People are finally catching on 👀

Something massive just happened in software engineering. For 8 straight months, every dollar spent on AI coding bought less code than the month before. January: $2.80 per expert hour of output August: $11.00 That's a 4x cost increase in 7 months. Teams kept buying more tokens. Code wasn't keeping up. And then we saw a huge shift in September... For the first time all year, output grew FASTER than AI spend. Cost per output hour: $11 → $9.39 (−15%) Code per dollar: 0.09 → 0.106 (+18%) The first wave of AI coding was brute force. Throw tokens at the problem, hope something ships. What we may be seeing now is the efficiency era. Engineers are learning which tasks to hand off, how to prompt, and when not to use AI at all. The engineers who figure this out first won't just be faster. They'll be dramatically cheaper. One month isn't a trend. But we'll see what happens this month.

Had an insane September. Shipped 40% more per engineer MoM. Few notes on it: 1. Agent parallelization: our engineers ran multiple agents in parallel on multiple tasks, and we've learned to multiplex. 2. We immediately incorporated new models into our workflow thanks to the router. 3. More Slack agents and coding agents. Our engineers are running more of both day to day. 4. We built an internal code quality check (more to come on that soon).

I jumped out of a plane for our Series A announcement. For the celebration, we’re staying on the ground. Thursday: run with us. Friday: party at the new office. DM me, and I’ll add you to the list for Friday. Luma: luma.com/a5ouvlt8