The One-Person Dev Team: 94 Commits in a Day

The Setup

> "I don't use Codex or Claude Code directly anymore. I use OpenClaw as my orchestration layer."

An orchestrator agent spawns coding agents, writes their prompts, picks the right model for each task, monitors progress, and pings on Telegram when PRs are ready to merge.

## Proof Points (4 weeks)

- 94 commits in one day — most productive day, had 3 client calls, never opened editor
- 7 PRs in 30 minutes — idea to production blazing fast
- Commits → MRR — real B2B SaaS, founder-led sales, same-day feature delivery
- Cost: ~$100/month Claude + $90/month Codex (start with $20)

## Why This Works: The Two-Tier System

Problem: Context windows are zero-sum. Fill with code → no room for business context. Fill with customer history → no room for codebase.

Solution: Specialization through context, not different models.

- Orchestrator (OpenClaw) — holds business context: customer data, meeting notes, past decisions, what worked/failed
- Coding agents (Codex/Claude Code) — focused on code, receive precise prompts from orchestrator

Each AI loaded with exactly what it needs.

## The 8-Step Workflow

### Step 1: Customer Request → Scoping with Zoe

After customer calls, meeting notes sync to Obsidian vault automatically. Zoe already has context.

Zoe:
1. Tops up credits (has admin API access)
2. Pulls customer config from prod DB (read-only, coding agents never have this)
3. Spawns Codex agent with detailed prompt containing all context

### Step 2: Spawn the Agent

Each agent gets its own worktree (isolated branch) and tmux session.

### Step 3: Monitor Progress

Zoe checks on agents every 30 minutes. If one dies, restart it. If stuck, provide guidance.

### Step 4: Review and Merge

When PR is ready, Zoe pings on Telegram. Review, approve, merge.

## The GStack Workflow

Our skills implement this pattern:

- Compass: The orchestrator mindset for product decisions
- Blueprint: Architecture review before coding
- Liftoff: Merge → test → version → changelog
- Mirror: Weekly analysis of what shipped

One person. Multiple agents. Team velocity.