Skip to content
Lewis Crook.Get the invite

How I run my Claude setup · Problem 1 of 11

Claude sounds sure when it's guessing.

Claude gives you a number and a source in the same confident tone whether it looked it up or pulled it from memory, so you can't tell which one you've got, and you usually only find out when someone else spots it.

Lewis Crook · Written November 2026

Try it on your own Claude

Honestly this is probably the biggest one. In Stack Overflow's 2025 survey, 66% of developers said AI solutions that are "almost right, but not quite" were a frustration, and the confident tone is what catches you out, because you only find out when someone else spots it. Anthropic's own advice is to give Claude permission to say it doesn't know, and that's where my setup starts.

Try this in a new chat
Roughly what share of UK small businesses were using AI in 2025? Just the number and where it's from.

If you get a number and a source but it doesn't say whether it actually checked that source in the chat, or that it's from memory and might be out of date, it's sounding sure about a guess.

What my setup does about it

In my setup it has to say what it actually checked in the chat and what it's guessing, and it's allowed to say it doesn't know, and that's in the instructions block in the kit.

I answer questions about this in a small free WhatsApp group. Get the invite.

If any of this was useful, message me.

Message me on LinkedIn Get the invite