2025-06-21 The “My Mate Dave” Test for Reasonable LLM Usage

Published on 2025-06-21

I think LLMs are powerful tools with lots of potential for value. I think, like many tools, particularly powerful and novel ones, they’re frequently misused.

Per the Law of the Instrument / Maslow’s Hammer etc. - “if all you have is a hammer, everything looks like a nail”, and gosh a lot of folks seem to be taking a bunch of good swings with ChatGPT et al.

We’re seeing:

And none of that seems great.

Some of this is laziness, but I think a lot of this comes from folks genuinely not understanding or appreciating the limits of LLMs, and treating them like Magic Knowledge Boxes rather than Fancy Autocomplete [1].

As such, I propose a rule of thumb to help folks identify appropriate LLM usage, which I’m calling The “My Mate Dave” Test [2] [3] [4]:

If you replace “ChatGPT” or “Claude” or whatever with “My mate Dave”, and it makes your plan/action sound ridiculous or terrible or at least somewhat concerning? Probably don’t do it!

Consider:

“I asked my mate Dave for some examples of case law” I’m sorry what?
“My mate Dave wrote this email / my performance review / that article I sent” you what now??
“My mate Dave says you should do X” gee thanks

Versus:

“I asked my mate Dave to read it and he had a few suggestions” ok cool!
“I was really stuck on the closing paragraph so I asked my mate Dave for some ideas and he helped me get it done” neat!
“I’ve got absolutely no idea here, but my mate Dave had some thoughts, and I’m not sure if they’re good or not, but they’re probably better than nothing” fair enough!

This won’t fix laziness, and it won’t fix not caring, but I hope it will help highlight some of the situations where people should really stop and think some more…

Note

To be clear, your mate Dave may be an absolutely fabulous chap - this isn’t about casting aspersions on them or their smarts! The point is that they’re a third party, and they’re an unknown quantity, and that is the bit that should be raising questions.