A blog post arguing that AI coding tools aren't the industry's real problem topped Hacker News this week - the real problem, it says, is that nobody remembers why systems were built the way they were.
The essay, posted to the personal blog ssp.sh, reframes the current anxiety over AI generated code. Its argument: code was never the hard part of software engineering. The hard part was always understanding system architecture and the intent behind past decisions - and that institutional memory is eroding regardless of who, or what, writes the code. The post struck a nerve on Hacker News, climbing to 258 points and drawing 170 comments.
This lands squarely in an ongoing industry debate about what AI assistants actually change. If the essay's framing holds, the risk isn't sloppy AI output - it's teams shipping code faster than they can explain it, AI assisted or not. That's a governance and onboarding problem, not a code-quality one.
Plenty of engineering orgs already struggle to document why instead of just what - AI just raises the stakes on a habit they never fixed.