Google Has the Same AI Adoption Curve as a Tractor Company

Steve Yegge shared a conversation with a Google tech director of 20 years. The breakdown: 20% agentic power users. 20% outright refusers. 60% still in basic chat mode. Simon Willison surfaced the thread, and Yegge's punchline was devastating: Google's internal AI adoption is "comparable to John Deere."

Google. The company that invented the transformer architecture. The company with Gemini, TPUs, and more AI PhDs per square foot than anywhere on earth. Their own engineers are using AI the same way farmers use it. Two out of ten go deep. Two refuse entirely. The other six treat it like a fancy autocomplete.

Yegge flagged something I haven't seen anyone else catch: there's been an industry-wide 18+ month hiring freeze. No external hires are arriving to flag how far behind internal practices have fallen. The people who've been inside Google for years don't know what they don't know, and nobody new is coming in to tell them.

This maps to what I see in my own work. The gap between someone who uses Claude Code as a chat assistant and someone who runs it as an agent with skills, worktrees, and CLAUDE.md context engineering is enormous. Same tool. Completely different output. The 60% in chat mode aren't lazy. They just haven't been shown what agentic usage looks like, and their organizations aren't structured to teach them.

The Stanford AI Index confirms this at macro scale: 88% organizational AI adoption, but "adoption" covers everything from "we have a ChatGPT license" to "agents run our SOC." The number is meaningless without depth metrics.

What to do about it: Audit your own team honestly. What percentage are in chat mode? Agentic power users? Refusers? The answer will surprise you. Then invest in the middle 60%. That's where the ROI is. One engineer going from chat to agentic usage is worth more than buying another AI tool.