The checklist didn’t work so it wasn’t useful for working through the suggestions and tracking my progress. Clicking the checkboxes did nothing, no toggle.
My guess is that markdown supports some notation that lets you generate checkboxes, and it “accidentally” did it, rather than a new feature, but it was interesting to see it happen and not something I’ve seen before so I thought I would share it with the community.
I use ChatGPT with 4o model everyday and very often. Don’t remember if I have seen a checklist before.
This is just speculation.
Before a new model is made public, meaning it can be selected or used specifically, some ChatGPT conversations or maybe just a specific request is sent to the new model and then if it is a reasonable reply returned, otherwise the result of the current model is used.
This looks like one of those cases based on the limited information given. If you search this forum you might find other such oddities noted soon before a new model is released.
I have been getting asked to rank two replies on an increasingly common scale though, but to my memory, this was just something it emitted in normal conversation flow.
