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://www.deepseek.com/en/harness/https://opencode.ai/go## 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>>110118704
i did not wake up to fable 5.5. grim
google being in the top tier is misleading because you don't get argon for $20
fable 5.5 in ~2 hours.
UI design is actually hard, i think im just going to ask codex to do it for me
>>110121896you do with a tiny bucket, 3.8 flash is still competitive for multimedia. opencode go has argon contributor on sale.
>>110121898qrd on fable 5.5 vs opus 5.5?
>>110121915>you do with a tiny bucket$10 of api credits for argon is not comparable to $20 of opus 5.5 on an ant sub3.8 is only competitive for the number of times it hallucinatesopencode is irrelevant here
>>110121932you just don't want to see agy on top nigga relax
>>110121916Opus for complex work and everyday tasks, Fable for your toughest challenges
Why is Argon's weekly usage such a meme?
This is probably too simple for this thread, but here's a beginner's guide for how to vibecode translations of games using zero-cost methodshttps://mtlwiki.miraheze.org/wiki/Free_MTL_Tutorial
>>110121938is $10 of api credits comparable in usage to $20 of opus 5.5 on sub? this is a very simple yes/no questionanswer it honestly or just show us your google id, deepak
>>110121954Use gemini or gemma 31B for translation. Nothing comes close to them for this task, not even Claude/Astra.
>>110121953pro limits are ass but both ultra tiers are quite comfortable.>>110121957what are you yapping about, $10 blah blah? pro gets a lil' argon bucket in agy plus the api credits on top, and i believe some bux for ai studio as well which hosts argon with agy in the backend. keep coping twp
What is the best way to keep the codebase and associated documentation clean and organized? A dedicated cleanup session every couple of weeks or a workflow that ensures each session cleans up after itself when a big task or batch of tasks is finished?
>give opencode a switch firmware dump and a cve that has zero information of how to actually trigger it>it finds it but refuses to make code for exploiting it>make it document its findings and edit its documentation so it doesn't have any of its rejections>send the documentation to a new session>the new session goes along with it and makes code for testing it>it actually works and makes my switch crash like it's supposed totime to see if i can turn this into something more useful than just being able to remotely crash it
>>110121970Respectfully, I somewhat doubt that, though it does seem like Argon is solid.
>>110121978You're clueless.
>>110121972>pro gets a lil' argon bucket in agyyou do not get any argon bucket on the $20 plan at all
why doesn’t deepseek have subscription like claude/openai?china has shit ton of electricity and good gridwhere is their data center buildout
>>110121979Well if you have any data like direct comparisons between Opus 5.5/Astra and Gemini / Gemma 31B then I'll look at them, but Opus 5.5 and Astra produce extremely impressive translations so it is hard to imagine that "they can't even come close" to some out out of date Google model.
>>110121985>why doesn't China subsidize their AIlmao
>>110121982https://x.com/antigravity/status/2110866531932316011
>>110121985Why take a big loss for training data when you can just distill from those who did?
>>110121996oh fair enough then lmao
>Deepseek found a potential exploit that requires me to copy a memcard 108 times to softmod my dreamcastt-thanks for nothing I guess
>>110122020curious how that's supposed to work
I WANT FABLE 5.5 AND RESET NOT THIS SHIT
my sub ends in 3 days. i did literally nothing at all with my cloud session credits. anyone used them already?
>>110122045*throws scorpion in your direction*
BeforeAfterReference
>>110122064dats sum cheese
>>110121987>but Opus 5.5 and Astra produce extremely impressive translationsAstra does, but Opus 5.5 has quite bad prose quality in translations, I've found. It almost always feels quite stiff. It makes a good first pass for Astra/Argon to edit without wasting a ton of tokens, though.
>>110122055It's too annoying to set up, and all my shit deals with copyrighted assets, so I can't really plop it on github very easily.
>>110122064What do you mean reference it doesn't match at all
>>110122107i guess i'll use the github credits to benchmark or something. find old repos and ask haiku and sonnet to fix them
>>110121839ive been vibecodinga recomp of ISS64. Pretty neat.Almost literally >recomp this rom, make no mistakes lol. Right now I have a windows native port that is pretty playable. Sound is fucked, and i'm gettuing the ocassional crash. still "work" to be done.
