A general for vibe coding, agentic engineering, coding agents, and shipping code with LLMs.You use Git — right, anon?## What “vibe coding” is, and how to do ithttps://simonwillison.net/2025/Mar/19/vibe-coding/https://simonwillison.net/2025/Mar/11/using-llms-for-code/## Related generals>>>/g/lmg/----## Frontier models using fully-general tooling — start here if you have $20 or sohttps://claude.com/product/claude-code — generally better currentlyhttps://developers.openai.com/codex/clihttps://antigravity.google/pricing## Near-the-old-frontier models for codehttps://x.ai/cli — no 5h limit for only $30/month## Best free tierhttps://opencode.ai/zen----## Promptinghttps://platform.claude.com/docs/en/build-with-claude/prompt-engineering/overviewhttps://developers.openai.com/api/docs/guides/latest-model## Skillshttps://github.com/mattpocock/skills — /grill-with-docs is a favoritehttps://github.com/Vuk97/forward-implementation-first — do less redundant bookkeeping## Other editors / terminal agents / coding agentshttps://osaurus.ai/https://pi.dev/## Is our AIs unlearning?https://aistupidlevel.info/## Will there be a codex reset?https://codex-resets.com/## What we’ve donehttps://vcg.gitgud.site## Previous thread>>110107867
good morning vibe GODS
FABLE 5.5 NOW
Is Free Deepseek on Opencode better or worse than chatgpt free plan chat? I know someone who exclusively uses chatgpt chat (yes, really) and I'm thinking of recommending them opencode, but I don't want them to have worse results.
vibe-threading...
>>110111484don't take my word for it but the best deepsneed mogs luna to the extent that i can't see a free quant being worse. it's also nicer to work with.
>>110111484What exactly are they using the chat for?
>>110111497Translation and translation quality checking (e.g. feeding a script and asking it to look for errors in X00 lines)
>>110111501the transition to multi-turn agentic bughunting alone will destroy a single-pass luna webchat check any day of the week
anyone have a vibecoded software for geological 3d velocity modeling? like leapfrog or petex?cc and sol are stubborn
gm gods
Why can Codex run Msys2 quietly but Claude can't? What is this shit, it either wants me to run scripts manually or it wants to take over my screen so it can take 30 seconds to navigate through folders by sight and click and double-click on a script.
>>110111511darksnail can do this, should i whip something up?
>>110111556nah somethings wrong there, I used all of those with claude both on windows and linux
how 2 implement vibe coding in a professional environmenti don't want to commit 100k diffs, there has to be a better way
>>110111572set a cron job to stagger commits?
>>110111558??? what's darksnail?
>>110111572just don’t use git or version control king
>>110111576can be a spectral analysis tool that runs locallyhttps://github.com/sneed-and-feed/projectdarksnail/
>>110111556I haven’t used the Claude desktop app but something tells me you are running on a cloud environment and not locally. Claude supports both kinds
>>110111468>You're doing this solo? Impressive.Thank you. Yeah it's a mix of:-My own blender activities (like two years worth)-Gemini 3 coding back a year ago-Suno for music-Elevenlabs for voices for the pieces and the story scenes-Opus 5.5 for modernizing the code-Astra for the UIBasically I made the game years ago, and had used gemini image generation for some of the art prior to even nano banana.When my blender work was done I tried to up it to a more modern look.
>>110111517I tried to Grok 4.6 code a time-awareness and memory system for a local LLM front-end but it kept making mistakes all the time, fixing one at a time while messing up others, and just overall didn't seem to be intelligent enough to handle the big picture. And no, it's not because of bad specs, it just failed at fulfilling the specs and was full of bugs that were hard to pinpoint.I wonder if I should give my limited NEET bucks to Claude? Is Claude actually intelligent or benchmaxxed with skills for impressive results? And should I use Opus or Fable? Is Fable even more intelligent or does it just bruteforce with more tokens? GPT Astra based on my research is kinda intelligent but too sloppy for something this complicated. What about Argon? What are Argon's strengths? Robust logic? But can it understand the big picture?
I'm going to ban all /vcg/ posters with no serious project
>>110111576Schizo slop that magically does everything from solving ARC-AGI to drones and now geological 3d velocity modeling
>>110111632I put minimax to work 3 hours ago on a small task and it's still going...
>>110111632who's moot?
Gatekeeping, cringe or based ?I saw a video with good (but obviously IA assisted) motion design. I asked the guy if he could share in a few words his workflow, he replied he'd rather not say. Of course I realize he's completely in his right.
>>110111598I'm running locally, though it might not help that I'm running the Microsoft app store version. I installed Desktop-Commander now.
>>110111556Claude chat won't run shit you have to open the code tab, yes it's dumb.
lol I REed this proprietary shit piece of software at work in one prompt and my boss is calling me a "tech wizard"
>>110111460Is there a reliable way to jailbreak chatgpt work? It seems much more restrictive than regular chatgpt or api tokens. I'm trying to vibe code lewd browser games.
