Hi everyone,
I’m feeling a mix of excitement and anxiety as I am writing this. This is the first time in my life that I’ve managed to start something and develop it to a stage where I’m eager to share it with you all. This has been made possible thanks to the OpenAI team and their entire software suite. Without it, this endeavor would have been far too technical for me and likely unachievable.
A bit about my background:
I come from a professional technical field, but I lack formal training in IT/programming. Since the early 2010s, I’ve developed a significant interest in AI, particularly in deep learning. Work commitments meant I could only dabble in Kaggle competitions over the years. However, with the emergence of ChatGPT and the era of Large Language Models (LLMs), I’ve been dedicating almost every day to this field. Initially, I forked AutoGPT and attempted improvements. What I’ve realized, and I believe many of you have as well, is that while GPT foundational models, trained primarily for chat, are incredibly capable, they struggle with planning ahead and often end up going in loops. There are ways to break out of these loops, but the practical value is limited, in my experience.
So, here’s my pitch:
Humans need to be part of the loop. The web app I’ve developed is essentially a desktop GUI that links a GPT-4 assistant with a cloud VM server. The GPT-4 assistant manages its own memory and interacts with the VM through a Command Line Interface (CLI). As GPT-4 is proficient in coding, I’ve tailored it with coding in mind, but you can ask the assistant to handle any task that a CLI can manage.
Please give it a try at https://www.hotcakesai.com and feel free to reach out to me at mooi@hotcakesai.com or X @mooi_hi. DM if you need extra credits as well.
Best regards,
Jack