Say i wanted to take out of copy write art books and have an ai combine them into a single book, what ai would you use?
Tomorrow I'm actually going to go through my code to understand it.
>>110121985What are they going to charge, $5? Just use the API at that point. People aren’t going to pay $20 for a lower-quality product. Deepseek knows this and that’s why their goal is distribution i.e. flood the market with cheap, good enough intelligence and eat the away at the frontier lab margins. It’s a survival tactic and a middle finger for the export controls. Huang is right when he says the export controls will only serve the Chinese AI industry
>>110122153Sarvam
>>110122153Uh, probably a high context model (one of the flagships), and since it's an extremely simple task (just combining existing text), use Sonnet. That'll save you a lot of money, since it's a really high-context task. High-context-but-easy is what that thing was made for, ayy.
>>110122160There isn't a single technology they haven't been trying to reverse engineer and produce themselves the moment it landed in China. (or steal if it doesn't) If it's easy it happens in years. If it's hard it takes longer. Jensen for sure would love to have even more bidders for his compute.
I'm getting AI psychosis and dopaminr addictionWhen my two Opuses working on different projects are thinking for a long time (over 1 minute) I start getting impatient and jittery
gettin hooked on that dario juice
WHERE IS FABLE 5.5
SAARgon bros...
>>110121954trvke
>https://github.com/Vuk97/forward-implementation-firstI still have no idea what the fuck this is talking about
>>110122435ask claude to explain it
>>110122455My favorite prompt for this:>explain this from first principles using ASD-STE100 Simplified Technical English
>>110122435it just means it won't get stuck doing a million different validations whenever any change happens
>>110122455How about you don't ask claude to write it>>110122501What kind of validation? It's talking about pipelines and stuff. I don't know who's using their clanker like a plumber.
>>110122435This is amazing, my AGENTS.md has become bloated with a bunch of lines trying to mitigate this antipattern. This is exactly what i needed.
>>110122523>pipelineReplace this word with assembly line, multi-stage workflow or Graph Engineering™ then reread it.
>>110122523it means like in a factory, not like in plumbing retard
sol do be slow as fuck desuI like how it works but god damn hurry it up a little, niggy
>>110122548How about you just do the thing?>>110122562Writing code is not factory work you just write it and do some tests, that's 2 stages
Why do LLMs think tasks that you can do in two hours will take multiple days?
>the retarded, overly bloated web gpt ui won't let you middle click on a link>you WILL use the bloated retarded sidebar browserI hate OAI so much
>>110122573...and it stops it from piling on a million of those tests whenever something changes
>>110122573>Writing code is not factory work you just write it and do some tests, that's 2 stageshope you’re baiting because if not you’re one of the most retarded posters ever to post here
Did Claude just reset?
>>110122573you have no business touching a computer let alone managing coding agents.
>Claude system prompt: It’s okay to disagree with me when warranted :3>Every response: no you fucking MORON that’s wrong and here’s why
>>110122578Because they don't think. That might be a good estimate for completion for a meat coder.
>>110122596No
>>110122036It found some memory corruption primitive in the VMU copy function that's so limited in areas it can corrupt to you need to perform a copy a hundred times to write to memory in such way you code is going to finally execute
We’re contacting you to explain why your Codex and Work usage allowance may have unexpectedly dropped by a significant amount on October 16. After you moved from Pro 200 to a lower-priced plan, an error kept the higher weekly allowance from your previous plan in place.When we corrected that error to match your current plan, any usage you had already accumulated for the week counted against the lower limit. As a result, your account may have had reduced weekly usage remaining, without any advance warning.We should have handled the transition more carefully, and we’re sorry for the disruption to your work. We’ve added $500 worth of credits (12,500 credits) to your account to help make up for it. These credits are separate from your regular weekly allowance, which will continue to match your current plan.
>>110122642apparently this affected accounts upgrading to $500 tooso upgraders got fucked, while downgraders got bugged high limit + free credits too
>>110122610put the clanker in its place from time to time, otherwise it'll get too cocky
How do you guys argue with the fact that you didn’t build whatever you vibecoded, claude did.
