@mati Oh this is amazing! I was just playing with elevenlabs and Claude to help build something for my father in law whose voice box was removed. He wants to recreate his voice back when he speaks thru the electrolarynx. Definitely interested.
Home time. The USA have been fantastic hosts! We’ve loved the welcome from every city and state we’ve been in 🇺🇸
If you want to hate the world, watch the news. If you want to love the world, travel it. 🌎❤️🏴✈️🇺🇸 #scotland#fifaworldcup
The World Lives Here.
Canada is all together different. Truly unique. And so is our relationship to the beautiful game. This is our opening to the 2026 World Cup. Narrated by Kiefer Sutherland. #FIFAWorldCup
1/8 Started Claude Code in January. By April: 1,658 session logs. 44 active days. 11 weeks.
Not a review. Here is what the data says and what I actually learned from it.
7/8 Senior engineers do not become less valuable with AI. They become more valuable.
Junior engineers working unsupervised with these tools is a risk most teams are not thinking hard enough about.
6/8 I caught hallucinated code not because I ran tests. I recognised the pattern was wrong before it compiled.
I made better decisions not because the AI flagged risks. I had seen those patterns fail before.
5/8 The biggest thing I learned had nothing to do with speed.
The engineers who get the most from these tools are the ones who already know where things break.
4/8 I learned each surface serves a different kind of thinking.
Terminal: long deep builds. Browser: architecture questions, live troubleshooting, quick research. Mobile: the gaps between everything else.
3/8 By April it looked nothing like that.
Bash: 22,393 calls Edit: 5,280 Tasks: 4,659 Subagent runs: 1,520
Avg session: 67 min. Longest: close to 14 hours.
2/8 February: watching more than building.
Top tools were Read and Glob. Sessions averaged 15 minutes. Every session subagent-driven. I was figuring out what was possible, not yet doing it.