Oct 14, 2026 · Victor Slompo
When agents scale, git can become a bottleneck
The problem stops being writing code. It becomes operating change.
The images are the original Instagram carousel, in Portuguese.
WHAT HAPPENED
Anthropic redesigned version control for agents
The company created Scion, its own versioning system, and migrated its monorepo off git in five months. The reason was straightforward: with agents producing more code, source control became the biggest development bottleneck.
THE NUMBER
30,000
code changes merged per day in September 2026
In May 2026, the average was around 2,500 per day.
THE REAL PROBLEM IN COMPANIES
What I see in companies when they bring in coding agents
- 01The team generates change faster than it can review safely
- 02The workflow is still designed for humans, not for agents that are always online
- 03Without merge and rollback governance, productivity turns into rework
I’VE ALREADY SOLVED THIS
What I did with the AskMoses.AI development team
- 01Each dev works with Claude Code on the project
- 02Another Claude reviews the pull request
- 03Deploy on Vercel with 1-click rollback and daily backup
MY TAKE
Simple that works is worth more than complex that impresses.
In practice, using agents well does not start with the model. It starts with the workflow. Review, human business decisions, safe deploys, and simple rollback are what sustain production. Without that, vibe coding just speeds up the chaos.
Sources
Want this working in your company?
I take AI off the slide deck and put it into production, with measured results. On Instagram, comment CLAUDE on the post and I'll send you the step-by-step by DM.