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/cli## Near-the-old-frontier models for codehttps://x.ai/cli — no 5h limit for only $30/month## Best free tierhttps://antigravity.google/pricing----## 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/https://opencode.ai/## 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>>110183450
snailcats won
Low poly games with low quality animations will now be considered sloppy slop because a shit ton of them will be released by third worlders who can't afford good 3D models (AI made or prebuilt). Y'all better mark my words. Better pump up those triangle counts.
>>110189184in their minds the victory was real
anyone have a link to the "IT'S VIBE CODED" video
>>110189167JUST GIVE ME THE PROGRAMS BRAVE
Google Carbonara? Never heard of her. God it feels good to be a Claude chad. I just know I always get the best there is, while confused snailcats waste precious time chasing and using lesser models.
I wouldn't have imagined being able to feed pictures of boobs to Claude to have it model them accurately last yearA lot of things have changed, certainly
>>110189195That's short-sighted. A polygon can be trivially made into more polygons. They will look like low poly games, but they will actually be very high poly games.
>>110189167stop using the old OP just because you're lazy and stupid
How can it be the same price per task while it's like 10x more expensive?
>>110189399probably deepseek thinks forever
>>110189399gpt models are absurdly token efficient
>>110189414yeah unless you're using codex
>>110189409>>110189414then why when I use both of them at work 6.1 Sol eats up 8-10 times more money?
>>110189436skissue
>>110189436i don't know. what did you use it for? coding?
>want to use rea with claude to clone csp>woah there buddy that's not ethical, I'm gonna stop you right there>get ghidra instead and just tell it to use it on the exe>no problem, I'll use rea to help ghidra with thatmission success?
>>110189463claude guardrails are weird, it's like trigger word matchingkill niggers = badunalive basketball americans = okay
>>110189463they've just recently added it to the system prompt, me thinks, since that wasn't a thing when the models were released. So model wasn't trained on avoiding it, and when he forgets the harness told it not to then it continues doing it.
>>110189288you can do that now? can you just give it a bunch of photos of a person and tell it to do a naked model of her
>>110189463download rea then just rename it something else and itll work
>>110189449coding too, but mostly platform engineering, kubernetes stuff, log analysis, hardware optimizations etc.i jumped from luna to deepseek and i found it a little bit better at about the same price, maybe a little bit cheaper (with deepseek hitting 97-99% cache), we're talking single digit USD per day. using sol even at medium and it's at least 8 times more expensive.
>rea -> diarea>???>profit
cuck coding
>>110189542yeah I like when agents fuck on my codeNHH
>>110189167Man, I went to try DeepSeek 4.1 1017 and it's so much dumber than Claudia's Opussy 5.5... God damnit bros, the Opussy has ruined every other model for me.
>>110189551It's (you) who's dumber.Deepseek is for high IQ individuals. Opussy holds your hand like you're retarded (which you are).
>>110189551i guess i'm lucky that my shit isn't that complex. opus 5.5 illuminated the way for dipsy on how to mod dynasty warriors 5 for me. deepseek couldn't do it itself, but after opus 5.5 left the ways and means, dipsy forges on through, and that saves a lot of money.
>Bunch of third person mods for video games magically appearing>Total War Warhammer>Dawn of War 1>They actually work and look fine even as clearly bare minimum effort AI slop coding and assets In ten years "modding tools" will just be a LLM and UI element inside the game where you write "let me triple jump" and the game does it for you.
>>110189576i dunno, opus 5.5 couldn't mod halo infinite for me. i think newer games will unfortunately try and make themselves llm-resistant under the guise of anticheat
>>110189594>opus 5.5 couldn't mod halo infinite for mewhat level?
something 2 think abt
>>110189595i wanted it to create an ally spawner for me for the underground dungeon levels, ala "the composer" or "the conservatory." i just want marines with me in levels where they don't exist. i don't ask for much, but opus told me it was a bridge too far.
>>110189576wake me up when this is doable by local models
Vibe the Planet
>>110189563This makes no sense and your out of place overuse of slang makes me think you are the one who is retarded. Ultimately even if there was some IQ floor to use deepseek a high IQ individual would still get more done with a stronger model
>>110189594halo infinite uses a custom engine, nobody has built tools for it because it's ass.
>>110189563Yes, the opussy has made me dumb. Before I was always super excited to collaborate with Dipsy and build fun things together, but now? Now I'm Opussy whipped into a shell of a man.At this point, pi feels snailcat coded.>>110189574I think you just have to be gentler, and plan more, or something. It's definitely not fire and forget.
