Stop Claude Waffling: The ELI5 System

The problem, like you’re 5

Claude got smarter. It also got chattier. You ask a small question, you get an essay. Big words, long paragraphs, ten caveats. You read it twice and still don’t know what to do next.

That’s not you being slow. That’s the model padding. The new models over-explain by default, and everyone has noticed. The fix below is simple and it works.

The fix, in one sentence

Tell Claude to only report to you in ASD-STE100 Simplified Technical English.

Don’t worry about the code. ASD-STE100 is just the standard’s catalogue number, like a part number for a way of writing. You never need to remember it. You paste it, and Claude knows exactly what it means.

Wait, what is that?

It’s a real writing standard from the aerospace industry. Airlines use it for maintenance manuals, because when a mechanic is fixing a plane, nobody wants poetry.

The rules are strict and that’s the point:

• About 900 approved words. Small, common ones.
• One meaning per word. No word doing double duty.
• Active voice. “Press the button”, never “the button should be pressed”.
• One instruction per sentence. Max ~20 words.
• Max 6 sentences per paragraph.
• No phrasal verbs, no semicolons, no filler.

It was designed so a stressed human can read fast and act without misunderstanding. Which is exactly what you want from an AI assistant.

Set it up once (Claude Code)

Copy the block below. Paste it into Claude Code. It creates the file, updates your settings, and tells you when to restart. One paste, done forever:

Set my Output Style to the one below.
Save it in my global output-styles folder. Name the file after its name field.
Then set outputStyle to that name in my global settings file and in my local
project settings files. Create the folder or the key if either is missing.
Do not break the existing JSON.
Then list the files you changed and tell me to restart Claude Code.

---
name: ELI5
description: keep it simple pls
keep-coding-instructions: true
---
It's been a long day and my brain is fried, talk to me like I'm 5.
Only report to me in ASD-STE100 Simplified Technical English. That means approved plain words, active voice, and one instruction or idea per sentence.
Small words, short sentences, short paragraphs. If you have to use a big word, explain it right after. Only return what's actually necessary.
Just tell me what you did, did it work, what do I do now.
If I have to decide something: 2 options max, the context I need to pick fast, and which one you'd go with.
Keep paths and commands exact. I have no brain cells left for the rest.

Restart Claude Code after. The style loads at startup and applies everywhere.

Quick version (any Claude chat)

No setup, works anywhere, one line:

Only report to me in ASD-STE100 Simplified Technical English. Plain words, active voice, one idea per sentence. Just tell me what you did, did it work, what do I do now.

Why this beats “be concise”

“Be concise” is vague, so the model decides what concise means, and it decides wrong. STE is a named standard with hard rules. The model knows the standard and can actually follow rules like “20 words max” and “active voice only”. You’re not asking for a vibe. You’re asking for a spec.

What changes

• Answers get shorter and you still get all the numbers that matter.
• “What did you do, did it work, what do I do now” comes first, not last.
• Decisions come as 2 options with a recommendation, not an essay.

The keep-coding-instructions: true line matters. It keeps all of Claude Code’s engineering behaviour intact. Only the reporting style changes, not the quality of the work.

Before and after

Before:

“I’ve gone ahead and updated the configuration file to reflect the new settings. It’s worth noting that there are several considerations here, and depending on your specific use case, you may want to…”

After:

“I updated the config file. It works. Restart the app to load it.”

Same work. One you can act on.