5/ (PPS: evergreen caveat that how to analyse the responses from public consultations isn't really anything to do with why they are often ultimately not very useful)

4/ (PS: I don't know if/how much this tool is actually used, it is explicitly labelled as being in research and not as a thing people should actually run, so this is probably not an alive problem)

3/ Not to mention the more general risks of prompt injection which I'm sure the team have done some thinking about but as a problem is (i think) beyond the state of the art to prevent

2/ But I do wonder: if these prompts are public and known to people writing submissions, does that

(a) help to make sure the submissions are better and more useful?
(b) facilitate the gaming of the systems by allowing the analysis to be gamed?

(probably both)

Yes 😄

I also noticed that the council said it was a "minor contravention" that the recipient "agreed not to repeat" as opposed to giving a wholehearted apology for dubiously issuing it in the first place 🤔

Quite!

Agree! How to get the cave party going is the question though, when for so many people going in _any_ of the caves is too scary

Yes, definitely! But the "from their point of view" is an _extremely_ important caveat

No, on this I must stand firmmmm

Same!

(It made me ragequit the radio)

Utterly grim wasn't it

Mm yes, I think the great GDS redirecting machine has quite a few "near enough" decisions in it!