10. The Future of Software Engineering¶
Video map
Section 2 — Future of Software Engineering
You have finished the fundamentals. Before building bigger things, a short chapter on the question every student asks: "Will I still have a job?"
What is changing — and what is not¶
| Changing | Staying |
|---|---|
| Who types the first draft of routine code | Deciding what to build and why |
| Time spent on boilerplate, config, wiring | Understanding the problem and the users |
| The tools: editors → IDEs → CI → agents | Judgement: design trade-offs, risk, security |
| How fast one person can ship | Responsibility for correctness and maintenance |
Tooling has changed many times: from text editors to IDEs, from copying files to a server to automated pipelines, from raw SQL to ORMs. Each shift removed a layer of drudgery and raised the level at which engineers work. Agents look like the next such shift.
The two bad reactions¶
- "AI will do it all; I don't need to learn." Then you cannot tell good output from bad. Reviewing code well requires more fundamentals, not fewer.
- "AI code is spaghetti; I won't touch it." Bad code has always existed — long before AI. The answer to bad code is review, tests and refactoring, and that applies to human code too.
The productive position sits between them: use the tool, and stay responsible for the result.
Skills that gain value¶
- Reading code quickly and critically.
- Breaking problems down into small, verifiable steps (this is exactly how you prompt well).
- Testing — tests are how you trust output you did not write (see chapter 16).
- System design — structure, boundaries, data models, failure modes.
- Security awareness — knowing what to distrust.
- Communication — turning a fuzzy business need into a precise request.
A habit worth building
After every task, ask: "What would I have to know to catch this if it were wrong?" That question tells you exactly what to study next.
A practical stance for students¶
- Learn the fundamentals deliberately, even when the agent could do it. Write some code by hand every week.
- Build in public with your own projects, not only tutorial copies.
- Keep a portfolio of reviewed work — projects with tests, a README, and commits that show your thinking.
- Be the person who can explain every line in a pull request, regardless of who typed it.
What's next¶
The remaining parts use Claude Code to take an idea all the way to a deployed product: planning (Part 3), building and testing, adding AI features, integrating email, and shipping. Everything you learned here — small steps, memory, plan mode, context hygiene, checkpoints, review — is what keeps that larger project under control.
Check yourself¶
- I can explain, with the carpenter analogy, why agents do not remove the need for engineers.
- I can name three skills that become more valuable with agents.
- I have a plan to keep practising fundamentals alongside agent-assisted work.