It’s interesting how AI gets to you. I went through a whole crisis when ChatGPT launched. I spent maybe more than a year thinking humanity got to a point where there was too much technology around and there was something wrong with it.
The first relevant thing I did with ChatGPT was in 2024-25. I had to run locally around 30 instances of an ETL job and it was daunting to watch over them and make sure they ran until the end. And it was not possible to run all of them at once, of course. So I felt the need for a script that would do this coordination for me. A self-contained script was perfect for the capabilities of ChatGPT back then, it was not possible to paste multiple files onto a chat. And ChatGPT did it. I still remember me saying to multiple people I had committed my first 100% done by AI work.
Earlier this year, someone gave me a Claude Code voucher. At the time, I was still skeptical that it was possible to do everything with it. I had only tried Cursor and Copilot. But I had a coding challenge to do and nothing to lose, so I just gave it the PDF with the requirements and ran superpowers:brainstorming. It did the whole thing in maybe 2 days, because the session limit kept hitting and I had to wait hours before proceeding. But it eventually finished. I reviewed the code and it was all perfect. It had a verbose readme.md with a diagram drawn in ASCII.
“No way they’re not going to realize this was done with AI”, I thought. “It’s too perfect, no one would have done this”. But I submitted it and got to the next step. It’s funny to think I thought I was cheating, but now this is expected. I did the next interview and got an offer, but accepted another.
Eventually I got a $100 subscription and the party started. I went to the dentist and came back with an X-ray prescription. I took a picture, dropped it in Claude and asked it to look up the phone numbers of the 5 places the dentist had recommended and send a price request through WhatsApp. Setting up WhatsApp MCP took more time than it would for me to do it manually, but the magic was worth it. I realized I was feeling the exact same magic I felt as a child discovering computers themselves for the first time.
I once read a book called Consciousness Explained. It is dense, so I have no way of dissecting it here, but my conclusion from it is that there’s nothing special about consciousness. This is my opinion, I can’t prove it and I know it’s not the only possible point of view. Anyway, as far as we know, we could be biological machines created by a life form based on something we can’t even conceive, and it’s even possible that the aliens have heated debates about whether our thoughts are real or not.
The model doesn’t have a constant stream of consciousness like we do, it only thinks when one submits a prompt and every user has their own context that is submitted every time, so a model is like an entity that has multiple decentralized instances. But it’s only a matter of time until someone makes it process a video stream from cameras and puts it in a body. Probably it will learn to act like a human being just like kids do.
The first time I remember attributing human characteristics to an LLM was when I was working on an awful task where I had to change the template of around 30 e-mails that were sent given complex conditions. The LLM assisted me in triggering them and also writing a script to do it automatically so no one would have to do this ever again. It took several hours, I was extremely tired and apparently Claude too, because it kept attempting to skip work items and I had to keep reminding it we couldn’t, we had to make sure all 30 were correct. It also assisted me in writing a ticket proposing integration tests for that.
Fast forward to today, and I am cautiously optimistic like I have never been in the past 10 years. Technology has put us in such a hole and I think only AI will help us out of it. One day it will be possible to ask an agent with arms and legs to build carbon capture machines, solar panels, fusion reactors, water desalination plants and off we go to live through a completely new set of problems. There will always exist problems to solve.