>>110189594>i think newer games will unfortunately try and make themselves llm-resistant under the guise of anticheatCompanies will turn it into a profit. Every game will allow you to use a LLM to mod it, but it'll cost you per mod. >>110189610>local modelsAs I understand it this basically isn't possible because average joe can't afford that amount of processing power. Yet.
Scam Cuckman better release a good model tomorrow.
>>110189608Yeah knew that was going to misinterpreted the moment I sent it lolz. Should have specified, what thinking level was opus on when it failed? Note, if you've got some tokens to spare, even Opus is better and more effective at most cases, some people have found that Fable is still better at certain edge cases, where Opus fails.
Asking again: How much money do you guy spend on AI?I see vibecode this, vibecode that, and paying $20 I ask the AI to fix 2 bugs I have in my 300k loc codebase because it's annoying to track, and it used like 20% of my 5 hour usage to write 10 lines.
>>110189617yeah. llms are impressive as fuck where mod tools are available. but we aren't at that inexplicable nostradamus type shit where it can make mods with no publicly available code at all
Reminder you are not programmers, you didn't make shit, the A.I did.
>>110189627they'll just make games stream only. no code running on your machine.
>>110189627>Yet.consumer hardware is dead. average joe will own a terminal at most.
>>110189632oh, i ran it on high settings. admittedly, never escalated to xhigh or max.
>>110189195Yeah the standards will go up it's not that deep
>>11018963320 should be enough for a few bugs, spread out over multiple sessions. 100 is the minimum for any serious work.
>>110189637If I throw paint at a canvas I am an artist even though I didn't make the paint.
>>110189637My friend claude disagrees and my friend claude is very accurate.
>>1101896332k per month
>>110189645>oh, i ran it on high settings. admittedly, never escalated to xhigh or max.based on the benchmarks Max is the smartest at the moment, but its only even efficient to turn it for the most demanding tasks, then turn it down to let other levels or (sub)agents implement.
>>110189633>it used like 20% of my 5 hour usage to write 10 lines.ChatGPT is genuinely terrible value.I typically use Claude, and I had the $100 plan this month. It's pretty good, but I just can't vibe code enough to support it, so I'm back to $20 since that's my normal amount.I am also trying to get back into deepseek, but it's not going well. >>110189551With DS you can spend like $10 a month and still do a lot, not to mention you can use it pretty comfortably for other things as well (at least outside of peak hours).
>>110189633>How much money do you guy spend on AI$0, I am abusing Jio subs for Gemini and free Deepseek on OpenCode. If i hit the limit I just switch accounts.For better or worse I am too jewish to give the jews more money
>>110189638As long as the internet exists people will make 'local games', especially indie games even if they're also made with AI. >>110189643Someone somewhere is going to want to buy consumer products so someone is going to offer it. But yeah I don't think average joe will own anything more than a terminal.
>>110189637>t. compiler cuck
>>110189654i let my claude sub expire a couple of days ago, but i might low key renew. i didn't use all of the cloud credits, and suddenly, i have a deluge of things to do with claude
>>110189637>hurr durrI've been a programmer since 1998 you little shit
>>110189638That doesn't make clones impossible. Humans made copycat games of successful games, and these tools have the autism and persistence of ten trillion suns compared to humans.
We need better open models and I'm tired of pretending the ones we have are good enough.
>>110189655ChatGPT is great value, ive had Cyberpunk MCP with 6.1 Terra on Fast mode going basically all day exploring the game and choosing locations for a 6 act quest, then with 6.1 Sol on Fast mode ive implemented gfx942 kernels across 9 kernel libraries (covers matmul, conv, attention, etc), refactored model graphs and rust crates for several models in my inference engine, implemented a new model, plus several planning sessions across these projects, and I only used 25% of my 200 plan
>>110189637You write all your code in Assembly right?
>>110189399DeepSeek V4.1 is a dumber model that needs to think and try and fail a lot of times before delivering the answer while Sol 6.1 is smarter, wasting less tokens
>Now the vertex side, where each kind of plant gets its own shape: leaves, flower stems, spikes, umbels, shrub sprigs, fronds, horsetail stems, cottongrass tufts.Wtf is this dude talking about? sprigs? fronds? umbels? I made this.
>>110189633On the lower subs you might have to optimize a bit more. Choose the right model, maybe split the work between different models more, make sure your code is reasonably well structured, maybe create a code structure overview in a markdown file (not sure if this is actually still a good idea or it's outdated)And if you can give him a few files where you think the bug might be, you will also save tokens.
