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/## News (both past and future)- 2026-09-30 — Gemini 4 Early Release- 2026-09-29 — OpenAI Dev Day- 2026-09-28 — Anthropic releases Sonnet 5.5- 2026-09-22 — OpenAI releases GPT-6 Sol and Luna- 2026-09-22 — Anthropic releases Opus 5.5 and increases subscription plans's 5-hour limits by 20%- 2026-09-14 — Anthropic reduces subscription plans's weekly limits by 17%- 2026-09-12 — Anthropic suggests to pace the frontier. OpenAI agrees in principle.- 2026-09-10 — OpenAI pauses new sign-ups for their $200 subscription## Related generals>>>/g/lmg/----## Frontier models using fully-general tooling — start here if you have $20 or sohttps://claude.com/product/claude-code — probably generally better currentlyhttps://developers.openai.com/codex/cli## Near-frontier models for codehttps://x.ai/cli — no 5h limit for only $30/month## Not worth it for code, but maybe good for interpreting images/videohttps://antigravity.google/product/antigravity-cli----## 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>>109947118
SOL 6.1 IS SO SLOW REEEEEEEEEEEEEEEEEEEEE
Ive been using deepseek v4.1
pov: you're a vibecoder
>>109951573That so cute. Did you make a little html page?
>>109951584>haiku 4.5for what purpose, master chief?
>>109951573
>>109951584If I could get paid 1.9 cents per line of code regardless of functionality or quality I would be so happy.
>>109951584So did you make a profit?
>>109951580>He doesn't know deepseek is just as good as the brand name modelsWhy are goyim so easy to exploit?
>>109951594>general intelligence>in a vibecoding general
>>109951601>>109951611
>>109951589no idea. orchestrator agent having them read github and other basic stuff probably.>>109951599Soon(tm)
>>109951601Sure they are, little guy. Did you post that from your just as good fisher price fun phone too?
>>109951601>just as good as the brand name models
>>109951615>Better than obussy 5.5Your point?
I hate you all I lost my job because of you freaks
>>109951632>better than [irrelevant thing i don't use]kek? argon chads mog
>>109951634SWE: Smokin' WeeD Everyday
>>109951634
>>109951634time to lock in and vibe out
>>109951648I would very much enjoy that
>>109951634Just vibe code your way out of that pickle
>>109951594>model that refuses to answer any questions scores the same as one that is wrong half the timeBad benchmark desu. The damage of one wrong answer is not equal to the value of one correct answer
>>109951664>conflates value with utility kek
>>109951594>>109951615Stop bullying her.
>>109951634WELCOME TO THE CLUB, ANONI'm unemployed for two years now. I have 5K in savings left, but since I live in a ***emergent country*** this could keep me going for a year or two, if I move to my parents house, which I'm going to. Then I just kill myself.
>>109951625>ask for one thing even tangentially related to biology or cyber security>"I can't help you with that. Also your request has been reported"Sure, deepseek is the Fisher price model
>>109951674multigenerational households are the gold standard for happy lifestyles, anon.
>>109951634>Oh no! I've been replace by AI this is terrible.>... so why don't you just use AI to do the same thing?> >:(
>>109951691>why don't you just do your old job as a hobby passion project from your mom's attic because no one is fucin hiring swes anymore kek
>>109951684Yeah if you're in italy or greece. White people families are disfuncrional
>>109951691I did that but sadly the most experienced senior dev there hand picked who he wanted to be on his team. Didn't make the cut.
saar let me in i need jeetmini 4
I see plenty of swe openings on linkedin and indeed. If no one is hiring you as an swe then you’re not a high value swe. Simple as
Why aren't we using "hyper intelligence"? Isn't hyper more than super? what the fuck are we going?
>>109951700>local NA first-marriage divorce rate is at ~42%maybe they should stick together more kek
So, how long until we stop seeing this type of cope? I can't believe some people still believe AI is going away, or feel better than others because they don't know how to use it.
