AI Pulse · Under the Radar

An essay showing up on Hacker News this week made a simple claim: Opus 5 feels worse to work with than the model it replaced. Not slower, not less capable on paper, just worse to actually sit down and use. The author’s argument is that Opus 5 asks fewer clarifying questions and guesses more confidently when your instructions are ambiguous. No benchmarks, no test suite, just one person’s read on how the model behaves day to day. The author is upfront that this is speculation, not proof.

What’s interesting isn’t the claim itself. It’s that the thread pulled in 545 points and over 500 comments, which puts it among the biggest discussions on the site this week. That kind of volume usually means a lot of people recognized something in their own work, even if nobody can point to a number that proves it.

Here’s why this matters if you’re running a business and using these tools daily. A model that guesses at what you meant instead of asking is faster when it guesses right and expensive when it guesses wrong, especially on tasks where the ambiguity was the whole point of asking a human to review it. If your team has noticed a similar shift, that’s worth flagging internally now, before it costs you in redone work or decisions made on the wrong assumption.

There’s no benchmark data here and no way to verify the cause. What we have is a strong signal that practitioners are noticing something and talking about it in numbers big enough to matter. Tracking behavior shifts like this across the tools your team relies on is exactly the kind of thing we build into an AI command centre, so you catch the pattern before it becomes a habit nobody questions.