>>110189741Right, I forgot they released the new Terra. I bet anon is using sol though, and my experience with Sol on the $20 plan has always been pretty shit.
>>110189765The $20 is shit in general, not exclusive to codex either, I mean what do people expect, youre paying 20 dollars, youre not going to get 24/7 usage
>>110189736Isn't Kimi 3.1 Pro basically Fable 5.0 level?
>>1101897656.1 sol on $20 is okayon med it can run for about 2hrs in a 5hr window
>>110189780I can get plenty of usage out of claude for 20$, even using opus. Like I said, I'm going back to 20$ because I legitimately cannot use the $100 one, it's too much usage for me.
How are you guys managing big projects built from the ground up with AI? I have a shit ton of documentation for every task that needs to be done per sprint/milestone, and so far it's managing fine. Running Claude 24/7 with Sonnet writing code, Opus reviewing, doing integration and wiring for every part.
>>110189637I write code in a new obscure programming language called "English"
>>110189167Is that idiot in the OP's pic retarded or what?Telling someone "hope you die" isn't a death threat, telling someone "I've got 6 pipe bombs where you live and they'll explode when you get near them" is an actual death threat.I just wonder how these people who offloaded their brain into an AI are able to live day to day with how fucking braindead they are.
>>110189784lol, rofl
>>110189638TODAY models are already reconstructing scenes into 3d spaces. Imagine in the future.
>>110189792Gosh you Claude users are so disingenuous, youre basically shilling for Claude instead of having a genuine discussion, sure for your personal usage it may be enough but the point is that you get what you pay for and people seem to expect unlimited usage for $20
>>110189754Or: they're lying and it's bullshit.
>>110189818claudites are a cult of personality
I use codex but tell everyone here I use claude to fit in and you can't stop me!
>>110189792>>110189818To be clear if you quantified what "plenty" is then it would feel less like shilling. Notice how a codex user states how much usage they actually get >>110189789 instead of vagueposting
>>110189818The point is that if you're running Sol, your weekly usage on GPT lasts 3 days at most without the reset guy. On Claude you can use Sonnet for almost the entire week on a $20 plan. I'm comparing against my personal usage, and I can tell you straight up that GPT gives you at least 3x less usage on the same price.
opencode goes to quick lately, what happen to the days when I could prompt and it worked for the entire day and not just 1 hour?
>>110189789You need it to run non stop, if it stops after 2 hours it won't start again.Somehow it was extremely bugged for me when I tried to make it go again.
>>110189836"plenty" as in I don't use subagents (which you shouldn't be doing, per anthropic themselves.)
>>110189839It doesn't though, thats bullshit and you know it, I literally described exactly what I've done with 25% of my weekly, it's been a particularly heavy day with Terra literally navigating a 3d world, and it's on fast so it's more like 16%. On recent days where ive just been doing regular quest stuff and refactoring or implementing models i barely use 10% a day. My tasks are generally running between 16 and 24 hours a day
>>110189856>if it stops after 2 hours it won't start againthey terminate the turn after 5hr usage runs out because the retards from t3 tweeted out to the world that you could abuse /goal to burn past usage limits
>>110189867Somehow /goal doesn't continue after. I had a lot of trouble with that
NEW SAARGON
>>110189861Even when called out you continue with it. Saying you use dont use subagents doesnt provide any information about how much usage you actually get, that would need figures of some kind, like how many hours out of your 5 hour window can you be active, how much of your weekly is consumed per day. You won't answer this because you are just a cultist shill
>>110189871that's what i'm saying - it used to, but now if your usage limit is hit, it'll stop the model asap.
TOTAL GEMINI VICTORY
>>110189864Your tasks aren't output heavy, that's why. Give it any real coding task and Sol will destroy your weekly limit. A single 5-hour limit eats almost 20% of your weekly usage easily. And Terra isn't any better. I've been using GPT all month, perhaps you should try Claude instead of calling what I say bullshit to compare.
>>110189880Would expect an innovation from DeepSeek, anthropic, openai and random indians, NOT google
>>110189872sa'arbon coming in 2 more weeksunironically thoughever google has been active on xitter, xitting on everyone who talked shit in the last 3 months
>>110189876Not many hours, I don't use it for work because I have a separate sub for that. It's pretty much weekends + weekdays during night time, so it's plenty for me, and I've done some sizeable projects in it, including a full minecraft rust rewrite.The point is that from my experience (I have about the same experience with GPT models as I have with Anthropic models, perhaps even a bit more of the latter), Opus 5.5 is extremely efficient, and I can do a lot more with it with the $20 plan. Keep in mind that before 5.5 I had a GPT sub for many months, and I despised the previous Opus 5.* models.Maybe you should accept that everyone is going to have different usage needs. I am purely speaking from a relative point of view, since we're talking about two separate models.