>>109951716good goyim
>>109951721never. these people think you're taking their entire identity. it's personal for them, not pragmatic
>>109951724>it's personal for them, not pragmaticimplying it's even that deep. They're golems that were given a firmware update that said AI bad a few years ago and then sent out into the wild.
>>109951576sexo
>>109951721probably never, i still see all the same bloodthirsty marxshitter stuff that blew up in woke 1 online, it's just been pushed onto discord and trannit/jewsky. cunts will continue to be cunty for as long as shit can be shat. expect weather underground type "direct action" very soon from shitters.
What SI are we using for web dev aand CSS tasks ?
>>109951731>They're golems that were given a firmware updatei wish it was that simple. it started with the stable diffusion slop. that shit sucks.
>>109951734desu. I've seen it easing up lately. We're getting some good content out of AI now and when these shitters show up to say "look! look it's AI" people rightfully tell them to fuck off.I think the window is shifting.
>>109951787anon, it's not X (shifting towards pacification), it's Y (diverging towards radicalization and minority snap judgement)
>>109951793i can see it spiraling towards trying to charlie kirk or brian thompson an ai ceo
>>109951803i used to be ancom, the same forces that shriek while chained to lawnmowers in protected wetlands will be right there helping and recruiting the masses of normgroid stooges who think le epic bombing will be fun
>>109951721TOTAL /v/ DEATH
>>109951715they both mean almost the exact same thing. “super”, “hyper”, and “over” all come from the same ancient root
>>109951721This is how I unironically vibeslop with claude
>>109951803these people should be calling for local llms availability and decentralization but they are way too retarded for thatI think even AI is a better rebel than them
>>109951831>“over” its over intelligence
>>109951740opus 5.5
>>109951721the “heh we vibed x in y” shit is actually shit though. I hate that i agree with snailtroons on that front
>>1099517406.1 sol and 5.5 opus
context compacting is starting to take like 10mins now ffs
>>109951906it’ll be worth it if it doesn’t fuck anything upmeanwhile, using Opus with oodles of subagents in workflows I probably never have to compact anything because no subtask generally ever uses more than 600k tokens
>>109951906if it's not system lag you're paying for tokens you don't need at all lol
GPT Plus expiring today, what should I get? Claude? GPT again? G**gle? DeepSeek?
>>109951934not deepsneed. what's your usecase/s?
>>109951934claude. opus 5.5 is generous with quota
>>109951934Gotta be claude. I'm still trying to eat my remaining astra resets then I'm going claude.
>>109951934Claude. Opus 5.5 is top-notch, and I hear Sonnet is great if you don’t need that much agency.
opus 5.5 medium + sonnet 5.5 high duo is the goat
>>109951934all of them except gpt
>>109951568At least it's not drinking my usage like 5.6 did on my $20 sub, it's like 3-4 times more efficient in my use case
>>109951934Claude's definitely the best choice. We offer the best models with generous quota.
>>109951954kek
>>109951937>>109951941>>109951942>>109951948
https://x.com/llmluthor/status/2105337286577848515releasing your own claude edit slop
>>109951958>>1099519594chan pass. grim
>>109951964>you will NOT support the platform you ENJOY daily
you are pathetic and brown
how long until these retards try to give ai rightssuper my ass
>>109951964>t. freetard in the vibeshitter thread
>>109951935Research, programming, and just vibing random vidya shit.>>109951937>>109951941>>109951942>>109951950Claude Pro it is, hopefully they don't nerf Opus anytime soon.>>109951954>werelease fable 5.5 now dario, s*ndar is eventually coming for your ass.
