The big #AI companies told everyone that they could build a whole app using an #LLM.
The real software developers like myself quickly found out that it was absolutely not true, unless it’s a simple little app/website. AI is definitely useful, but it still requires you to know programming if you want to build something awesome.
Because of this, much of the hype has died off. There’s still hype, but people are getting much more realistic about what current LLMs are capable of.
I think a lot people has gotten an app idea at some point - not only developers, but also completely normal people.
The normal people were told that they could use LLMs to build their app ideas. Then they tried, and quickly figured out that it’s actually way harder and the AI still requires a lot of guidance. You can’t just write a prompt and expect the AI to give you a perfectly usable and functional app. If you don’t know how to program, you can’t read and understand the code that the AI writes, which makes it impossible to guide it in a precise way.
I suppose this is the reason for why tools like Lovable are getting less Google searches. A lot of these LLM tools were made for people who couldn’t code and now those people seem to be using AI a lot less for trying to build their app ideas. Basically, they gave up, which means many AI tools are experiencing less growth and perhaps even a decline.
Frederik Handberg
npub1nj0c...2gqz
23 🇩🇰 I'm studying for a degree in Software Engineering while building fun projects and experimenting with drone technology.
I'm also learning to sew garments as a hobby 🪡 Sometimes I work as a freelance photographer for different news outlets 📷
I mostly share progress updates as I'm working on my software and drone projects, and as I learn to sew.
Basically, I just write about my hobbies...
Sora 2 is crazy good. Like really good!
But at the same time, it's so damn sad... It's never been easier for scammers to manipulate people. #AI
I don’t think the #AI bubble (if it is actually a bubble) can be compared to the .com crash. Everything with a .com name had crazy valuations without creating any real value.
LLMs aren’t like that. The AI companies do have super high valuations, but they DO create value for customers. Think about all the startups that can now work much more efficiently due to LLMs.
I can see it myself. The progress I am making on my notes app is amazing, and not something I’d have been able to if I didn’t have #Claude helping me do certain tasks, so I can concentrate on other parts of the coding.
View quoted note →
The question is, if everyone came to Nostr and used a client that let them tell an LLM what they want in their timeline, would they exclude brain-rot, or are people naturally drawn to it?
The cool thing about Nostr (and other open protocols) is that the people are in control, not a company.
I think it’s likely, we’ll see a client with the functionality I have described before, regarding an LLM you can tell what you want in your timeline, which then builds a custom recommendation algorithm tailored just for you. #AI #LLM #algorithm
View quoted note →
I have a desktop PC with an RTX3090, Threadripper 3990x, and 64GB of ram, but the thing is, I never use it as I’m not a gamer anymore… I don’t use it for work either, because I prefer my MacBook Pro.
I think it’s time to wipe the NVMe and install Linux on it. Then, I’ll install something like `text-generation-webui` and run an AI model.
I will then fetch all notes from users I follow and feed each note to the AI model. The model should then return a relevance score (1-10) based on my description of the content I wish to see in my timeline.
In that way, I will (hopefully) be able to build a recommendation algorithm to only show me notes in my timeline, that I would actually want to see.
It’s gonna be a fun experiment! #AI #LLM
View quoted note →
I would love to learn what kind of compute it takes to run an algorithm with an LLM that can curate me a timeline of exactly what I want to see here on Nostr.
Basically, I tell the LLM what I like, and then the AI curates a timeline based on my description.
Some people hate algorithms and just wants their timeline to be chronologically from newest to oldest posts, but that’s not what I want. The vast majority of what I see on social media (including Nostr) is just noise that brings me no value and I’d rather not see that at all.
I prefer having an algorithm, but only if I’m in control! #algorithms #AI #LLM
We’re almost at a point now where I can barely distinguish the ChatGPT AI voice from a human voice. It’s both impressive and scary at the same time! #AI
Using LLMs for coding can be risky when it comes to learning.
I often see people immediately asking the AI to solve a problem instead of attempting to solve the error themselves.
Then people will feel like they’ve learned something, but the next time they encounter the same error, they have no idea how to solve the problem, and once again, they immediately pull up ChatGPT or whatever other LLM. And as a result, they never really learn how to debug.
Some beginners rely way too much on AI for debugging their code. #dev #ai
Okay, Meta AI is not perfect.
Now it keeps giving this error:
"Sorry, I can’t help you with this request right now. Is there anything else I can help you with?"
Like, literally, with every prompt.
Funny thing is, it shows the answer, but then quickly removes the answer and shows the error message. #Meta #AI
View quoted note →