Should AI Coding Sessions Be Part of Your Git Commits?
Original: If AI writes code, should the session be part of the commit? View original →
A New Question for the AI Coding Era
The GitHub repository Memento earned 221 points on Hacker News by asking a deceptively simple question: if AI writes the code, should the AI session be part of the commit?
In traditional software development, commit messages record what changed, and comments or documentation explain why. But when AI tools generate significant portions of code, much of the context lives inside the AI conversation session — and then disappears.
What Gets Lost
When an AI session ends without being recorded, several valuable artifacts are lost:
- Which prompts produced which code
- What trade-offs were discussed with the AI
- Which alternatives were considered and rejected
- The reasoning chain that led to a particular implementation
For future developers — or even the same developer months later — this missing context can make AI-generated code harder to maintain and reason about than hand-written code.
What Memento Proposes
Memento provides a mechanism to capture AI coding sessions and link them to Git commits. This creates a traceable record from the original intent and dialogue through to the final code, enabling true code archaeology in the age of AI assistants.
Community Debate
The Hacker News discussion revealed strong opinions on both sides. Proponents argue that session logs are essential for understanding AI-generated code. Critics raise concerns about repository bloat, privacy implications of storing AI conversations, and whether this shifts too much complexity onto developers. The debate reflects a deeper question: how must version control evolve as vibe coding becomes the norm?
Related Articles
Simon Willison reflects on how the once-clear line between careless vibe coding and responsible agentic engineering has begun to blur in his own work, raising sharp questions about trust and accountability in AI-assisted software development.
With AI coding tools handling implementation details, the classic arguments for Python's simplicity carry less weight. But the real reasons to use Python - its ML ecosystem and readability - remain as strong as ever.
A high-engagement Reddit post surfaced TechCrunch reporting that Spotify engineers are using Claude Code and an internal system called Honk to accelerate coding and deployment.