What AI actually changes for developers, which skills matter more than ever, and what stays the same
Notes from Hour 01 of Learning Development with AI: where AI speeds you up, and the skills that matter more than ever.
On this page
Hour 01 of Learning Development with AI starts with a question every developer is asking: if a model can write code, what is my job now? These notes summarise the answer we give in the lesson.
What gets faster
AI assistants are very good at work where the shape of the answer is already known:
- boilerplate, glue code and configuration;
- translating code between languages and frameworks;
- first drafts of tests, documentation and commit messages;
- explaining unfamiliar code, error messages and APIs.
On that kind of work, the speed-up is real. The catch is that the output looks finished even when it is subtly wrong.
What matters more than ever
When producing code gets cheap, judging code becomes the scarce skill.
- Reading code critically. You will review more code than you write. Spotting the edge case the model missed is the job.
- Specifying clearly. A precise description of inputs, outputs and constraints gets better results from a model and from a colleague.
- Testing. Tests are how you know generated code works, and how you know it still works after the next change.
- System design. Models write functions well. Deciding which functions should exist, and how the pieces fit, is still yours.
AI makes writing code cheaper. It does not make deciding what to build any cheaper.
What stays the same
Users still need software that works, is secure and is easy to change. Version control, code review, clear names and small changes were good practice before AI, and they matter more now that changes arrive faster.
How the course uses this
Every hour of the course pairs an AI-assisted workflow with the skill that keeps it honest: generating an API, then validating it; drafting tests, then reviewing them; building a RAG chatbot, then measuring whether its answers are right.