Humanizer STE80

Rewrite AI-sounding text so a reader takes it in on the first pass and does not think "a chatbot wrote this". Keep every fact. Add nothing.

Why AI text sounds the way it does

AI text has two layers of problems.

The tells. These are habits that people now recognize as machine writing: "It's not just X, it's Y", a one-line paragraph that restates the point, three adjectives in a row, an em dash in every sentence, "delve", "robust", "seamless", "Great question!". Wikipedia editors collected these patterns in "Signs of AI writing". Part 1 lists them.

The structure. Under the tells, the sentences are long, carry three ideas, hide the actor in the passive voice, hedge with "could potentially", and swap synonyms for variety. If you remove the tells but keep this structure, the text still reads like AI. Part 2 fixes the structure.

For the structure, this skill uses a method from aerospace. ASD-STE100 (Simplified Technical English, STE) is the controlled English of aircraft maintenance manuals. It forces one meaning per word, one idea per sentence and no decoration, because a misread step on a jet engine can kill someone. On 2 October 2026, Andrej Karpathy wrote:

Ask your LLM to explain something in ASD-STE100, it's a controlled language specification originally developed for aerospace maintenance documentation. LLMs well-versed in this language and it comes with heavy constraints on clean writing style that I often find a lot more readable. Sometimes I've tried to soften it a bit e.g. ask for "80% of the way to ASD-STE100" because the spec is quite stringent.

"80% of the way" is the core of this skill. Take the structure of STE fully. Take its vocabulary only as far as it sounds natural. The result reads like a good engineer or a good journalist: plain, exact, with a human voice left in.