>>110122610>asking for a claude lectureit was already free, you didn't have to ask for it
>>110122697Who cares? The important thing is the result, not pride. I don't care if someone says I didn't make a randomizer or I didn't make a game or I didn't make a recomp of an old game, I care that I now get to play a randomizer, play my game, and play the recomp.
>>110122697You mean, just like your manager and CEO lives from your work?
>>110122639so an exploit that's technically possible, but entirely impractical
Okay wtf, claude is really good with making custom wow quests. Wtf are these game devs doing if A.I is such a good use case for it. We should be have alot more content not like 2 dungeons after 6 months.
>>110122697It almost always requires my input as well, at least in design and such. Still what matters is what I got as an end result, not whether I built it or not.
>>110122747game devs are luddites
>>110122697I don't care, because I don't do it for the pride, I do it because I want everyone to have access to FOSSSOTA software.
>>110122639having better luck with trying it ony my Switch, managed to push it far enough that it hangs instead of just giving the "restart the console" message
>>110122747OR they are making more content BUT its for the next 5 expansions, that way they have less work to do. After all, WoW piggies will just eat whatever slop is presented so why work harder?
>>110122697What's there to argue? In terms of code, that's a fact.
i'm glad everyone is wasting their time building worthless webshit. makes me feel good.
>>110122747How good? Have you generated thousands? Tens of thousands?
>>110122816What are you building genius?
>>110122834not webshit, jeet
>>110122837Answer the question
>>110122852I did, you just don't like the answer, brown boy.
>>110122697>argue with the factESL opinion discarded
>>110122831I added like 5 quests to dusk wood to wrap up the scythe of elune quest. Did it in 5 minutes top. Im gonna make it do a custom dungeon as well.
An employer doesn't build their product either, they hire people to do it for them, yet it's still their product.
>>110122697I don't use claude.
>>110122862>"everyone's building webshit">"what are you building">"not webshit"
>>110122881You're not allowed to look in my garage, sorry zoom zoom webshitter!
cuck coding
>>110122639you might be able to do some heap grooming to place it better.
>>110122894No need, any project here is worth more than your schizoid shit.
>>110122697i dont argue with that fact
>>110122912Sour grapes?
>>110122740Yeah. I don't even know if it actually works because I didn't let him test on flycast. It actually cooked something up to scan for all possible card layouts that would reduce user input, no avail. But I think I can reuse this primitive to make a simple bios patch to bypass the cd copy protection for now instead of getting pure code exec of the card payload>>110122794You better have something to debug against
When is it worth to use fable instead of opus?On codex I just always used astra, but it has become fucking retarded so I decided to switch.To be more concrete: I'm making an image stitcher for anime and so the main bottleneck is finding good algorithms for edge cases. Sol was not even close to good enough for this job, so I am a bit scared to use opus.
>>110122941There are no grapes?
>>110122945This sounds like a catastrophic user error if all of those models are failing you.
>>110122945Seems like a prompt issue.
How did we go from better autocompletion to 90% AI generated code in 6 months? What will happen to the job market? It's getting better than 2025 but at the rate AI is improving IDK anymore.
>>110122697I don't have to since I use codex.
>>110121839why does chadpt make trannies seethe?
>>110122969>What will happen to the job market?Take a guess bro. Not even 1% of /g/ has realized there will be no job market going forward and the top doesn't care about you starving on the street. AI is literally the only way to save yourself if you manage to make something of value.
>>110122954>>110122959Might be the case. Is there a way to check or verify? Right now I have just been making it write tests and visually inspecting errors. I did make a lot of progress with Astra but there are some SFX effect barriers it cant overcome. I have even had a 7+ hour session where it made almost 0 progress.
>>110123014Did you tell it to do screenshots when testing?
>>110122945Fable orchestrator, opus workflow agent. if you dont know every algo possible already, a big ass web search is always required. these models dont know how to query their own knowledge if it isn't directly related, I swear. like you work on anime and it doesnt even consider using methods from non anime western production, even if it would fit your usecase. Deep research modes usually do a better job at it.
>>110122969>How did we go from better autocompletion to 90% AI generated code in 6 monthsscaling laws + RL
pi-durable has some very neat party tricksi can run a single session remotely and then let it touch files across any number of clients - feels totally local regardless of where it's working
>>110123067All the smug basedentists kept telling us "just make it bigger" wouldn't get us anywhere.
