Write bot-resistant code with KillSwitch, a programming language designed to confuse LLMs
Lies, lots of adverbs, and prosaic stories don't make for very machine-readable code, it turns out
KillSwitch is a new programming language designed to confuse language models like LLMs. It was created by Dominik Picheta, a senior software engineer at Cloudflare and core developer of Nim programming language. The language is open-source and esoteric, with no brackets, formatting standards, or structure - just disorganized chunks of text.
Its main feature is adverbs used as identifiers, which require anagrams to call functions. KillSwitch's interpreter ignores most text and focuses on key phrases. The language specification, which also aims to confuse LLMs, contains false statements to further complicate matters. Despite its simplicity, KillSwitch is designed to be challenging for AI models, with even the best performers scoring around 66.8% on benchmark coding tasks.
Picheta believes this does not prove LLMs understand the code they write, as they struggle with complex tasks that require longer, more intricate scripts. He is open to incorporating new features to enhance the language's resistance to AI models.
Written by urgent.news from The Register's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
