work, still directing everything, but the pace of progress when the stars align is unbelievable and addictive.

What makes this psychologically interesting is that DHH is now in the consensus for the first time, and he is uncomfortable about it. He is a contrarian by nature. But he has concluded that at some point you just have to admit the math maths and that AI agents are incredibly capable. He was wrong about the timeline and says so openly, drawing a direct parallel to his Facebook prediction: sometimes you underestimate how quickly a paradigm shift will change everything.