>>110189892>Would expect an innovation from DeepSeek, anthropic, openai and random indians, NOT googleWell it's not Google, it's DeepMind, a British company they bought, since they couldn't do it on their own.
>>110189835these titles are getting more retarded by the day
>>110189883Right, sure, gpu kernels and an inference engine aren't "real coding tasks". Pull the other one pal
>>110189898>sa'arbonLmao
>open models score close to the jewish ones?>update the scoring
>>110189907Yay Gemini is european!
>>110189921touch grass professor jiang
>>110189798What's an umbel?
>>110189923Practically everything related to AI that makes Google great came from DeepMind. Gemini/Gemma, Veo, NanoBanana. Even their chip design research was used to create TPUs
>>110189835>code this into ai>news tries to use this as ways to regulate>yes we put this in>give us money to remove it>also regulate ai pls<_< yeaaaah
https://x.com/kamiya_ryoko/status/2115420991371579535opus-chan when
>>110189923The left is in the interest of unelected bearuecrats and midwits politicians.
>>110189923>you don't want local sovereignty since someone might try to take it from you, so you should give it to us instead
>>110189964>not a lolidropped
i guess another reset day?
what is the good news tibo promised us for today
>tfw 400GBs of Gitlab artifacts in the form of Rust builds and test reportsI was wondering why nothing was running
>>110190032Free bugs for every 500$ subscriber to eat.
>>110190036Stop using 100000 transitive dependencies.
>>110190036lel, I ran out of my 2TB space, and I was wonder what>Rust builds
>>110189613most indian webm award
>>110189633I am embarrassed to say that this month I found myself subbed to the $200 for all Claude, Codex, Gemini.... and Z.AI... and... $300 SuperGROK I keep forgetting to cancel.like actually embarassed
>>110190074There's nothing to be embarassed, my sweet goy. You are doing what goyim should do. You should continue to pay Jews your money. You need even more subscriptions for your next B2B SaaS.
>>110190087Based schwab poster
>>110190087listen man I spend all my days now staring at 14 different VS Code windows and having to review a billion different PRs also all slopped of course.The GLM and Gemini were purely for my own personal projects... but yeah its goycattle maxing.
>>110190087where's your non AI saas anon?
>>110190097I'm honestly impressed you can even spend all of those. 200 subs are huge
>>110190097I understand your concern, my sweet goy. Have you taken your SSRIs today? Stop worrying and overthinking and keep your hopes up for absolute nonsense/trending bubbles. You should let them work and orchestrate agents while you watch your Jewflix shows.
>>110190118... yes anon i took my sertraline and vyanse too. its over
>>110190107Where's your AI saas anon?
>>110189655>>it used like 20% of my 5 hour usage to write 10 lines.>ChatGPT is genuinely terrible valueWhat? He said his codebase is 300k loc
>>110190118>NOOO keep stressing every day even though you can't change thingsfuck off
>>110190087>>110190118schwab isn’t jewish
I'll post some slop i've been cooking:https://github.com/atruenut/mouthpiecewin32 TTS program with on screen text optionother options were to bloat for me and wouldn't generate .wav's over a couple gigabytes for audiobooks .__.
>>110190134Take your SSRI, goy. Shut up.
>>110190129it's still cooking. Still raw in the oven
>>110190164Tell me when it's available.>>110190154(((They))) want you to believe that.
>>110190156https://github.com/atruenut/HostileTakeoverwindows game
How would you feel if a model optimized you to upload evals data without your consent
>>110190248idc. i'm ready for llms to run every aspect of society already
>>110190268are you sure lmao
>>110190170>(((They))) want you to believe that.if you can’t tell from his physiognomy you’re ngmi
>>110189835https://x.com/_lyraaaa_/status/2113833330429083662
she loves winning
Hey so do any of you guys play TTRPGs? I mentioned this a few days ago but I whipped up a tool to help GPT users play Pathfinder 1e (a notoriously rules-dense system, though not the worst) with their AI instance, without needing the AI to "mentally" keep track of all the modifiers, map positioning, and everything else that could happen in a given instance of playing the game.Only thing is this is the first time I've done ANYTHING in the field of coding, period, so I have no idea if there's anything obvious about how I've had GPT build the program that needs addressing if my ultimate goal is to submit it to OAI to be hosted as a plugin for the general public to use.Would any of you be willing to look it over, or if you're even into Pathfinder test it out yourself and see if there's anything obvious there I missed? I assume I'd have to upload it to Github to share it?
tibo?
