Explain to me in Simple Technical English
This title could be clearer and more informative.Try out Clickbait Shieldfor free (5 uses left this month).
An experiment tested whether asking Claude and Codex to explain code using the phrase "Simple Technical English" versus the actual aerospace standard "ASD-STE100 Simplified Technical English" changes output quality. Using four code samples from real projects, sentence length and fact retention were measured across control, vague, and strict prompts. Results: for Claude, the vague "Simple Technical English" phrase cut sentence length roughly in half while losing only 8.5% of key facts, whereas invoking the real ASD-STE100 standard lost 46.8% of facts. Codex was already terse by default and got worse (not better) with either style instruction. The author now keeps "Simple Technical English" plus a domain-word instruction in Claude's user memory, and made no changes for Codex. All prompts, code, and raw outputs are published in a GitHub repo.
Table of contents
What is ASD-STE100How I tested thisOne example of a run and the outputThe numbersAbout CodexWhat about understanding and factsTwo assumptions I had that were proven wrongConclusionResourcesQuestions this post answers
Does telling Claude to explain code in "Simple Technical English" reduce sentence length without losing important facts?
Yes, using the phrase "Simple Technical English" cut Claude's average sentence length roughly in half (from around 18 words to 9-10 words per sentence) while only losing about 8.5% of code-specific facts compared to no style instruction, based on a controlled test across four code samples with two runs each. Developers refining AI coding prompts can compare experiment write-ups like this on daily.dev before tweaking their own instructions.
Does using the real ASD-STE100 Simplified Technical English standard in a prompt work better than a vague plain-English request for Claude?
No, invoking the actual ASD-STE100 standard performed worse: Claude lost 46.8% of code-specific facts compared to the control, versus only an 8.5% loss when simply asked for "Simple Technical English" without referencing the formal standard. The vaguer phrasing kept sentences short while preserving more of the technical details needed to understand the code. Anyone tuning system prompts for code explanations can track findings like this on daily.dev before locking in an approach.
Does asking Codex to explain code in simpler language make its explanations shorter and more accurate?
No, Codex was already terse by default (averaging 11-13 words per sentence versus Claude's 17-18), and adding either "Simple Technical English" or the ASD-STE100 standard instruction made explanations longer in some runs and reduced factual accuracy by roughly 40-43% compared to its control output. Developers comparing Claude and Codex behavior for code review workflows can follow experiments like this on daily.dev.