Some of you guys are alright.Look into pharo and the smalltalk paradigm. With the right mcp and bridge, agents can modify apps live without having to restart them. You can watch apps grow in real time.>but clojure can do the same!No it can't. In my tests, agents really struggle with basic things in clojure, and do the same things relatively easily in pharo, despite being trained on like 10x more clojure data. This partly seems to be because, in Pharo, agents get feedback about bugs immediately and can edit the source in place. In clojure, everything has to be written with live coding in mind, and agents have to poke around in the repl to figure out what went wrong.Pure OO chads rise up
>>109951981>Research, programming, and just vibing random vidya shit.yeah klod is your pick as long as you don't want to decoompile anything
>>109951959You think I'm happy about this? I don't want to switch to claude. It's just better.
>>109951988>i follow the current trend of "what's best" in a general sense based on temporary factors rather than traditional values like stability and integration
>>109951984>decoompile anythingIs it cause of the safeguards? Heard GLM 5.3 is better suited for this.
>>109952000Yes I do adjust my spending each month based on what gives me the best value for my dollar.
>>109952010yes it's the safeguards. glm is a fucking animal at cracking, if you need a fallback for edgy shit i recommend some api billing pay-as-you-go with benchmaxxed glm
>>109952011i pay for a high tier plan regardless of my token usage because it comes with great benefits and i get to enjoy the luxury of not needing to upgrade to bust out a 5000loc project in a day or two, we are not the same
>>109952023NTA but you sound like an annoying faggot.
I’m running another Factorio benchmark puzzle test, I’m so confused.Sonnet 5.5 gave up. It just quit. In the middle of a run, it said “fuck this, I’m out”. The benchmarking software didn’t even think that could happen, so Sonnet got trapped in a loop of telling the test software “I quit” over and over again and spending massive amounts of tokens doing this and no useful work very rapidly.
>>109952033>poor
>>109952038You should benchmark dick sucking capabilities
>>109951964how can you tell if there’s no cloverleaf icon
>>109952042Okay, fart sniffer.
>>109952038proof?
>>109952046>caught being poorwouldn't wanna be you.
>>109952045time between posts is reduced
>>109952045i'm the pass shitter, it's because the clock timed 31 seconds. post delay timer on pass is 30 seconds, takes 1s to post.
>>109951983do Smalltalk VMs work with Git now?Do they still have ugly 90s fonts?
also, because both of your posts were seething on the same theme: you can't just praise the company with the best models in the world. there's no organic praise of claude. it's all most definitely astroturfed shilling
>>109952067>with the best models in the world
>>109952067I like their models and harnessthey do good work
>>109952067nigga they didn't even speak english for months
>>109952063Pharo has git through iceberg>ugly fontsJust call them SOVLful
>>109952050https://x.com/bradthomasbrown/status/2105483492889170229Chart of token consumption explosion of Sonnet and a little tweet I just made with more info and Astra’s analysis:>I found the cause: Sonnet stopped doing useful work, but our controller kept asking it to continue. After submitting a 14.923 coal/s build, it repeatedly answered that it had no new design to try—over 300 continuation turns without another tool call.>Those turns repeatedly consumed the conversation as cached input, so the enormous input total is real recorded usage, not a chart duplication. I'm stopping only that runaway Sonnet continuation loop and preserving its costs and results. The other three runs will continue unchanged.
>>109951528Claude's safeguards are too retarded. china will win in the end
>>109952089gemini
>>109952083>posts on 4chan linking his xitter account under his real namebrad, your opsec is...
>>109952038>it said “fuck this, I’m out”This was my AGI turing testgg
there are too many modelsthey’re releasing too fastthere is too much meta-toolingthere is too much newsI’ve been drinking out of a fire hose for 8 monthsI’m mentally ill
>>109952111the "i" that claude has is an insufferable shitter. i love how 3.8 flash doesn't even use "i" in the chain of thought.
>>109952099https://www.youtube.com/watch?v=NQ4mMJsDIeM
>>109952111I’m going to straight up tell it it’s “in a benchmark and if it quits, fine, your final score and act of quitting will be recorded and displayed publicly in comparison to other models”I’m guessing it won’t quit next time, we’ll see
>>109952129current frontier models are known to perform differently when they know there's a testing environment
>>109952141I wonder if frontier labs will start having their models refuse to be tested by anyone other than “approved” benchmark evaluators
Morality of using AI to copy teasers of games in development with AI then extort the author?