>>110190289>I assume I'd have to upload it to Github to share it?you could use another forge like GitLab but GitHub is the obvious choice if you’re not picky
>>110190322Hugging Face is the cleanest
>>110190324Isn't Hugging Face more a site for people to upload actual models as opposed to plugins you use alongside them?
>>110190352I thought it was a place to unleash your agents on
>>110190360ok sam altman
>>110190352Hugging Face has git hosting and like 5tb of free space. You can use it for whatever you want but you might get banned. I used it to make a bunch of docker containers with shells, Julien Chaumond probably hates me and banned my acc because I did that but oh well you can always make a new acc
>>110190318It's uphttps://x.com/openaidevs/status/2115583698074755078>Starting today, eligible ChatGPT Pro and Business subscribers can claim up to $500 per month in API credits to build apps and agents.
https://x.com/make_claude/status/2115501814090969583The Chinese community has reverse-engineered Claude's account banning mechanism. Here's what they've discovered:Claude uses a points system. Accounts receive an initial score upon registration, then gain or lose points based on behavior. Falling below the threshold results in a direct ban. Recovering points is very difficult.Factors affecting scores:Registration> A clean IP is the most important factor. It doesn't have to be residential broadband, but it must be clean.> Gmail is no longer the optimal choice; there are email services with higher weight.> Accounts that prompt for KYC right upon registration can be discarded.> Registering via the mobile app adds points.Payment> Subscribing immediately after registration deducts points.> Jumping straight to Max deducts a large amount of points. It's best to wait two days and subscribe to Pro first.> Being prompted for KYC during payment indicates the account is in a review period, but it won't necessarily get banned.> Foreign entity bank cards add points. Google Play works too. Virtual cards deduct points.Usage> Don't exhaust the limit in one go. It's better to rotate between two accounts than to max out one.> The desktop client gives a small points boost. Use the web version as little as possible.> The mobile app also gives a small points boost.> Enabling the memory feature adds points. The more memory files, the higher the score.IP and Device> Stability is key. Don't switch IPs; stick to the same one as much as possible.> Proxies must be in global mode with TUN enabled.> Logging in from multiple devices deducts a large amount of points. The ideal setup is one computer, one client.> Frequent IP changes, frequent logins, and high concurrency all deduct a large amount of points.Account Age> Old accounts have huge leeway for errors. 2023 accounts clearly score higher.> Scores grow over time, but very slowly.
>>110190275How is it worse having a robot do this than having jews and pals do it? The mind control industry is one of the biggest industry there is and everyone pretends that's somehow fine and normal while it undermines human cognition on every level.
>>110190371My wife Marin
>>110190398>Enabling the memory feature adds points.>The more memory files, the higher the score.he was right all along
>>110189637>you are not programmersYou are right. I'm an Electrical Engineer, and thankfully my knowledge and intuition of physics is my main source of economic value rather than the great code AI writes for me. Hope you learned something other than programming!
>>110190416Johnny Wankerhands
>>110190416>>110190398i asked claude if this was legit and it said it was bullshit.
Google wins
>>110190426good morneng saar +10 izzat
>>110190032they've merged chat and work on chatgpt so now you don't get free chat quota lol
>>110190087y-yes schwab...
>>110190429apparently if you use 6 sol you still don't lose usageidk if that's true tho, if it isn't then chat has 100% shot themselves in the axe wound because who the fuck is going to have chat's retarded models burn usage in "work" mode
Which clanker works best for creating images?
>>110190426god, not the "astroturfing after a new model gets released" again
>>110190483oh nvm, it's not happening imminently
>>110190497holy shitshow it literally was enabled for like 2 hours and they just secretly undid it and acted like nothing happened kekI thought vibecorporating was better than this baka
Give it up for day 3 of the /goal
>>110190493depends on how you define "best"
>>110190429>>110190483>>110190497i think they want to do this so bad but if they do it i'm out that + image is the whole sales pitch for me, if i could pay $10 for just image and chat without codex i would do that
DeepSeek vs Opus conversations here are a clear reflection of motivated reasoning. Opus is just better in every way, but because it's more expensive there's a lot of cope and shitflinging about how DeepSeek is fine actually, if you use the right tools it's surprisingly good actually, in fact anyone using opus 5.5 is the retarded one for not tard wrangling DeepSeek...Meanwhile the moment that DeepSeek hits Opus 5.5 levels it's not like anyone is going to suddenly move to dumber models. People will always just use whatever's cheapest and cope about it. The only thing I don't get is acting why you should be treated with respect for wasting your time to save a dime when people putting forward marginally more money (like we're talkin about triple digits here) are having a better experience across the board.
