GitHarness paper: managing agent memory like a Git repo to handle mid-task requirement shifts
rohanpaul_ai · x · 2026-10-05
A new arXiv paper introduces GitHarness, a pluggable Git-style framework that organizes user requirement states and agent work states into a branchable version history for long-horizon LLM agent collaboration.
- Problem: Users rarely give a full spec upfront — they add, revise, or drop requirements mid-task, and agents either carry obsolete information or turn local revisions into global rewrites.
- Approach: Joint requirement tracking and local update; a trainable Git Agent resolves requirement changes and selects a semantically compatible historical state, then a unified version interface restores that state and creates a new branch, excluding stale info and inheriting compatible work.
- Takeaway: Build versioning (commits/branches) into any agent that takes user feedback.
Related event: GitHarness Manages Agent Memory Like Git, Cutting Tokens 73.6%(3 posts)→
More from coding & agent
- Railway launches Cloud Agent Tasks: machine, conversation and result are separate — reusing a VM doesn't mean memory — krishnan · 2026-11-20
- Eric Schmidt: founding an agentic AI company is the easiest way to get rich in the next two years — AICopyLab · 2026-11-20
- Apple reportedly building proactive AI agent that acts autonomously on users' behalf — burny_tech · 2026-11-20
- NormGym benchmark shows agents learn from own norm violations but not from others — xuanalogue · 2026-11-20
- Every adopted 'compound engineering' company-wide so its codebase gets easier to maintain — every · 2026-11-20
- Not every agent task needs a general LLM — specialized small models by task type — ziv_ravid · 2026-11-20