July 8, 2026
Autocorrect, but make it chaos
Unicode's transliteration rules are Turing-complete
Your keyboard’s boring text rules might secretly be a computer, and commenters are losing it
TLDR: A researcher showed that Unicode text-rewrite rules can be powerful enough to act like a real computer, turning a humble text feature into a surprising math machine. Commenters swung between delight and existential exhaustion, joking that every file format now seems one step away from becoming its own programming language.
The big reveal here is wonderfully chaotic: a researcher says a part of Unicode — the giant standard that helps computers handle text and languages — can do full-blown computation. Not just swap an accented letter for a plain one, but keep rewriting text in a way that can simulate real programs. In plain English: the same kind of rule system used for language and typing tricks may be powerful enough to run logic puzzles forever, and the internet immediately responded with a mix of awe, dread, and "of course it does."
The Hacker News discussion quickly turned into a support group for people realizing that yet another innocent-looking format can secretly become a computer. One of the loudest vibes was pure resignation: "At this point it feels more difficult to ensure that your format cannot compute than to ensure it can," joked one commenter, basically summing up the entire thread’s mood. Another instantly jumped to the obvious next act of madness: someone should build a compiler for this thing, because if the internet discovers a weird machine, it must try to make it worse.
And then came the memes. People compared it to Microsoft Word autocorrect somehow being able to compute too, which only added to the sense that your documents, keyboards, and text tools are all one bad idea away from becoming accidental programming languages. There wasn’t much outright fighting — more a rolling wave of nerdy disbelief — but there was one practical mini-drama: does this work everywhere people actually use it, or only in one Python library? Even that question fed the suspense, because if the answer is "modern operating systems too," then this stops being a quirky math party trick and starts sounding like your text stack has been hiding a tiny chaos engine all along.
Key Points
- •The article argues that Unicode UTS #35 transliteration rules are Turing-complete under unbounded semantics, unlike bounded core Unicode algorithms.
- •Transliteration rules support ordered rewrites with contexts and a cursor-revisiting feature that can cause newly written text to trigger further rules.
- •The proof approach compiles a universal 2-tag system into transliteration rules, relying on prior universality results for 2-tag systems.
- •A worked example uses Liesbeth De Mol’s 2-tag encoding of the Collatz function with productions `a → bc`, `b → a`, and `c → aaa`.
- •The article shows the construction running both through Python with PyICU and through ICU’s stock `uconv`, with an execution trace matching expected tag-system states.