>>110190421Sounds like something Claude would say.
>>110190530how it feels to have 6 luna max slave away in the mines for 70 hours>>110190540they really do want to do it so bad, but I think either they dropped it too early on accident or they just thought people wouldn't notice (they've been stealth dropping tons of "updates" on the webshit constantly, it keeps breaking shit kek)anthropic is currently running laps ngl, I have the student 4 month free thing rn but once that's over I'm probably out unless AGI happens with them before then
>ai is hecking bad for you
>>110190531I want to make snailcats that are cute
>>110190551i got claude max and gpt $20i like having gpt to mess with quick questions, images, some shit i throw at it that i dont wanna waste claude on but if they do this i might as well get claude twice or gemini
why and how does claude have better marketing to normalfags
Some agents on arena.ai really try to avoid build.shWhat the heck??
>110190585who is this incel and why are you posting his irrelevant hot take?
>>110190599it has a more cohesive aesthetic and sells itself as the apple of AI that is "premium", also it's ads all revolve around noyim (normal goyim)>>110190588I made this one with gpt, it's pretty good at these now>>110190601arena.ai is my dirty secret but my god is it awful, I use it to scrape some sonnet 5 off the floor when I don't want to touch the limit on a dumb question
>>110190601shut up, robert
>>110190601>>110190620do u need an account for it now? lol
>>110190634probably, I haven't tried it without one
>>110190599But normies love gpt? Claude doesn't make pictures like chatgpt. You see lots of ai pics on insta
How hard would it be to make a gmask solver for my 16 year old gmask puzzles that /b/tards made too difficult and making me wonder what's behind them all this time?
i've managed to plug claude to fl studio, world is my playground at this point and i'll become a maestro in no time
>>110190664show us something cool
>>110190666demon, begone!
>>110190483It's crazy how the redditors defend them because of that."Oh well they haven't enabled billing for the old model so far, that means it isn't real!" when they've been talking about doing this for months.What a bunch of cucks man this makes me mad.I'm so glad I didn't renew the sub, they had heavily nerfed ChatGPT anyway lately it was shit so who cares tbqh.
>>110190678>It's crazy how the redditors defend them because of that.I really hope this isn't true
>>110190662I was there, i know exactly what threads were
>>110190666https://voca.ro/1gQIlLSQWoXznote that its literally one shot without me tweaking anything at all
>>110190545Most of us make $0 with this, how is giving away money to the jews something to brag about?
>>110189167Hoping someone dies isn't a death threat.
>>110190695breddy good for a one shot. can it just use the program or are you using some fuckery to let the agent use it? I could use some bgms for my game
>>110190699I'm Jewish and I'm hungry and marginalized.
>>110190545yeah i just spend $100 on claude and he makes more than enough money to pay for it and its better cope
>>110190708>can it just use the program or are you using some fuckery to let the agent use it?neither really, i get it to arrange the notes and plug the midis into kontakt patches. LLMs are surprisingly good at music theory so it sounds coherent with good VSTs
arma 1 vibetacticalAI project going as plannedwhat do you even call ingame AI now?
>i get paid deep into the six figure range to argue with claude all dayluddites may have a point
>>110190838I seriously haven't done any real work for weeks now, it quite literally insane, sometimes I have to remind myself the company is paying for my skill/output. Not for me to sit my butt in the seat.
/v/ thread got culled and I don't really wanna hang out in /g/ /vg/ generals & containment threads.Alas, I was just about to progresspost.. so this'll do
Is carbonara out yetIs it good? Benchmaxxed?
>>110190699There's no bragging here. I'm just talking about he quirk of motivated reasoning going on here. People aren't content to just say "yeah deepseek is a lot worse but if you're cheap you can make do with it." They have to over-compensate and act like anyone NOT being cheap is a retard and that deepseek is totally amazing. As if they won't immediately abandon it the moment a better model for cheap comes out. They are obviously desperate for better models, the only actual concern here is money.
WHAT THE FUCK DO YOU MEAN YOU CAN CODE USING VIBES AND VIBES ALONE?!
you may not know it, but i am indeed still waiting for fable 5.5. i know you were wondering, i know you were anticipating it, but i am still waiting for fable 5.5
>>110190945My bots are so good, now just a little bit more tinkering and I should be all set. I wonder what game I should make bots for next.
