Agentic
Engineering

Sixty minutes to understand how AI agents really work — and what you can build with one.

No degree, no math, no vendor pitch. Just the mechanism: how a model thinks, how a loop turns thinking into doing, what breaks, and how to start this week.

Start module 01 → See the whole path

10 modules

60 minutes

18–55+ no prerequisites

6 interactive exercises

Every module is a part of this circle.

The path

Ten short modules. One working mental model.

Read in order the first time. Each module is one idea, one diagram, one thing to try, and one question to check yourself. The minute counts are honest — the whole path is about an hour.

01 What an agent actually is The four parts. The autonomy ladder. Why "agent" is mostly a marketing word. 6 min 02 How the model thinks Tokens, next-word prediction, context windows, and why confidence is not correctness. 7 min 03 The loop: think, act, look The single mechanism behind every agent. Step through a real task trace. 7 min 04 Tools: how it touches the world Function calling, permissions, sandboxes, MCP — and the risk of a tool that can delete things. 7 min 05 Context and memory The budget, retrieval, and the difference between "remembers" and "was told". 7 min 06 Patterns: six ways to wire a loop ReAct, plan-then-execute, routing, reflection, human gates. Pick by task shape. 6 min 07 Many agents: power or paperwork Supervisors, hierarchies, swarms — and why one agent usually wins first. 5 min 08 Evaluation: how you know it works Golden sets, benchmarks, failure modes, and how agents get hijacked. 6 min 09 Production: money, safety, sleep What a task really costs, what to log, and how to fail gracefully. 5 min 10 Start here: your first 30 days A minimum viable stack, three project recipes, and what to ignore. 4 min

Core path60 min

After this hour you can

Explain an agent to a colleague without buzzwords. Read an agent trace and see where it went wrong. Estimate what a build will cost. Know which framework question to ask first.

What this course is not

Not a prompt-engineering tutorial, not a vendor comparison, not a promise that agents replace your team. It is the mechanism, honestly described, including the parts that still break.

How to take it

One sitting. Keep a note of anything that makes you say "wait, really?" — that's where your first project lives. The timer in the corner is optional; the questions are not.

Who it's for

Curious beginners, career switchers, operators and managers who keep hearing "we should use agents", teachers, and anyone between 18 and 55 who wants the real picture rather than a demo. If you can use a spreadsheet and read a diagram, you have every prerequisite.

Free to read and share · CC BY 4.0 · Starphix Course Library 001. Content is drawn from a curated research vault; the underlying papers, standards and docs are listed in Sources.