My ideas have always run faster than my ability to build them. I can sketch a circuit, breadboard a sensor, and read a datasheet without much trouble. But turning an idea into working, deployable software always sat just out of reach. If that sounds familiar, this course was built for you, and it is available today.
What this course is
Most makers can write a short script or adapt an example they found online. Turning a real idea into a working application is a different problem. The gap between “I can code a little” and “I can ship software” used to take years of practice to cross. AI coding assistants collapse that gap, but only if you know how to use them well.
That last part is the whole point. Pointing an AI assistant at a vague request and hoping for the best is how you end up with confident, broken code. This course teaches the workflow, the judgement, and the safety practices that turn these tools into something you can actually rely on. Every module ends with something working in your hands, not just a page of notes.
Who it’s for
The student I had in mind is a hardware maker, an electronics hobbyist, or a technically-minded solopreneur. You are comfortable with microcontrollers, PCB design, and the basics of Python. You have ideas for tools and small applications, but you have always hit a wall when it comes to structuring, building, and deploying real software. You are not a professional programmer, and you have no interest in becoming one.
If you are a software engineer looking for advanced architecture patterns, this is not your course. It prioritises practical outcomes over theoretical depth, and it never apologises for that.
What you’ll build
You build four projects, and each one teaches a different shape of software using a different AI tool:
- A RC filter analyser that runs entirely in your browser, built with Claude in the browser at claude.ai, no editor and no terminal. It plots a Bode magnitude curve from your resistor and capacitor values and updates live as you type. This is your first taste of framing a problem clearly enough for an AI to build the whole thing in a single HTML file.
- A serial data plotter that reads live data from an Arduino over the serial port and plots several sensor channels in real time. It is a Flask application that streams readings to the browser with server-sent events, and it includes a mock serial mode so you can develop without hardware on the bench. I built it inside VS Code with a coding agent, Cline or Claude Code, and drove it through the full workflow from brainstorming to deployment.
- A datasheet question-and-answer tool that lets you upload a PDF and ask it plain questions, such as “what is the maximum supply voltage?”. It uses retrieval-augmented generation so the model answers from the document itself rather than guessing, with an LLM wired in at runtime through its API. The agent built it, and the lectures show where an LLM genuinely earns its place in an architecture and where it does not.
- A multi-user resource booking system as the capstone, a full web application for a makerspace or lab. It runs on Flask with a PostgreSQL database through the SQLAlchemy ORM, handles user accounts and booking conflict detection, and deploys to a real server. This is a multi-file project I built with the agent, the same kind of stack a solopreneur ships to paying customers.
What’s inside
The course runs across six modules plus a closing debrief:
- Module 0: Setup and orientation. Install the tools, understand what AI assistants can and can’t do, and get oriented to the model landscape before you write a line.
- Module 1: AI coding fundamentals. The conversation loop, the five levels of AI-assisted coding, how to frame a problem so an AI can solve it, and the failure modes to watch for. You ship the RC filter analyser at the end.
- Module 2: AI tools and models. VS Code as your base, the coding agents Cline and Claude Code, how to treat the LLM as a separate concern, and choosing between local and cloud models.
- Module 3: The AI-assisted workflow. The full cycle from brainstorming and specification through implementation, testing, iteration, documentation, and deployment, applied end to end to the serial data plotter.
- Module 4: AI in your IDE. When to put an LLM at runtime, API basics, retrieval-augmented generation, preparing PDFs for it, and managing cost, latency, and graceful failure, applied to the datasheet tool.
- Module 5: Ship a real project. Multi-user application design, database design with an ORM, building a multi-file project with an agent, external APIs and background jobs, and finishing and shipping the booking capstone.
Each lecture runs between five and fifteen minutes. I built it that way on purpose, so you can make genuine progress in the gaps between everything else you do. You do not need a free afternoon to move forward.
What you’ll be able to do
By the time you finish, you will be able to:
- Frame a software problem clearly enough for an AI assistant to solve it
- Build and deploy a real web application using AI-assisted development
- Recognise and recover from common AI coding failures before they become costly mistakes
That last skill matters more than it sounds. These tools fail in specific, recognisable ways, and knowing how to catch a wrong pin mapping or a hallucinated function before it reaches your hardware is the difference between a useful assistant and a liability.
Enrol at the launch price
To mark the launch, the course is 30% off for a limited time. The discount applies automatically at checkout during the promotional period, so there is nothing to type, though the code is AISHIP if you want to confirm it on the order summary. The offer runs until Monday, 6 July 2026, after which the course returns to its standard price.
If you have questions or comments about this course, please leave them below. I read every message, and I would rather answer your question now than have you guess.