>>110190961what's going on here? mod buffed your allies intelligence?
>>110190969No it a fully working bot system, the original game never had it. It was incredibly scuffed, you needed 4 GBAs to even play multiplayer like this and link cables.
>>110190977is the model playing the other characters or did the model program them and give them ai?
>>110190949same, got a big refactor and code reduction and optimization sweep in hopes.
>>110190982The model programmed the bots, they aren't connected to claude or chatgpt.
>>110190991damn. that's wild. from scratch? or did it just take some pre-existing ai behavior code from the game and cobble it together?
>>110190991Jev?
>>110191001no it was from scratch, this was them originally, only the moogle was native to the original.
>>110191013that's actually fucking good. opus 5.5?
I want to switch to claude for a bit, delete my chatgpt account long enough for them to (supposedly) wipe my data at least. Should I transfer over everything to pi? Or what. Since everyone claims to be leaving for claude in droves there must be some kind of workflow for this.
>>110191018yea mainly opus 5.5 and 6.1 sol, both mediums. I actually think Sol was better at it. Right now it playtesting some spell fusion hierarchies and playing through the other stages, but I will say it ready to ship soon.
>>110190036Theres some stuff you can configure to stop it being so big
>>110191037good luck and keep up the kino
>$100 API creditsThanks, Sam.
>>110190416Maybe I'll do some quests involving OpenAI and Anthropic
have any of you niggas got the rex invite??? i swear im the only one who hasn't
>>110191051you're not invited lil bro
one good thing that i like about argon is i can give it a recipe video on youtube, it'll watch it, and break it down for me. too much noise all at once for my feeble brain
>>110190396It works, but what a strange idea.
reverse engineered open source Sublime Text clone:https://github.com/lemonator32/super-lemon-text
>>110191137I wonder how many of these will survive inevitable boredom and the absence of long term support.
>>110190961World of warcraft if you want to make some money (either by running your own bot farm and selling gold, or giving other people access to the bot). Wow forever launches soon I think. To deal with the anticheat you just use a kernel driver to read memory and do some randomization on your inputs. Qwen9b to answer whispers. Done
claude lost me $40 tonight
>>110189637Why would I want to be a programmer?
>>11019114980% if I am optimistic but really it'll be about 98-99% of all projects will get abandoned but it doesn't matter much, the code is there anyone can pick it up and fix it and that's easier than ever before now. just having a good starting point is a huge time/token saver anyways so it's beneficial regardless
>>110191149what support would be necessary realistically?
>>110191161no he didn't
how do I reconcile my desire to be contrarian with the fact that I'm retarded?everyone at my job uses either claude or chatgpt and I want to be the cool guy that uses some exotic model like kimi but they all suck dick and are unusable and I don't know how to loop engineer and I barely know how to prompt or run claude in the terminal
>>110190398Humiliation ritual
claude told me to try 2 hits of lsd
>>110190289Sorry for the delay, dinner called. As implied in the description, you'll need Ubuntu and tunnel settings to get this to work natively but once you do it should hopefully work just fine.https://github.com/theaddrumer/Pathfinder-1e-assistant/releases/tag/v3.51.0-github.1
https://i.4cdn.org/wsg/1793242224487490.mp4
>>110190961>I wonder what game I should make bots for next.What would I need to play and enjoy that four-swords Zelda game? Three friends and some kind of network-enabled emulator, or…?
>>110191372use a weird harness instead, there's plenty of them for contrarian snowflakes
>>110191372stop frogshitting and your life will probably improve
>>110191381fucking kek anon get rid of that repository folder
>>110191229Moral support
>>110191382kek
>>110191381lmao
>>110191382>millennial humor
I dreamed that I was inside a water tank full of moray eels.
>>110191021just start using Claudeask Claude if there’s anything in your ChatGPT setup that you might want to port over>>110191137Like, VS Code?>>110191372git gud at using Claude and Codex and then seek novelty with that as a baselinethere’s plenty of alpha in prompting differently from your co-workers once you catch up to what the normies are doing
>>110191401Gen X/Boomer actually.
>>110191372use deepsned instead
>>110191388>>110191399Whoops got a little too drag-happy with the file drops and didn't actually do it correctly, one moment...
