Vibe Coding & AI/SI Agentic Programming General - How to become the a 10x engineer you always knew you were despite being absolutely inept in every imaginable manner

  • 🔎 If you see an alert for analytics.usips.org, this is a self-hosted system I am running as a test. It is GDPR compliant and collects no invasive information.
  • Want to keep track of this thread?
    Accounts can bookmark posts, watch threads for updates, and jump back to where you stopped reading.
    Create account
Similar to the the post by @Smooth Jazza I also started a little side project. What originally started as mod downloader/organizer for Rimworld now has become a sort of universal mod downloader/organizer. Consists of a core in which you can input links from say the Steam Workshop, Nexus Mods, Github, Gitlab, etc.. they get staged downloaded and installed. Then instead of going for a general approach for all games since that is impossible what I decided is doing modules for each game to handle their own peculiarities. First is Rimworld but I could see this expanding towards Bethesda games or others. Plus I do have some aspirations to eventually make it a sort of modded save archiver also, in which it can give you a zip with the mods as files, a save or multiple saves and a .txt indicating game version and the like.

Posting here because im using the thread to begin learning how to vibecode more efficiently and better, and also because using vibecoding for these specialized, autistic niche tasks when one doesnt have time to code/learn how to code is a godsend.
 
Lammy said:
I afterwards tried OpenCode with Nemotron 3 Ultra Free and it worked well to get my GUI to work exactly how I wanted it in about 20 minutes, so I've completed my personal vibecode slopproject, and I'm content with it as it is for now

I did also install Unsloth with Qwen 3.8, so that's what I'm gonna try out next time I want to vibecode something, since it's nice to have the reliability of it being run locally and not being rate-limited
Click to expand...
I would say check out Hermes for a project like that! Hermes basically gives the AI a body so it can do more then just talk.

https://hermes-agent.nousresearch.com/docs/guides/local-ollama-setup

Their free model gives you 1 million tokens.
 
Male Idiot said:
Star Wars Empire at War
Click to expand...
This already got a patch in November 2023 to convert it to 64-bit.
Dawn of War 1 got made 64-bit with the remastered version.
You might be able to do it for Dawn of War 2 and the original Dawn of War 1 release, but it won't be a one shot prompt; you'd need to a lot of co-engineering and hand-holding with a paid frontier model.
 
Eaw got a patch? It is still bugging out with mods and has memory leaks.

I was thinking if a 20 dollar Claude 5.5 could have a try at it.
 
HootersMcBoobies said:
It is impressive so much of it works, but when it gets it wrong, it's really wrong.
Click to expand...
I've been forced to use gpt to code things for work, and it's wild to me the things it tries to do. Something stopped working in a pretty standard data analysis pipeline. I handed it over to the LLM to check and it started shoving checkpoints in at random, and writing a bunch of shit that would tell me about missing files that didn't matter, and other random things. Like, my nigger, calm down and do something like analysing the input files for a second.
 
@Null It might be due to me not really doing anything of interest with it but do you see a noticable change in quality when you use Fable instead of Opus?
 
Rungle said:
It might be due to me not really doing anything of interest with it but do you see a noticable change in quality when you use Fable instead of Opus?
Click to expand...
For me, the difference was may more apparent with Opus 5 vs Fable 5.1, but Opus 5.5 brought them a lot closer. Opus 5 tried to suggest some stupid implementations of things that I needed to steer it away from often, so at that time I was using plan mode in Fable and the passing that plan to an Opus subagent so Opus didn't have to make any decisions- since Opus dropped I just kind of stopped using Fable so I could make the most of my budget.
 
Rungle said:
@Null It might be due to me not really doing anything of interest with it but do you see a noticable change in quality when you use Fable instead of Opus?
Click to expand...
Opus is faster, cheaper, and capable of doing basically everything.
What large enterprise companies with lots of money to spend A/B testing report is that Fable is capable of handling very, very long projects without losing focus. It handles compaction better and can run for days.

Supposedly, once Opus's context gets used, it starts spending more time reasoning, which closes the spend gap.
 
Opus 5.5 for 20 bucks is really great value, I'm knocking out ideas left and right and I'm struggling to hit the weekly usage limits. Started having it make a game mashup mod just to burn the extra tokens when I'm not making something else.
 
All I need to do is zip together a few dozen slop Python Tkinter Serial hardware GUI's (accumulated over two years of free LLMs so they are probably utterly trash) I have never read even once into one complete GUI with a shared API so that I can run my shit from one window and/or GCode style script instead of a mouse and keyboard macro across 6 monitors.

Which LLM will let me just paste these .py into their website UI and achieve this for the worth of a random VISA prepaid card from a raffle I won that is at least $20 and not require any Proof of Identity?
 
