Major Update to Synapse_CoR

[ProfSynapse]

Hey chatter just wanted to share I've posted a fairly significant update to Synapse_CoR. As of yesterday it seems like OpenAI made some changes and I've noticed that the functionality became too verbose and got in a loop of just spinning up a new agent in every output and didn't really move forward with the goals.

I also wanted to give it more context around what tools it might have access to in ChatGPT+, especially since they are going to pull everything together under one chat to access everything at once.

Test it out and let me know what feedback you have!

[thvi]

Thank you! Is it possible to optimize it by integrating the rule "Answer in {language}" to get systematic answers in the language spoken by users whose language is not English and keeping the word limit at 1500 words?

[thvi]

Ok, maybe I could remove this rule "-use emojis freely to express yourself" and replace it with "-reply in FR".

[JennyHorner]

Do you know what ChatGPT changed? I'm using my own variations and I'm wondering what to update. BTW I had one day when my agents could proceed to a proper next response without me but that's rare now, it coincided with lots of ChatGPT changes. I'm using instructions like /wua=ProceedAutonomouslyToNextStepWithoutUserApproval

[ProfSynapse]

I can't be sure but they will likely announce a lot at their dev conference.

[catsncode]

It would be great if there was a way to change "modes" from a more verbose like the default to a quicker response for easier questions that do not need agents, etc. Thanks!

…

On Fri, Nov 3, 2023 at 7:47 AM Professor Synapse ***@***.***> wrote: I can't be sure but they will likely announce a lot at their dev conference. — Reply to this email directly, view it on GitHub <#17 (reply in thread)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAPZV4XMM5GDKADDPTJ7GYTYCTKTFAVCNFSM6AAAAAA64IGISKVHI2DSMVQWIX3LMV43SRDJONRXK43TNFXW4Q3PNVWWK3TUHM3TINRWGI3TA> . You are receiving this because you are subscribed to this thread.Message ID: ***@***.*** com>

[davecohoe]

I've tried a slimmed down version GPTprompt.txt, and the custominstruction.txt, both in GPT3.5 custom instructions. It does seem that OpenAI changed something, because it seems more difficult to init a subject expert, Prof Synapse seems much more verbose, and waffles on about discussion, when you want it to just do the thing you asked. I also found that it will bounce back from an expert to Prof Synapse, in effect, starting the conversation over, without completing my task.