AI Engineering · Seven-part route

Start with one task.Learn the Agent step by step.

You do not need to know the terms first. The series follows one Agent fixing a payment test, then answers six questions: how to state the task, prepare material, execute actions, advance one run, hand work across runs, and prove the result.

AI Engineering route map connecting Prompt, Context, Harness, Agent Loop, Outer Loop, and Evals

Reading route

See the complete story, then learn each part

PART 1 · MAPHow an Agent Task Actually Gets Done

Follow one task from request to verified result, then name the six engineering problems along the way.

Start reading → PART 2 · PROMPTPrompt Engineering: Make the Task Requirements Clear

Start with a vague request and add the goal, background, constraints, done criteria, and output.

Start reading → PART 3 · CONTEXTContext Engineering: Prepare the Right Material for the Next Step

Distinguish the library, retrieved candidates, and the material the model can actually see now.

Start reading → PART 4 · HARNESSHarness Engineering: Give the Agent a Safe Place to Act

See how “run the tests” passes through tools, permission, isolation, execution, and records.

Start reading → PART 5 · AGENT LOOPAgent Loop: Help the Agent Finish a Task Step by Step

Follow repeated rounds of reading, editing, testing, and deciding inside one run.

Start reading → PART 6 · OUTER LOOPOuter Loop: Let Runs Hand Off Long-Lived Work

A run ends; learn how the task preserves progress, recovers, verifies, and starts another run.

Start reading → PART 7 · EVALSEvals: How Do We Know the Agent Really Succeeded?

Start with one repeatable task, then add outcome, process, safety, and operations checks.

Start reading →