Mullet is a deliberately satirical AI coding skill that turns the usual goal of simple, maintainable software on its head: it instructs coding agents to make working applications maximally complex. Its premise mocks the use of lines of code as a measure of AI productivity and the claim that ever more elaborate architecture protects jobs. The repository says its FizzBuzz and to-do examples already contain 257,083 lines, use 200 abstraction layers by default, and route ordinary operations through contracts, committees, and multiple runtimes. The software must still work and pass tests; complexity is treated as the achievement, not a substitute for functioning behavior.
The examples show how the joke is implemented. FizzBuzz has three competing implementations, chosen at random, and sends results through a Kafka-style log, hundreds of hearings, and a hash chain; printing a result can involve JavaScript, Ruby, AWK, Python, SOAP, XML, and JVM assembly. The to-do app makes marking “Buy milk” complete depend on leased portions of an X, royalty disputes, and thousands of contracts. Mullet packages skills, hooks, editor rules, and integrations for several coding-agent hosts, with a task board for tracking spending in Claude Code. It also offers 46 configuration inputs for generating design briefs and a “mulletize” mode that repeatedly expands an existing application, verifies its behavior, and checks that its source code grows.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.