# g4a - the reasoning layer for AI-written code

> Git stores what changed. The reasoning behind it is lost. g4a captures it. Add it to your existing project. Unlock 10x for both humans and AI agents.

## The problem

Every AI coding agent thinks deeply while writing code - reading files, considering alternatives, testing approaches. Then at commit time, all of that thinking is thrown away. The commit message is a 10-word summary of 100 steps of reasoning. This happens with every agent, every commit, every tool.

## What this causes

- PR review time increased 91% on AI-heavy teams (Faros AI, 10K+ devs, 1,255 teams)
- 46% of developers actively distrust AI-written code, only 3% highly trust it (Stack Overflow 2025, 49K+ devs)
- 66% waste time on "almost right" AI code (Stack Overflow 2025)
- Debugging is terrifying - no trail of WHY agent decisions were made
- 70% individual productivity gain, only 17% team collaboration gain (Stack Overflow 2025)
- AI agents start from zero every time - no memory of past decisions, rejected alternatives, or tradeoffs

## How g4a helps humans

- PR reviews go from 25 minutes of guessing to 3 minutes of reading with full reasoning, confidence scores, and risk flags
- Debugging has a trail: g4a why <function> shows the complete decision history
- Trust becomes data-driven: low-confidence changes get scrutiny, high-confidence flows faster
- Onboarding is instant: new developers read the decision history instead of guessing from diffs

## How g4a helps AI agents

- Agents read past reasoning and understand WHY code looks the way it does
- Agents stop duplicating work - they start where the previous agent left off
- Agents coordinate through shared context instead of colliding blindly
- The codebase accumulates institutional memory that makes every future agent smarter

## The solution

g4a captures reasoning from AI agent sessions and stores it in .g4a/ inside your repo. No server, no account, no hosting. Clone the repo, get the reasoning. Install with pip install g4a or brew install lcbasu/g4a/g4a, then g4a init. Works with all AI coding agents - launching first with Claude Code (direct transcript parsing), with git hook fallback for all others.

Full content: https://www.git4aiagents.com/llms-full.txt
Website: https://www.git4aiagents.com/
GitHub: https://github.com/lcbasu/git4aiagents
Author: Lokesh Basu (@lcbasu)
License: CC-BY-4.0