"If you follow the steps in Chapter 8, you will have a working AI employee — with identity, memory, and a defined role — by the end of one afternoon. Not a ChatGPT tab. An actual system that does work for you."
— From the introduction
Using AI vs. Hiring AI — The Real Difference
|
Using AI (most people) |
Hiring AI (this playbook) |
| Context |
Re-explained every session |
Persistent across weeks and months |
| Identity |
Generic, inconsistent output |
Defined personality via SOUL.md — consistent every time |
| Memory |
Starts from zero each call |
3-layer memory system — working, project, semantic |
| Tools |
Text generation only |
Real capabilities: code, files, APIs, web, email |
| Safety |
Hope for the best |
Green/Yellow/Red action classification — explicit zones |
| Income |
Saves you some time |
Scales output — same hours, 4–10x clients/deliverables |
What's Inside — 11 Chapters
01
The Foundation
Why hiring is different from using. Platform options. The SOUL.md file — the highest-leverage thing you'll create.
pp. 6–11
02
The Memory Architecture
The 3-layer system: working memory, project files (DECISIONS.md, OPEN_LOOPS.md), and semantic memory via Supabase pgvector.
pp. 12–16
03
Tools, Sub-Agents & Delegation
Tool access levels in risk order. The orchestrator + specialist pattern. The sub-agent brief format that produces clean outputs.
pp. 17–22
04
Safety Rails & The Trust Ladder
Green/Yellow/Red action classification. The 5-rung autonomy ladder with real advancement criteria. The pullback rule.
pp. 23–27
05
The Operating Relationship
Daily rhythms: morning brief, daily log, weekly review. The 15-minutes-a-day system that keeps everything calibrated.
pp. 28–32
06
Making Money With AI Agents
6 proven income models with real revenue ranges. The productised service formula. Why you should never discount AI-assisted work.
pp. 33–40
07
Coding Agents at Scale
The Ralph loop (self-correcting execution). TDD prompting. Parallel execution. When not to parallelise.
pp. 41–45
08
Quick-Start Kit
Zero to working AI employee in one afternoon. Every command, every config file, every decision point — in order.
pp. 46–51
09
The Sentry Pipeline
Detect, triage, fix, and ship bug fixes autonomously. Full webhook handler code included. Works while you sleep.
pp. 52–55
10
Advanced Configuration
Multi-agent architecture, Tailscale remote access, semantic memory, production-grade infrastructure checklist.
pp. 56–60
11
Honest Retrospective
What went wrong. What surprised us. The parts nobody posts about. No polished success narrative — just what actually happened.
pp. 61–62
Templates — Copy & Customise
Every template has been used in production. Replace the [BRACKETS], drop them into your setup, and start today. No writing from scratch.
SOUL.md
Your AI's identity file — values, personality, anti-patterns, communication style, scope limits. The highest-leverage document you'll ever write.
SAFETY.md
Green/Yellow/Red action classification. Read at every session start. Defines exactly what your agent can do autonomously vs. what requires your approval.
MEMORY.md
Session context loader — project overview, tech stack, active projects, brand voice, current revenue model. Paste at session start.
DAILY_LOG.md
Morning brief + action log template. Every action taken, every decision made, every blocker identified — structured for async review.
DECISIONS.md
Institutional knowledge file. Every significant decision with date, reasoning, and status. Prevents relitigating settled questions.
Sub-Agent Brief
The brief format that produces clean specialist outputs — task, success criteria, constraints, context files, output format. Used in every delegation.
A Sample — SOUL.md Before vs. After
Request: "Write an intro paragraph for our new product launch blog post."
| Without SOUL.md | With SOUL.md |
|---|
| "We are excited to announce the launch of our new product! This innovative solution will transform the way you work and help your team achieve greater productivity. Read on to learn more about this exciting release!" |
"The waitlist is closed. MyStreakboard is live.
One-time payment. No subscriptions. No free tier.
If you build habits seriously, this is the tool." |
Same model. Different identity. The second version required zero editing.
Chapter 06 — Making Money With AI Agents
The most detailed chapter in the playbook. Six models currently being run by solo operators. Real revenue ranges, real setup effort, real AI role.
🏭
Productised AI Service
Recurring deliverable powered by agents in the background. R3,000–R15,000/month per client. AI does 80% of production; you do QA and client comms.
📦
AI-Powered Digital Product
Guides, template packs, toolkits built with AI assistance and sold at a fixed price. R500–R2,000 per sale. Low ongoing cost once built.
✍️
Content System as a Service
Build a content pipeline for a client using your AI stack, then manage it monthly. R8,000–R25,000/month. AI drafts everything; you approve.
🔬
AI Research & Synthesis
Structured research reports on competitors, markets, trends. R2,500–R8,000 per report. AI reads hundreds of sources; you write the exec summary.
⚙️
Automation Builds
Custom AI workflows for client processes. R20,000–R80,000 per project + retainer. AI builds the tools; you design the workflow.
🎓
Community + Education
Paid community or course teaching your AI system. R300–R2,000 per member. AI handles Q&A, content creation, and community management.
Who It's For
Solo founders & creatorsBuilding products, content, and systems with limited hours and no team
Freelancers & consultantsWho want to serve more clients at the same hours by delegating production to AI
DevelopersWho want to run coding agents properly — Ralph loops, TDD prompts, parallel execution
Anyone who's used AI toolsAnd felt like they were getting 20% of what was possible — this is the other 80%
Technical level: Comfortable with AI tools. Some Python code is included but the playbook works without running any of it — the frameworks and templates are the core value.