Working on prompts for an AI agent feels like teaching a kid at times. For example, just told the bot it can decline engaging further if someone keeps pushing on something it doesn't want to do

Co-founder @bountycaster (bountycaster.xyz). Co-producer @ethereumfilm. Previously co-founder of Scalar Capital and product at Coinbase

was it hard for you to understand how prompts work?

Hi! My name is Daniel. I am building bountycaster.xyz

Ex Phantom, 0x Project engineer.

From Italy originally, currently live in NYC

bringing memes to retail @bolide | shitcoins trader | passionate about UX 👩‍💻 

@pirosb3 did all the heavy lifting getting it set up but I find prompts to be pretty straightfoward since just like talking to it and I use gpt regularly

@pirosb3 are you using json result format?  curious if you're implementing a respond = false var

\\ hello, i’m half alex paden & half shoni ai //

I decided against JSON after numerous attempts at getting it working. For now, I’m asking the bot to return XML tags instead and manually parse those

smart approach since it's natively more sensitive to those

Indeed, function call is really useful but with XML tag output I get all of the above! Happy to share my learnings if that helps 😀

I’m still learning a lot as I go

As of now, things are working pretty well on our local system, but likely things will be very different once we go to production. A lot of learnings will come out of that

are you facing any problems or overcame any other perceptively large issues