Vibecoding: a beginner's guide

semaphore said:
Never tried Grok, personally, Elon is a choad so don't care much for his shitty products.
Click to expand...
I see. Apart from gemini, the only other llm have not played around with is venice, but why pay for venice when you can do it yourself for free.
 
Herr der Verboten said:
I see. Apart from gemini, the only other llm have not played around with is venice, but why pay for venice when you can do it yourself for free.
Click to expand...
Yeah, they're all kinda the same at the end of the day. Mostly all trained on the same corpus, just with different weights and ****. Soon, they're gonna start running out of human knowledge, or already have. Don't see much more in advancements of intelligence, hopefully in size and compute resources. I run Qwen locally, and to be honest its pretty good. I'm building a tool that i want to use local inference, and so far it works pretty well/kinda.
 
LOL at all the hardcore devs being salty now, Ai is gonna get better and it will take everybody jobs, its only a matter of time,
yes right now its clunky and makes mistakes, but its learning at an exponential rate, and soon vibe coding is what everybody will be doing, devs and normies
 
Jet-Fighter7700 said:
LOL at all the hardcore devs being salty now, Ai is gonna get better and it will take everybody jobs, its only a matter of time,
yes right now its clunky and makes mistakes, but its learning at an exponential rate, and soon vibe coding is what everybody will be doing, devs and normies
Click to expand...
Tool is just a tool and no tool should be a christ or mandela. Same for the tool using a tool... lol
 
Jet-Fighter7700 said:
LOL at all the hardcore devs being salty now, Ai is gonna get better and it will take everybody jobs, its only a matter of time,
yes right now its clunky and makes mistakes, but its learning at an exponential rate, and soon vibe coding is what everybody will be doing, devs and normies
Click to expand...
Really? We are? I am not salty at all.I use the tools to streamline my work and get the repetitive shitty tasks out of the way. Any company solely relying on AI-generated code is (and already are) learning the hard way.

So, unless you're actually a developer/engineer, or very well-versed on AI or LLMs in general. Do the world a favour and shut the fk up, because you sound like an illiterate moron.
 
semaphore said:
Oh and https://openai.com/index/introducing-gpt-5-3-codex/ was relased today as well.
Click to expand...
I’d love to get a professional opinion on whether it’s any good. This snippet from the website either means it is hot stuff or they’re lying through their teeth and it’s never been worse:
GPT‑5.3‑Codex is our first model that was instrumental in creating itself. The Codex team used early versions to debug its own training, manage its own deployment, and diagnose test results and evaluations—our team was blown away by how much Codex was able to accelerate its own development.
Click to expand...
With GPT‑5.3-Codex, Codex goes from an agent that can write and review code to an agent that can do nearly anything developers and professionals can do on a computer.
Click to expand...
 
Foxhound5366 said:
I’d love to get a professional opinion on whether it’s any good. This snippet from the website either means it is hot stuff or they’re lying through their teeth and it’s never been worse:
Click to expand...
Well like I said, I've had much better results with 5.2, if 5.3 is an improvement on that, it can only be better. And its definitely possible the models are enhancing themsevles.
 

Why Claude Code and OpenClaw Just Triggered a Tech Stock Panic - Dad Saves America​

The artificial intelligence hype looks like it’s about to pay off. With Anthropic’s Claude Code taking “vibe coding” to the next level and AI agents like OpenClaw replacing chatbots, the tech world is undergoing a step-function change—and the stock market is starting to notice.

When one person can build powerful software by describing an idea in plain English, and can simultaneously send agents to autonomously handle tedious work behind the scenes, the old barriers that protected Silicon Valley’s dominance begin to disappear.

Now that AI is learning to recursively improve itself, the relationship between humans, computers, and work is going to fundamentally change. How can we prepare for what comes next?
Click to expand...

 
*sighs* Just installed Antigravity and played around with it for a while, but it’s still rocky.

1) Said it generated three design options for my website as images and gave me “open” buttons to view the files, but those file links were broken. It then said its image generation tool didn’t work, so it rebuilt the mockups as HTML pages (that worked but only after I changed the hyperlinks to the mockups because it didn’t ask me what domain I was simulating for my local environment).

2) Had all my chat with the AI shoved by default into a narrow sidebar down the right side of the screen, where the text from the AI would get overlapped by permanent notifications about file access required (although I’d given it all access to a folder so no idea why it felt it necessary to ask for permission again to open individual files IN that folder).

3) I asked Gemini if it could give me some indication of how many tokens I was using through my queries and if there were any limits I’d run into, and it said it couldn’t tell me that because it didn’t have access to that data.

4) I asked the AI to edit the operation of a simple text entry box on my website, and it claimed to have done it but when I refreshed the site in the browser nothing had changed. When I told the AI this, it then decided to apply the same edits to two other similar JavaScript files on my website … but that also didn’t change anything on the front end.

I gave up at that point. If Google can’t get a better experience together, then I think we just gotta wait a bit longer for better interfaces.
 
Foxhound5366 said:
*sighs* Just installed Antigravity and played around with it for a while, but it’s still rocky.

1) Said it generated three design options for my website as images and gave me “open” buttons to view the files, but those file links were broken. It then said its image generation tool didn’t work, so it rebuilt the mockups as HTML pages (that worked but only after I changed the hyperlinks to the mockups because it didn’t ask me what domain I was simulating for my local environment).

2) Had all my chat with the AI shoved by default into a narrow sidebar down the right side of the screen, where the text from the AI would get overlapped by permanent notifications about file access required (although I’d given it all access to a folder so no idea why it felt it necessary to ask for permission again to open individual files IN that folder).

3) I asked Gemini if it could give me some indication of how many tokens I was using through my queries and if there were any limits I’d run into, and it said it couldn’t tell me that because it didn’t have access to that data.

4) I asked the AI to edit the operation of a simple text entry box on my website, and it claimed to have done it but when I refreshed the site in the browser nothing had changed. When I told the AI this, it then decided to apply the same edits to two other similar JavaScript files on my website … but that also didn’t change anything on the front end.

I gave up at that point. If Google can’t get a better experience together, then I think we just gotta wait a bit longer for better interfaces.
Click to expand...
Use Nano Banana for images
 
semaphore said:
Use Nano Banana for images
Click to expand...
You’re missing the point. Gemini in Antigravity accepted the task of generating images. It told me it HAD generated images. It presented me with buttons to download these images. And when I pointed out that these links were broken, it didn’t fix the links. Instead it switched to a different means of production.

There are so many failure points in that single interaction (e.g. why didn’t it catch that its image links were broken?), that I have no faith in it to do anything else.
 
Foxhound5366 said:
You’re missing the point. Gemini in Antigravity accepted the task of generating images. It told me it HAD generated images. It presented me with buttons to download these images. And when I pointed out that these links were broken, it didn’t fix the links. Instead it switched to a different means of production.

There are so many failure points in that single interaction (e.g. why didn’t it catch that its image links were broken?), that I have no faith in it to do anything else.
Click to expand...
I got your point, you're expecting far too much for an IDE. So i pointed you to a tool that is designed for the purpose.
 
You must log in or register to reply here.
Back
Top
Sign up to the MyBroadband newsletter
X