In the Lab

How I Get AI to Stop Rambling

AI answers were becoming a chore to read. Adding “use ASD-STE100” to my prompts helped cut the clutter. Here’s the difference it made in my finance work.

I started to feel overwhelmed reading AI answers and documents, particularly from Opus 5 (the model that loves to talk).

Then I stumbled on an article from Claude pro Ruben Hassid at 'How to AI' about how he "de-slops" AI.

It's the best fix I've found to get AI to get to the point: "use ASD-STE100"

I add it to most of my prompts now. Like Ruben, I've found it just helps avoid the clutter that most models spew out so I can get to the point without the B.S.

What is it?

It's a simplified English writing standard. In the 1980's aircraft manufacturers wanted to make maintenance manuals easier to understand.

They basically made a playbook to make English more useful by capping it.

You can see the difference below from the prompt above about indirect cash flows:

Another example: I ran a prompt to review forecast assumptions, one where I included "use ASD-STE100" and the other where I didn’t:

*The one on the left I asked to use ASD-STE100 while the other doesn't*

Here's the difference between the HTML outputs. Guess which one used ASD-STE100:

Spoiler: the one on the left. Better right?

Where not to use it

Now, this is great for your own knowledge work but as you can imagine it kills creative content. If you're writing a poem, or a genuine email to a colleague, I probably would't use this.

I’m more selective when I want warmth, humour or a personal voice. For documents others will read, I’d use it to improve clarity, then check that the result still sounds natural.

Note: Ruben created a skill based on ASD-STE100 principles, which you can find in the link above, which I've started using and really like.

Try it out this week on a piece of work.

Try it out now by taking an AI answer you found hard to read and ask: “Rewrite this using ASD-STE100. Keep the key facts, assumptions and caveats.”

Timon

Smarter model or more effort? A finance guide to picking AI settings

How I pick between a smarter AI model and a higher effort setting, with real finance examples of when each one actually fixed the output

Ship Faster Without Lowering the Bar

Old me would have disappeared for three days, built the pack across every entity, made it beautiful, and sent it Thursday night. Then he'd have told me the second half wasn't what he meant. Instead he had a first draft within the hour. Half a day of actual work, all in.

I Started Asking AI to Ask Me Questions, and the Work Got Easier

I've started ending my prompts with "ask me clarifying questions first." Two minutes of questions saves twenty of rework.

From Preparer to Reviewer: A Simple Framework for Building Claude & Codex SKILL.MD Files That Actually Work

Most skill.md files I've seen (including my early attempts) are mediocre. The difference between a skill you abandon after two runs and one you rely on every month comes down to how you build it. Here's the framework I use.