Signal Roads

Agent Foundations
for Marketers

An AI agent in an editor you control, put to work on real marketing tasks. No code or terminal experience needed. Works with Claude, Gemini, or Codex.

8 lessons · ~1.4 hrs of reading

Building alongside takes most people a few sittings, and it is where the value is.

Start the course

Subscribe to get new lessons and courses as they land.

Prerequisites None. You don’t need to code or to have used a terminal before. A laptop and access to one AI coding tool (Claude, Gemini, or Codex) is enough.

What changes for you

This course teaches you to drive an AI agent in an editor: handing it whole pieces of marketing work and getting finished work back, instead of asking a chat for answers and assembling the rest yourself. Agents do two kinds of marketing work, and you meet both. The making you do on your own product, a landing page built from one prompt. The thinking you practice on a real company’s files that come with the course, so the deep work is safe to get wrong while you learn the machinery behind it: context files, data connections, checkpoints, saved workflows.

  • Hand the agent whole pieces of work, a landing page, a positioning read, a research pass, and get finished work back instead of text to assemble
  • Feel what context does to output, by interrogating a real company’s reviews and positioning and watching generic answers turn specific
  • Know both roads into your stack, connectors and keys, and the least-access habit that keeps a session gone sideways from touching anything
  • Run it on real work without fear, with an undo button under every change and a permission gate in front of anything you could not take back
  • Read and run saved workflows, so you know what a system looks like from the inside before you build your own

Curriculum

Part 1 · Understand

  1. 1What is an AI agent? A marketer's guideThe shift that makes agents click, before any setup7 min
  2. 2How to pick and set up an AI coding toolPick your editor and agent, set it up, and run your first task14 min

Part 2 · Build

  1. 3How to build a landing page from an AI promptOne prompt becomes a real page in your browser, nothing plugged in11 min

Part 3 · Context

  1. 4AI context file: what it is and what it changesThe file that turns generic output specific, felt on a real company10 min
  2. 5How AI agents reach your data: MCP and API keysMCP and API keys, scoped read-only first so nothing breaks13 min

Part 4 · Operate

  1. 6AI models and credits: how to optimize usageMatch the model to the job and stop wasting credits on cheap work11 min
  2. 7Git and GitHub, explained for non-codersThe undo button, the online home, and the door to going live9 min
  3. 8AI agent skills: turn workflows into systemsFrom one-off tasks to saved workflows that run themselves11 min
Nicolaas Spijker

I'm Nico. I run real client work through an agent in my editor every day, and I write about it in the Build Log. This course is the how-to version: the same workflows, set up so you can run them yourself.

Get the rest as I write them.

All eight lessons are live. I keep refining them in the open, no set schedule, just when they're good.