Silver Coin Anon said:
Which LLM will let me just paste these .py into their website UI and achieve this for the worth of a random VISA prepaid card from a raffle I won that is at least $20 and not require any Proof of Identity?
Click to expand...
Buy OpenRouter credit, install OpenCode, connect OpenCode to your OpenRouter account, select any model you damn want (honestly even GPT-6 Astra is reasonably priced on there), drop all your Python files into a directory, put what you want in `spec/01-what-i-want.txt`, start opencode, then type "Read and implement spec/01-what-i-want.txt" and let it work. Check it all in to git when it's done (or just ask it to handle version control for you if you don't know how that works).

I hate that Openrouter got bought by Stripe, but it's still got decent prices and they haven't screw it up yet. It's basically a broker between you and model providers. They support most of the Anthropic, OpenAI, Z.AI, xAI and other major models open and closed.

$100 of openrouter credits easily lasted me three months, and I feel like I use quite a lot of it.

Checking in your prompts means you can re-run a spec on a branch with a different model if you want to see if the Chinese models like GLM-5 or Kimi 3 write better code or do a better job if you're into that sort of testing.
 
So I was watching a Naomi Brockwell video about some project that was vibe coding replacements for Adobe software in order to avoid their telemetry/spyware bullshit and I thought, okay, nice.

Then about an hour later I decided, Hey, want can't a former coder not vibe code a replacement for 2010's Dreamweaver CS5.5, the only Windows software I miss, but on Linux and only containing the features he wants?

So I went to ChatGPT and started the project. I expect it will take me about two or three months, but I already have a solid road map that will gradually get me to where I want to end up.

If I go ahead with this, I'm busy these days, I'll post occasional updates about my progress.
 
Hypercop said:
Anyone else think vibecoded games might derail the games industry soon? I have a feeling we're about to pass from, "Vibecoded games are slop!" to "Vibecoded games are actually decent" and a new generation of vibecoded shit is going to completely take over steam
Click to expand...
companies are already using vibecoding to port games. it's no surprise all these 40 yr old games are suddenly showing up on steam.

unfortunately vibecoding enables retards to post ports that miss basic features like, say, gamepad support. like any idiot can port a game, but making the port quality is a horse of a different color.

vibecoding, in the right hands, CAN produce good quality games, but we're also in for a tremendous amount of bullshit because generally people who are too lazy to code are also ass at doing software QA. coding being hard and boring put up a gate on quality that has vanished.
 
dreamworks face said:
vibecoding, in the right hands, CAN produce good quality games, but we're also in for a tremendous amount of bullshit because generally people who are too lazy to code are also ass at doing software QA. coding being hard and boring put up a gate on quality that has vanished
Click to expand...
it used to be that you had to have executive functioning, the ability to micro/macro, and have the patience to see the project through to its conclusion. ai kind of demolishes the patience and micro/macro aspect of it so now it has become a pure hard skill check for your ability to architect something
 
big pauper said:
it used to be that you had to have executive functioning, the ability to micro/macro, and have the patience to see the project through to its conclusion. ai kind of demolishes the patience and micro/macro aspect of it so now it has become a pure hard skill check for your ability to architect something
Click to expand...
It's similar to art where the process of learning to program without AI often teaches these related necessary skills and builds your sense of taste, without those even personalized garbage seems impressive. If you have a soul you can see the flaws in your own AI generated work and feel compelled to refine them out, making criticism and revision a higher priority than being able to create from nothing.
You could also draw parallels to manufacturing, quite a few people could get together enough resources to buy some sort of equipment like a CNC machine or 3d printer and start making similar products, but established companies live and die on less tangible things like QA and consistency.
 
I mean I'm just saying said:
You could also draw parallels to manufacturing, quite a few people could get together enough resources to buy some sort of equipment like a CNC machine or 3d printer and start making similar products, but established companies live and die on less tangible things like QA and consistency.
Click to expand...
Right now, vibecoded games still lack that certain OOMPH or je ne sais quois that most good games have in their art, their physics, their sound design, their progression etc. Something about it just doesn't feel satisfying like human art yet, but that could change.
 
Hypercop said:
Right now, vibecoded games still lack that certain OOMPH or je ne sais quois that most good games have in their art, their physics, their sound design, their progression etc. Something about it just doesn't feel satisfying like human art yet, but that could change.
Click to expand...
that's because actual good gameplay systems come from iteration and right now the whole gimmick seems to be showing off how much work they can do in short periods of time. but when you can iterate with a machine that works that fast and can afford to throw prototype systems away instead of comitting to them because the cost was like... 15 minutes of your time, then shit's gonna get absolutely crazy
 
You must log in or register to reply here.
Back
Top Bottom