August 13, 2026
Math doomers caught in 4K
NP-Overrated
People are roasting the ‘impossible problems’ myth and saying class lied to us
TLDR: The article says many “impossible” computer problems are only terrible in rare edge cases, and modern tools solve useful versions every day. Commenters mostly agreed—but argued hard over whether that’s thanks to genius algorithms or because engineers simply fence off the nightmare scenarios.
A spicy little nerd revolt is brewing around NP-Overrated, a piece arguing that the scary “some problems are basically impossible” lesson many students got in school is wildly overblown. The author says real-world software doesn’t usually hit the nightmare cases, and points to big wins in package installs, scheduling, route planning, and even the famously nasty logic-puzzle world behind SAT solvers. Translation for normal humans: some tasks are theoretically awful, but in everyday life, clever tricks make them work shockingly well.
And the comments? Absolutely not calm. One camp basically yelled, “You learned the wrong lesson!” with commenter porridgeraisin insisting the theory never said all useful cases were doomed, only that there’s no magic fix for every possible input. Others piled on with a more practical vibe: most real jobs never wander into the explosive worst-case swamp anyway. andrewla and WCSTombs stressed that humans often reshape the problem so the computer avoids the dumbest dead ends.
But then came the reality-check brigade. Guvante argued the article skipped the biggest secret of all: industry often just bans the dangerous cases entirely. That sparked the juiciest tension in the thread—is this a heroic story about smarter algorithms, or a sneaky story about refusing to play on hard mode? Meanwhile, jvanderbot delivered peak comment-section swagger by saying fancy math can get “obliterated” by a few brutally simple passes that computers love. In other words: the professors brought apocalypse; the commenters brought receipts, side-eye, and optimization gossip.
Key Points
- •The article argues that NP-hardness is commonly misinterpreted as meaning practically useful algorithms cannot exist.
- •It states that complexity results only guarantee bad performance on some inputs, not necessarily on the inputs that matter in real applications.
- •The article names dependency resolution, some forms of type checking, scheduling, traveling salesman, and SAT as prominent NP-hard problems.
- •It claims optimization problems such as scheduling and traveling salesman can often be solved optimally in reasonable time, and cites a reported 450-billion-fold algorithmic speedup from 1991 to 2015.
- •The article uses Amazon's reported solving of a billion SMT problems per day as evidence that satisfiability-related problems are handled at industrial scale.