I am an entrepreneur (second-time founder), developer, and research engineer.
I am currently building ogre.run, an AI-based code dependency hypervisor - https://ogre.run
Hi! GPT 3.5 does work but its performance is not as consistent as GPT 4’s. We noticed that very often its response drifts away from what the prompt requested, generating invalid requirements files. GPT 4 = less prompt engineering.
We will soon enable other model options, including one we fine-tuned internally. Stay in touch: ogre.run
Great use of the GPT model, I love stuff like this that helps to boost the speed on starting new projects.
This is the kind of things that come to my mind when people says “work smarter”.
I do have a question tho. Why is GPT 4 a requirement? Does GPT 3.5 produces inconsistent results?
Great work!
Hi! GPT 3.5 does work but its performance is not as consistent as GPT 4’s. We noticed that very often its response drifts away from what the prompt requested, generating invalid requirements files. GPT 4 = less prompt engineering.
We will soon enable other model options, including one we fine-tuned internally. Stay in touch: ogre.run
I've notice some drifts from prompts as well, that's why I was asking. Thanks!