The Skill That Explains Anything to Anyone
mlj.solutions
mlj.solutions Když se na totéž zeptáte technika a vedoucího provozu, dostanete dvě různé odpovědi a obě mohou být správné. Dobré vysvětlení se přizpůsobuje tomu, kdo poslouchá: manažera zajímá dopad a riziko, uživatele to, co se pro něj mění. Na tomhle principu postavil vývojář Andrew Ou nástroj ELI5 pro Claude Code. Model podle toho, komu vysvětluje, upraví slovník, přirovnání, tón i hloubku. Inženýrovi popíše architekturu, manažerovi dopad a riziko. V autorově malém testu splnila vysvětlení s tímto nástrojem 83 % hodnoticích kritérií, bez něj jen 42 %. Největší rozdíl byl u vysvětlení pro manažery, kde úspěšnost vzrostla z nuly na polovinu. AI umí vysvětlovat dobře. Nejdřív jí ale musíte říct, s kým mluví. A pokud tě zajímají novinky ze světa AI, nezapomeň nás sledovat!
View on InstagramWhat the study is about
ELI5 is an open-source skill for Claude Code that adapts an explanation to whoever it is meant for: a child, a manager, an engineer or parents. The author verified its benefit with his own tests against the base model without the skill.
Statistics
83.3%
success rate with the skill
With the skill, 10 of 12 checklist criteria were passed.
41.6%
success rate without the skill
Without the skill, only 5 of 12 criteria were passed, a difference of 41.7 percentage points.
0 to 50%
success rate for managers
For the manager audience, the success rate rose from zero without the skill to half with it.
What this means
- Same topic, different explanation, the skill adjusts vocabulary, analogies, tone, depth and framing to the audience.
- The biggest benefit is for non-technical audiences, the most notable improvement is for managers, who need impact and risk rather than technical details.
- A skill can be measured, the test script runs every prompt with and without the skill and scores the result against predefined criteria.
- Installation is simple, you just copy the skill folder into the skills directory in Claude Code.
- The results are indicative only, there are few tests and the evaluation was done by the author himself using Claude.
Our take
The same information needs a different language for a developer, a manager and a customer. With corporate AI assistants, we therefore expect the answer to be adapted to the audience.
Michal Dobrovolný
Founder & CTO
Zdroj
DreambigOu. (2026). ELI5: Explain Like I Am 5 [Computer software]. GitHub. https://github.com/DreambigOu/ELI5
Don't miss these posts
When More Agents Pay Off and When They Don't
More AI agents don't necessarily mean a better result. A study shows that what matters is how much information is lost when it is passed between agents.
AI Remembers a Claim Even When You Tell It It's False
When a model is fine-tuned on documents that deny a claim, it often learns the claim as true anyway. The authors named this effect Negation Neglect.
A Correct Answer Is Not the Same as a Verified Analysis
TrustDABench tests whether AI working over tables can refuse to answer when the data isn't sufficient. The best model reached only 24.21% in reliability.
Want to implement AI in your company?
We'll go through your needs together and propose an approach that delivers measurable results.
Write to us
