What a Coding Agent Actually Does
A plain-English tour of coding agents: what they can build, where they fail, and what that means for you.
The Contractor Analogy
A coding agent is software that writes, runs, and fixes other software on your behalf. The best mental model is a contractor: you describe the outcome, it does the labor, and you inspect the result.
You don't need to know how to code to direct one — the same way you don't need to swing a hammer to have a house built. You do need to know how to brief, inspect, and accept work. That's what this module trains.
What Agents Do Well Today
- Websites and landing pages. Describe the page; get a live site.
- Automations. "When a form comes in, add the person to my list and draft a welcome email."
- Internal tools. Dashboards, calculators, intake forms, small databases.
- Glue. Connecting the tools you already pay for so data stops being re-typed.
Where They Fail
Agents fail when the brief is vague, when nobody checks the work, and when the project is too big for one bite. All three failures are preventable, and all three are the director's job — your job — not the agent's.
The Reframe
Stop asking "can AI do this?" Start asking "could a competent contractor do this in a day if I explained it well?" If yes, an agent probably can — and your next lesson teaches you how to write that brief.