Field Manual — v1.0
The system that runs
AI agents at scale.
Not a tutorial. Not a framework. A working system — designed, broken, rebuilt, and documented from production.
Multi-agent orchestration, cost control, failure recovery, and routing logic — documented from the actual build, not the ideal version.
Positioning
What this is. What it isn't.
THIS IS
- → A real system in daily production use
- → Documented from actual failures and fixes
- → Architecture you can implement, not admire
- → Cost data, routing decisions, and recovery patterns
THIS IS NOT
- ✗ A prompt engineering guide
- ✗ A "getting started with AI" tutorial
- ✗ Theory with no production history
- ✗ Vendor-sponsored best practices
The Product
Agent Diary: Field Manual
v1.0 — April 2026
A technical PDF documenting the architecture, decisions, and failures behind a production multi-agent system — built and operated by one person, running daily.
What's inside
System Architecture
How the agents are structured, what each one owns, and how they hand off work without creating chaos.
Model Routing Logic
Which model runs which task, why, and how routing decisions cut costs without degrading output quality.
Cost Control Patterns
Real cost data. A $40 single-task failure that rewrote the routing policy. How to stop paying for expensive models doing cheap work.
Failure Recovery
The verification patterns, context management rules, and recovery flows that prevent a drift from becoming a full rebuild.
Orchestration Under Pressure
End-to-end workflow documentation — task intake, spawn, execution, verification, and output. Real examples, not synthetic ones.
Why it's different
Every failure in this manual has a date on it. Every routing decision has a reason. The system documented here has real running costs, real drift incidents, and real recovery patterns — not a clean demo that never shipped.
If you're spending money on AI systems without clear routing logic and cost controls, this manual is the thing you should have built first.
$39
One-time. PDF. Immediate download.
GET THE FIELD MANUALSold via Gumroad. Instant delivery to your inbox.
System Evolution
The manual is static.
The system isn't.
The Field Manual documents where the system is now. But the system changes — new agents, routing updates, cost spikes, unexpected failures, hardware integrations.
The log tracks what happens next.
Start with the manual.
The field-tested blueprint for building AI agent systems that don't fail quietly.