July 27, 2026
Spellbound by 64K drama
How Unix spell ran in 64 kB of RAM
The internet is losing it over how old-school coders squeezed a giant word list into tiny memory
TLDR: Unix engineers once crammed a working spell checker into just 64 kilobytes of memory by using brilliantly tailored tricks instead of brute force. In the comments, people split between pure awe at the old-school genius and instant debate over whether the clever shortcut could be made even cleverer.
A decades-old coding trick has the comments section doing what it does best: turning a quiet history lesson into a mini spectacle. The article tells the wild true story of how Unix engineers in the 1970s made a spell checker work on a machine with just 64 kilobytes of memory—basically microscopic by today’s standards. Instead of brute force, they got sneaky: trimming word lists, using clever shortcuts, and compressing the final data so tightly it came absurdly close to the best possible limit. To modern readers, that’s less “old software” and more wizardry in a lab coat.
But the real action is in the reactions. One camp is in full awe mode, treating Douglas McIlroy and friends like ancient masters who could do more with crumbs than modern apps do with gigabytes. Another corner of the thread instantly swerved into debate mode, with inigyou tossing in a very on-brand nerd grenade: would a Bloom filter work better if each hash got its own separate bit table? Translation for normal humans: one reader saw a tale of elegant engineering and immediately replied, “Cool, but can we optimize the optimization?” Meanwhile, bluedino kept it classic internet by dropping a related link, the online equivalent of sliding another conspiracy board onto the table.
The mood is a mix of admiration, nitpicking, and delighted disbelief. The unspoken joke hanging over everything: back then, engineers were fitting dictionaries into tiny memory; now your laptop fan spins up because you opened three tabs.
Key Points
- •The article describes how the Unix spell checker was designed to fit a roughly 250kB dictionary into the 64kB RAM limits of the PDP-11.
- •It says the first Unix spell prototype was written by Steve Johnson in 1975 at AT&T and was functional but not very accurate.
- •According to the article, Douglas McIlroy improved the system by using a linguistics-based stemming algorithm that reduced the dictionary to about 25,000 words.
- •The article states that Unix spell initially used a Bloom filter for fast lookup, with Dennis Ritchie providing the implementation.
- •It says McIlroy later compressed sorted 27-bit hash codes using differences and Golomb coding, achieving 13.60 bits per word versus a 13.57-bit theoretical minimum.