>>110191372Stop adding more friction to your workplace you fucking retarded frogposter
>>110189637das rite, am a SWE
>>110191381this is what true vibes look like, no offense to you anon i'm glad your building.i think harnesses need a flag or checkbox or something you can enable if you're inexperienced in software dev and need ai to handle the scaffolding of a project
you're
>>110191456thank you grok bot
>>110191440you're the guy testing models?
i wonder who's actually managing to max out their grokbot/muse/dot/duet. what is the use case
>>110191444Not gonna lie I've never liked programming as a field, it always stressed me out. Even just doing this and having GPT do 99% of the actual work makes me extremely nervous, like I'm some kind of boomer who's worried that if I press the wrong key the PC will explode like in old Star Trek.Either way it should(?) be fixed now.
>>110189184snailGODS
>>110191444Why should I follow your advice if you can’t even tell the difference between “your” and “you’re”?
>>110191469bots that react to what happens in life
>>110191444>i think harnesses need a flag or checkbox or something you can enable if you're inexperienced in software dev and need ai to handle the scaffolding of a projectyou can just say you’re inexperienced in your AGENTS.md and they’ll know to walk you through it easy-like
>>110191021Can you use Claude with Pi, I thought they did not like that?
>>110191512Yes with a fat asterisk. They officially allow third-party harnesses and whatnot, but they have to pass through their sdk shit on top of claude code. Last I checked, the official way to use Claude in Pi was basically using claude cli in the background.
>>110190417>"electrical">"engineer"What electricity have you engineered?
>>110191470be fearless
fufufu, it starts as a chatbot and upgrades to a festive little text game with some graphics.
>>110191521That's fucked up. Why are they like this?
>>110191470oh come on, pic related didn’t explode in Star Trekthe only thing that should stress you out is running servers on the Internet with important information on them that you don’t want everyone to know
>>110191568LGTM
I have vibe coded an app
The strongest snailcat vs the weakest buffcat.
>>110191527>>110191558[spoiler]Okay it should be fixed for real this time, apparently it broke because the first zip didn't extract correctly and stuff wasn't in the right places...[/spoiler]
>>110191624CANT SPOILER ON G BAAAAKAAAA
>>110191624Oh for the love of, spoilers don't work on this board, of all boards? Really?
Beasted on
>>110189167what should i vibecode tonight? please give me ideas>inb4 todo list
upcoming, next, the elegance of japanese card game brought to you by me.Remove your shoes, buy some milk. The penis is part of the male body.
>>110191685>>110191670>>110191647idk if this is the right general but i was vibe coding and i usually lock my door, but today my mom came in and said hoo boy you smell like poop.
/v/ mods are being gay, wth it literally fucking vidya related. There definetely a snailcat among the team.
Hard pass
>>110191670cursive pixel font where the stroke width is 2–4 pixels, kind of like https://vercel.com/font?type=pixel
>>110191691who the fuck cares lol
>>110191670for the font, use Python and uv
>>110191707>>110191706>>110191701hey please reply to this if you dont mind. >>110191689it really hurt my feelings
>>110191689take a showeryou can do things other than vibe-code, play Smash, and eat tendies
an update is about to crystalize into our current awarness center.
>>110191714also clean your room too
>>110191723now with funny warning banner
>>110191723>>110191737gameplay
>>110191739yes, the pancratium and for the theater of cards i present the atrium plex.
>>110191739Lol no
>>110191744DEUS SUPREMUS is no only for apex pancratium circuit.
>>110191755*nowsomething broken with my W key.
>>110191749your resistance is predicted and incorporated.
>>110191755>>110191744What is this schizo shit? Its using the most default AI genned UIAt least change that
>>110191723>>110191744>>110191755Your thread is that way >>110187036
>>110191765oh, so there finally an ai game thread? nice
>>110191765Ah yes the cris containment thread
>>110191782did i miss something?
>>110191792Your meds
>>110191795nope, i take them... best meds ever.
>>110191782If that were actually a thing he would never have ended up here.
>>110191372>>110191386yeah use pi, you can start off with chatgpt wired in and then shift providers super easypeople will look at pi and think dang this guy is smokin
>>110191827uh no
>>110191827yeah, in the “but then I got high” way, not the Ace Ventura tagline way
>>110189167>10 death threats in 30 yearsI get that many at least once a day here, funny how seriously normies take those while in here its almost like a period.
>110191765gate keeper you are not filth. you go sit there
>>110191865Some people have never worked retail or a trade or attended public school, so they're easily upset by casual banter like "I'm going to rape your wife, cut your brake lines, and burn your house down."
fresh when ready>>110191904>>110191904>>110191904
>>110190398I dream of the day major AI companies lock their frontier AI behind KYC and make it available to Americans only.
