Finance
Close, forecast, reconcile, approve.
Building With Agents
A zero-to-builder guide for creating useful software, reliable agents, and your own operating system.
Start with no assumed technical knowledge. Learn how computers and software work, build your first useful tool, and progress toward an AI operating system you can understand, operate, and improve.
Zero technical background → system builder
- 01Problem
- 02Workflow
- 03Tool
- 04Agent
- 05Your OS
Built for complete beginners
Start with zero. Build what matters.
No coding vocabulary is assumed. Each chapter introduces only the concepts you need for the next build, then returns to them at greater depth as your software becomes more capable and consequential.
People & HR
Hire, onboard, support, protect access.
Customer support
Triage, investigate, resolve, learn.
Marketing
Research, create, review, publish, measure.
Operations
Coordinate handoffs, exceptions, and delivery.
Sales
Qualify, prepare, communicate, follow through.
Product
Connect evidence, decisions, experiments, outcomes.
Leaders & owners
See the truth, choose, delegate, verify.
18 original chapters
The field guide
18
Read front to back. The sequence is deliberate: plain-language mental models first, one useful build next, agent engineering after that, and your own operating system at the end. Every chapter includes a case file, technical pattern, exercise, reusable artifact, and proof checks.
18 of 18 chapters
- 01The Builder's Starting LineSee how the book will take you from domain expert to responsible builder.18 min read
- 02Your Computer, ExplainedBuild a usable mental model of files, apps, the terminal, local servers, and secrets.24 min read
- 03How Software Fits TogetherTrace a request through an interface, logic, data, identity, and external services.26 min read
- 04Start With the WorkMap a real workflow before choosing tools or agents.19 min read
- 05Build Your First Useful ToolPlan, build, test, and share one complete vertical slice with an AI coding agent.28 min read
- 06Make Changes SafelyUnderstand Git, branches, worktrees, diffs, tests, and deployment well enough to supervise agent-made changes.30 min read
- 07Prompts Are SpecificationsWrite specifications that survive different models, contexts, and runs.23 min read
- 08Models Are ComponentsEvaluate models as replaceable system components instead of brands or personalities.24 min read
- 09Give Agents Jobs, Not VibesCreate worker lanes that cooperate without blurring ownership.22 min read
- 10Tools, APIs, and the HarnessChoose and engineer the runtime around an agent instead of blaming every failure on the model.25 min read
- 11Truth Before IntelligenceBuild a truth layer agents can inspect instead of improvise.22 min read
- 12Patterns for Real SystemsUse a small catalog of agent patterns without turning every workflow into a swarm.25 min read
- 13Approval Is a FeaturePlace human judgment where consequences become real.21 min read
- 14Receipts or It Didn’t HappenDefine evidence for local, shared, live, and installed work.22 min read
- 15Operate and Debug the LoopFind failures, control cost and latency, and improve the system from real trajectories.25 min read
- 16Design Your Own OSDesign a clear interface over specialized agents and tools.20 min read
- 17Make It Ready for Real LifeKnow when a workflow is ready for autonomy and how its state survives beyond one session.26 min read
- 18Build and Evolve Your OSConnect interface, truth, workers, approvals, and receipts.24 min read
Three steps. One system.
01
Understand
Plain-language explanations build a mental model before introducing commands or architecture.
02
Build
Replit, Claude Code, Cursor, and Codex examples show how real work becomes useful software.
03
Operate
Tests, approvals, receipts, recovery, and live proof turn prototypes into systems you can trust.
Beginner-to-builder path
Zero to your own OS.
The curriculum grows from basic computer literacy through software anatomy, first builds, prompts, models, tools, context, agent patterns, approvals, evaluation, operations, and long-horizon systems. Retool appears only as employment and domain context; the building examples use Replit, Claude Code, Cursor, and Codex.
Zero to system thinker
Start without assumed technical vocabulary. Learn how computers, applications, data, APIs, identity, and requests fit together.
First useful build
Map real work, ship one complete tool, and learn Git, branches, worktrees, diffs, tests, environments, and rollback.
AI building blocks
Write specifications, choose models, give agents bounded jobs, and understand tools, APIs, skills, and harnesses.
Reliable agent systems
Design context, patterns, approval gates, receipts, debugging, evaluation, recovery, cost, and operational proof.
Your own operating system
Create one front door, separate control, data, and runtime, build a complete vertical loop, and grow it responsibly.
You made it. You built your OS.
You now have the clarity, tools, and systems to build, run, and evolve your own operating system.
Keep building