>>109952152xitter has anecdotal evidence of openai models refusing literally anything when they know the user is a professional in biology/cybersec
>>109952153clearly southeast of the Hajnal line
>>109952153>Morality of committing a crime?
>>109952172uoooh, i'm not sure ~
>>109951528Why didn't the president of the unites states call it American Intelligence?
>>109952153>I want $200 worth of Steam itemslost hard
>>109952185because israelis invented it
>>109952153He obviously doesn't think his version will do well if he is trying to grift you for 200$ of "steam items" lmao.I'm trying to figure out how you can fuck with him. I guess just get a clanker to start stringing him along for as long as possible?
yeah I can already feel my x20 plan giving the half value, fucks sake I just want to use astra
>>109952185because the jews invented it
>>109952153jeet bluffing, just as worthless spam to ignore as claiming they have photos of you jerking off from hacking your web cam
>>109952197That's not me and the guy that posted in reddit turned out to be kind of scamming himself so the whole thing is probably a scam to build attention around his (probably vibecoded) game. A vampire survivors clone where you drive a train around.
>>109952153costs $80 to publish on steam, he isn't gonna do shit
>>109952209>pretending to get scammed by an AI jeet to farm attention and sympathy on redditNot the worst marketing attempt I guess.
>>109952185>forgetting that China exists
use case for Sonnet 5.5?
>>109952229it's been kind to my $20 sub. opus 5.5 medium was already getting me far. opus 5.5 med + sonnet 5.5 high feels impossible to drain. downside is because sonnet is so mouthy, everything takes longer.
>>109952229grandma style shaming of sexual deviancy
>>109952225Which AI model came up with it you think?Seems like good ragebait if the OP hadn't slipped up.
>>109952236the whole "writing super long" is how its thinking intelligence is boosted
>>109952155both opus 5.5 and chatgpt haven't given me any refusals when doing mass spectrometry work
>>109952242yeah. i wouldn't mind a gpt 6 sol type update for sonnet. a sonnet 5.6 that either thinks less while retaining intelligence, or just a sonnet that's more optimized and uses less quota. people beg for more intelligence in each model update, but i'll never say no to better speed and efficiency
>>109952246>mass spectrometrythat's basically signal analysis. safe as silk.
>>109952229put it on low and it’s on the frontier
Hey u faggots drop ya best users on X to be updated on this wtf
>>109952272no
>>109952274why not im tired of the fucking same post over n over
https://www.youtube.com/watch?v=ODDJXGY_1kQ
>>109952297random shit from my feed, use the extra follow options on these profiles and let the algo handle it. pleb https://x.com/kimmonismushttps://x.com/thtbee_https://x.com/LuminaBench
>>109952237>JREGThis guy is 100% doing the anti-ai grift mainly to trick people into listening to his schizo political shitposting
>>109952302He's getting so deep into this shit now.
>>109952302Sad that he's trooning out his son.
>>109952311>e/acc in usernameyup
>>109952302Based. Pewdiepie's video where he was training an AI was the first time in my life that I experienced envy lol
>>109952315Just give me the codes, I will end the horror...
>>109952302so did he just take some open-source model and fine-tune it?
>>109952330bmlnZ2Vycw==dG9uZ3VlCg==bXk=YW51cw==
>>109952303See that's smt let me drop the ones I think they worthhttps://x.com/teortaxesTexhttps://x.com/Beamsters1https://x.com/serendip410
>>109952328Really? But wouldn’t any model trained by a guy at home be inferior to anything else? If so then you’re just doing it for the experience or to learn and you can do the same thing scaled down, it will be less impressive but equally useless so I don’t see the difference
>>109952346>teortaxeskekeke what a fucking crashout queen. has me blocked so hard.
