I don't really take notes, but I want to start documenting my thoughts and ideas. Might as well document them publicly. What's the worst that could happen?
This is my first note. I wanted to make it because I saw that I could download an entire AI model to my computer and run it offline/locally. I know I'm pretty late to this, but better late than never. This seems like a really cool idea because in the case that the internet or power grid goes offline, I can still use my model to answer any questions I have (assuming that I have power). I know for a fact that I wouldn't surive out in the wild without internet, but having a local model could definitly guide me in the right direction. Hopefully, my odds of survival would go up.
Though, in the case that the world doesn't end, I do think it's still pretty cool have your own model running locally, kind of like having a true personal assistant that doesn't intrude on my privacy. I could use it offline, have it more private and secure, and tune it just for me. There are also uncensored models, which would give me more freedom to use it for whatever I want. Maybe I could pack a model (or multiple) into a small device and use it as a wearable assistant. Imagine all the possibilties that could come from that.
I just released my first public project. How exciting! It's an independant reporting website that uses AI to automate almost everything. Pretty cool how it works in my opinion. It's crazy how all of this is now possible. A few years ago, this was all just science fiction (for me at least). Now I'm a Claude power user using the 20x pro plan. Lmao.
Right now, AI has enabled me to work on many projects that I've been envisioning for a while. AI is amazing in that sense. Whatever my idea is, I tell it to claude, and in a few hours, I have an MVP up and running. Of course, it's not perfect, and it takes days or weeks to get that project to a state that I'm ok with, but a few years ago, it would've taken me months to do what I can do now. But everyone already knows this.
I'm a person who has a lot of ideas. I get bored getting bogged down in the techincal work, and I'd rather focus on the bigger picture. For some people, they love getting extremely techincal and nerding out on how they optimized the last function they worked on. That's cool. But for me, that never really felt rewarding. I feel rewarded when an idea I have comes to life. I like it when I build an idea from the ground up that I genuinely care about, and it works. I don't really care too much about how it was implemented because then I lose sight of the bigger picture. Obviously, the implementation, software architecture, and optimizations are crucial to having a great app, but tbh, the idea matters more.
But I do understand that solely depending on AI can build fragile, slow, and insecure systems. If you vibe code harde enough, and your context is nearly maxed out, any button that is created could potentially drop your entire database. That's why understanding how the AI works, and optimizing for AI development is also important. Techincal knowledge also does come in very handy. Understanding the tools and systems you're using/building is vital. As long as you know what the right tool/framework/architecture is good for the job, and you know how to debug and quality test your product (which takes the longest time), I think products will come out fine. And with AIs getting even better, and as you learn to give the AI clearer instructions, this testing time will also reduce.
Anyways, I love AI because a lot of my ideas can come to life. It's still a lot of time and effort (and money), but it's a lot easier now, which I appreciate
PS: I pasted my thoughts into chatgpt, and it gave me this insight: "Instead of caring about implementation details, care about failure modes." I shoud always have a security mindset first when developing AI. Asking questions like: What’s the worst thing this system could accidentally do? How easy is it to detect when it breaks? How reversible are my mistakes? Because of my interest in security, I think I've sort of been doing this unintentionally, but now I realized how important it is to vibe code with a 'prevent failure first' mindeset. Thanks chatgpt!