AI Pulse · Under the Radar

A small Claude skill called i-have-adhd is getting attention because it tackles a familiar AI problem. You ask a simple question, and the agent responds with context, caveats, a plan, and several paragraphs before it tells you what you actually need to know.

The skill pushes agents to put the direct answer first and cut the preamble. It reached 156 points and 128 comments on Hacker News’ front page on the same day, which suggests the frustration is widely shared. This is not really about ADHD as a label. It’s about respecting the user’s attention. Most people using AI at work aren’t looking for an essay. They want the answer, the next action, and any important exception.

For business owners, this matters when you’re designing AI assistants for real teams. An assistant that produces thorough but slow answers can become a drag on adoption. People stop asking it questions if they have to scan five paragraphs to find the one useful line. Set a default response style that starts with the conclusion, then gives detail only when it helps. Let staff ask for deeper analysis when they need it.

That terse-output instinct is already central to how we think about practical AI systems. This is the kind of thing we build into an AI command centre, so different users get the level of detail they need without having to fight the assistant for it.