the golden age of bad software

a couple months ago microsoft decided to update the beloved notepad app by adding features nobody asked for. this was the 14th update in the past 12 months. for context, in the entire decade between 2010 and 2020, notepad received a grand total of 1 update.

notepad was perfect. it opened instantly, it edited text, it never crashed. it did one thing and did it well. now it has tabs, ai integration, and a rewrite that broke shellexecute for thousands of developers who relied on it as a lightweight text handler.

this is the story of modern software in miniature.

the dichotomy

we have type systems that catch bugs before they ship. ci pipelines that run thousands of tests in minutes. ai that can generate entire features from a description. by every measure, building software has never been easier.

and yet the software we ship is worse than it was ten years ago. apps crash more. pages load slower. updates break things that worked fine yesterday. the gap between what we could build and what we actually build has never been wider.

the reason isn't technical — it's cultural. the industry optimized for speed of delivery over quality of delivery. "move fast and break things" became doctrine. technical debt stopped being something you paid down and became something you raised your next round on.

good enough won

somewhere along the way, the industry decided that software doesn't need to be good. it needs to be first. it needs to be updated. it needs to have a changelog that justifies the headcount.

notepad didn't need 14 updates. but someone needed to ship 14 updates. that's the difference. the product wasn't driving the roadmap — the org chart was.

this plays out everywhere. every app on your phone is slower than it was three years ago. every website loads more javascript than it needs. every service has more features than anyone uses. and every update introduces bugs that the previous version didn't have.

what ai changes

ai code generation makes this worse and better at the same time. worse because it's now trivial to produce massive amounts of mediocre code — the 14 updates become 40. better because the bottleneck was never writing code. it was understanding systems deeply enough to know when to stop.

the engineers who will matter most in the next decade aren't the fastest coders. they're the ones who look at notepad and understand that the best update is no update at all.