How are you all reversing applications and systems using Claude without it refusing or dropping back down to Opus 4.8?
>>110111709use a chinkmodel
>>110111710By using Codex
>>110111632im still building my own game... animation is hard...
>>110111705Most people, even those working in tech, are still so unaware of what the current models are capable of. The shit I hear some of these people say is just straight out of 2025.
>>110111665Based when I do it cringe otherwise.
>>110111717Which ones are good and easy to jailbreak? I've never used chink models for agentic tasks so im out of the loop. I also can't code for shit and am familiar with the ease of use of Claude and chatgpt work.
>>110111738>>110111705Not as impressive but I’m a nocoder who uses this old open source program that is outdated, full of bugs and constantly crashes or brings up error messages.All I said was “fix this program”. And it works, flawlessly.
>>110111764glm 5.3 is the dog standard for eating decomps alive but it's not equivalent to american
>>110111710i dunno. i just asked claude to use the rea mcp to figure out how internet download manager is so good at capturing videos, and migrate that method to xdm.and it did. of course, rea mcp fucking omega rapes your usage on $20, so i have to wait for the weekly reset tomorrow
>>110111772surely deepseek is miles ahead of it as of 1017?
>>110111787not sure about 1017's restrictions, objectively it's better thoughever
>>110111572wait for scion to be publicly available
>>110111738The other day a software engineer told me "I spend more time fixing the AIs mistakes than it probably would take me to write it myself". Like my dear brother, why are YOU fixing its mistakes? What exactly are you using to drive your LLMs? They fix their own mistakes just fine these days. But I just nodded and did not engage further. These people are utterly clueless.
any way to do really good video game animations with the current models?
Eurobros, is now a good time to switch jobs? My current position doesn't pay much, but there's no stress at least. My boss recently said we're going to stop programming soon and only make decisions for AI, and test. He insists I don't have to fear for losing my job. I have a permanent contract so they can't get rid of me easily either. I got an interview coming up for a senior position that offers a much higher salary and a car. I'm scared AI might progress so quick this company might get rid of me first, since I'm the new guy.
it's so unlucky that soviet died decades before AI boom
>>110111821yeah leaving a permanent contract is a very very bad idea, that shit's rare nowadays
>>110111671No, but even though you're running the app locally, is it a cloud or a local session? or even a web chat?You're most likely not using the right thing at all even if you're using the installed app. Where you install it from has nothing to do with it.
>>110111841>that shit's rare nowadaysIs it really? Companies in the netherlands are required to give you one after 3 years (or 3 consecutive temp contracts). Most people I know have one, and most at my current company get one after 2 years.
Codex limits usage has too be broken, no? I can barely get 40 minutes of usage out of it now on xhigh. Last week it would work for 2-3 hours easily before draining my quota. I guess I'll go down to high to get 20 minutes extra out of it, lol.
>>110111856It's just a local project with a project folder added.
so what bullshit "improvement" is going to be added today to avoid a reset? a fart button on the desktop app?
>>110111656your papa
mega, checkmediafire, check
>>110111511
>>110111910oh yeah, I think he died.. I hope so
>>110111926what does this do? are you guys geologists?
grok 4.7 in cursor harness for $30 plan currently has usage limits similar to a $200 plan. idk what’s going on.
>>110111898They are gonna stealth nerf usage again and say they fixed some irrelevant issue.
>Microsoft updated 18 evaluation configs from GPT-5.6 Terra to GPT-6.1 Terrano more fake sol
>>110111511done, v0.1.0 is building on runnerhttps://github.com/sneed-and-feed/g3vm>>110111978sa'ar is a geologist in bangladesh
>>110111988what a cluster fuck this is. so how much is this gonna cost when 6 sol exists?
>>1101119886.1 Terra is a thing? Thought they retired the Terra line
why does everything cost so much more now, tibo is real quiet. fuck that nigga
>official now: anthropic is preparing the release of their new model(s). Via The Informationsam gayman... tibosaar... are you ready?
In terms of error detection it seems like OpenCode Deepseek 4.1 is fairly competitive with free plan GPT Luna. It caught 3 more minor bugs but missed an equal amount of major bugs. More testing is valuable, but so far not open and shut
6.1 terra is probably gonna end up competing with deepseek 1017 in capabilities (but definitely not cost)
>>110112035can it do math? I bet it can't :<
If everyone has these super powerful models available to them, how are there still people struggling to put food on the table?
because we are too busy vibe GODDING and not eating like a fat fuck
>>110111921bro why are you still using this boomer software. jdownloader exists.
do a fresh medium-size pretrain with all the bel thinking trace then crush claudeI know you can do it sammy
>>110112039deepseek isn't even performing at it's best, considering that it's not using claude code or dsh, which it's better trained on than opencode. luna has been left behind completely. can't even put them in the same group anymore. it'd be more fair to pit deepseek against 6.1 sol low
