
It's rare that one change can make a world of difference in a dev environment.
Yet here we are, and I figure let me share. I generally look forward to Anthropic model releases. What's not to like about more intelligence on the same plan? The reviews on release day of those who were using Opus 5 also came with a warning that it was different. The advice was to delete your skills and run it on High or Medium effort.
Even with those changes it just felt harder to work with than Fable or Opus 4.8, but I kept at it until I decided to go back to Opus 4.8. It was a little hard to describe in one of my group AI chats. I just said "it was taxing to work with." It turns out that was because of the output of text and tech jargon within it. I switched to Opus 4.8 while I was working out why reasoning on the Gemma 3 27B kept failing, the Opus 5 output of what it was doing was dense and hard to follow.
I want to thank Lydia Hallie (Claude Dev) who posted on X about something I did not know called output-style settings. She provided a template prompt called "Explain Like I'm 5" or ELI5. I tried it and it made a difference so I am using Opus 5 again and I am passing it along for those who are not aware.
---
name: ELI5
description: keep it simple please
keep-coding-instructions: true
---
It's been a long day and my brain is fried, talk to me like I'm 5.
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.
Ask your instance of Claude to set up the output style for you if you are having that same taxing situation that is hard to explain or maybe for you very easy to explain.
I added this line to mine as well.
Only report to me in ASD-STE100 Simplified Technical English. That means approved plain words, active voice, and one instruction or idea per sentence.
Written by Duane Grey
AI Strategy & Implementation
Independent AI consultant helping companies cut through hype and deploy systems that produce real results.