Code Is Cheap Now. Software Isn't.
AI crushed the cost of writing code — not the cost of understanding a problem. Why we're entering an era of personal, disposable software, and why lasting systems still need real engineering.
The rise of AI coding assistants like Claude Code, Copilot, and Cursor has fundamentally changed what it means to build software. But the implications aren't what many people expect.
The barrier to entry has collapsed. Tools that once required years of learning can now be approximated in an afternoon. Non-developers are becoming architects of their own solutions — spinning up custom subscription trackers, niche Chrome extensions, and personalized fitness apps tailored exactly to their needs. Software is shifting from a commodity you purchase to a personal utility you generate.
We're moving from SaaS to scratchpads. Much of this new software isn't built to last. It's designed to solve a single problem, right now, and then be discarded. When creating a tool takes five minutes, permanence becomes optional. It's a return to the original spirit of spreadsheets: reason through a problem, get your answer, move on.
But here's the catch: code is cheap, software is still expensive. LLMs have crushed the cost of generating lines of code. They haven't touched the cost of truly understanding a problem. The subscription tracker breaks when a bank changes its export format. The Chrome extension dies when a website updates its layout. Real software costs come from maintenance, edge cases, UX debt, and data ownership — not the initial write.
Engineering isn't dying. It's evolving. The value of an engineer is shifting from the how of syntax to the what and why of systems. Architecture, abstractions, rate-limiting strategies, distributed caching, security — these still require human judgment. AI hides complexity; engineers manage it.
Distribution remains the hard part. With everyone able to ship functional apps quickly, noise has reached an all-time high. Claims of five-figure MRR on weekend projects are often marketing plays, not technical innovations. When code is no longer the bottleneck, success depends on taste, timing, and understanding your audience — things that can't be automated.
The hype suggests we're entering a golden age of SaaS. We're not. We're entering an era of personal, disposable software — where anyone with twenty dollars and a few hours can ship something functional. But building systems that last, scale, and actually matter? That still requires the fundamentals of good engineering.
The tools have changed. The job hasn't.