Earlier this week, my colleague Jyoti reported on the great lengths Meta Platforms engineers are going to show off their AI prowess. An employee-created leaderboard tracks who burns through the most AI tokens—part of a broader trend known as “tokenmaxxing.” The article has sparked a flurry of armchair analysis on how much this strategy could cost Meta (and potentially benefit Anthropic), such as here and here.

The drive by companies to use AI for coding could boost demand for tools that help developers organize code generated by AI agents. One startup developing such tools is GitButler, whose software streamlines how developers coordinate their code alongside AI coding agents. The Berlin-based startup just raised a $17 million Series A led by Andreessen Horowitz, GitButler CEO Scott Chacon told me. Existing investors A.Capital Ventures and Fly Ventures also participated.

Tools that help developers see, organize, and track each others’ code changes, known as version control systems, have been around for the past few decades. But this software wasn’t ready for AI that writes code and works alongside human developers, says Chacon. So three years ago he co-founded the company with startup entrepreneurs Anne Leuschner and Kiril Videlov.