Major variation in Formless

  • 1 November 2023
  • 4 replies

Hi there! I’m loving experimenting with Formless, but I’ve come up against a big issue. When I test out my “form” in the creator “app”, the AI’s responses are rich and nuanced. When I test it using the link on, say, my iPhone, sometimes it’s comically bad. For example, I’m looking for appreciative and critical feedback on something, and the AI - when accessed by phone - will say, “Input appreciative feedback” or something similarly non-conversational. Whereas in the creator app, it will say, “Tell me something you appreciate about X...” Is this the difference between GPT3.5 and 4? If so, it’s a much bigger discrepancy than i would have imagined. Thanks!

P.S. Is there a special forum somewhere for Formless beta testers to share experiences? Thanks!

4 replies

Userlevel 7
Badge +5

Hi @jonathanlal Oh this is an interesting difference - thanks for noticing and asking here. 

@cate.anti or @dokuniev do either of you happen to know? 

Userlevel 3

Hi @jonathanlal,

firstly I’m glad you’re trying our new product, and thanks for the feedback.


We’ve not experience differences among mobile and desktop, and this shouldn’t be the case. Are you testing the conversation in preview mode in the Desktop? Could it be that not all the changes have been published, therefore when trying the link in the mobile, the behaviours are different?

If this is not the case I would kindly ask to share the conversation URL so that we can investigate further and try to reproduce the issue.


Feel free to email us to


We don’t have a specific section of the community for Formless yet :) 

Thank you. I just tested it again on my phone and didn’t run into this issue. I don’t believe it was a failure to “publish” since what I was getting was verbatim what I had written in “info to collect”. Like, “Strength 2:” or something like that. Seems to be okay now, but if I run into it again I’ll send a screenshot.

BUT I have two other questions!

  1. Is there anything I can do to encourage the AI to ask follow ups when answers are too brief / not detailed enough (without having to program in follow-ups by adding too many elements in “info to collect”)?
  1. How can I stop the “reflecting” langauge (e.g., I respond, “ABCDEFG” and the AI replies, “Thank you for sharing ABCDEFG” before asking the next question). I find this clunky but can’t figure out how to stop it! Right now, it’s happening on my desktop but NOT on my phone. I’ve set the tone to, “concise, professional, neutral”.


@cate.anti or@dokuniev

I spoke too soon. Now it’s happening again, on my desktop...

Some screenshots.

It’s like sometimes it forgets to be an AI ;)