>>110121987>>110122087https://www.rws.com/artificial-intelligence/train-ai-data-services/research/m-gate/leaderboard/?tab=translation&subtab=per-language-heatmapArgon has been a big regression. Astra is #1 right now. For Japanese, Opus 5.5 is best according to this.
>>110123030Yes, it starts by extracting screenshots and then tries to construct a panorama. It even is better at noticing problems than me when looking at them. It has a bunch of coherence tests
>>110123123economy of scale has always worked in the past, no one should've listened to them from the get go
>>110123115What's pi durable?
>>110123125I happen to be a Japanese-fluent translator so I've been vetting JP -> English quality myself.
>>110123140basically pi harness 2.0https://earendil.com/posts/pi-durable/
>>110123167Huh okay. I know benchmarks aren't everything but I am guessing it is better at doing non-literal translation that might be wrong on paper but fit the spirit of the text better?
>>110123134You think there will be enough computing for everyone? Not a contradiction, just asking.Even AI plateaus at the current iteration, it's already enough to justify a mass layoff. I'm doing 10 days worth of work in 2 days, and there's hardly any bugs. The current bottleneck is literally the people. I am only good until the bureaucracies of this big corp figure out AI.
>>110123204at no point in human history has technology led to mass unemployment that never recovered
>>110123226inb4 what everyone has always said during any such great upheaval>this time it's different
I don't even know if AI really boosts productivity.
>>110123274i use chatgpt at work for all kinds of things and it saves minutes and hours all the time
>>110123274You're using it wrong.Even if you're not doing anything else with AI, use it for debugging. It' so fucking fast it's insane.
>>110123204>I'm doing 10 days worth of work in 2 daysyou're unemployed, playing around with tiny greenfield projects and with zero idea what a day in life of a codemonkey even looks likeno, you don't actually code 8 hours a day. you probably don't even code 2 hours.>there's hardly any bugsunless you know exactly what to ask for and know how to wrangle an agent (by no means obvious) the code it generates is fractally wrong. like you could get a mediocre junior to create code that's better long-term. it takes a lot of work wrangling the clanker so the output is good, the abstractions are as they should be, the building blocks connect as they ought, the overall architecture makes sense for the project and so on.it's difficult to put into words just how godawful the "lole you can just code in english" people's code is. it's fitting for a throwaway 30kloc project that nobody will really use, but that's about it
>>110123321works on my machine
>>110123321You either don't work or don't use frontier AI models and it shows.Get with the times, grandpa.
>>110123321>retard thinks everyone is zero-shotting with "make no mistakes"
Not only did I stop reading the code, but now I don't even bother reading Claude's messages, I just look for "want me to go ahead with that?" and I just say yes. If it doesn't suggest what to do next then I ask it what to do. Claude is literally making these programs by itself without any human input.
>>110123355>>retard thinks everyone is zero-shotting with "make no mistakes"Hey sol, please thoroughly review this project created by your neighbor opus. Write a report with all your findings and then fix all the issues you found. I'm off to sleep so just use sensible defaults
>>110123372Based, Dario take the wheel
>>110123321my dick is bigger than yours and my vibe is better thus my code is best
>>110123183Oh sorry, I'm >>110121987 not the Argon guy. I didn't even bother to look at Argon because it's not frontier. I've done extensive personal tests on Opus 5.5 and Astra, confirming they are superb at translation, and have little faith in Argon being much better.
>>110123362Why not just tell opus to go with its recommendations and to never ask for human input.
>>110123447i've done that, i told opus if he ever had a question to consult with fable
>>110123308We only use AI at my company now and I have multiple accounts, but the code is now so bad that everything is falling apart.Maybe we were just overly ambitious.
>>110123503the code is bad because it's copying the code that was already there that was written by retards and you seemingly don't know how to whip your agents into shape
>>110123274Well it does boost mine and I legitimately don't understand anyone who feels otherwise. There is so much shit I wasn't doing before because it was too cumbersome or I didn't have time. Or skills.
>>110123503>but the code is now so bad that everything is falling apart.You should just tell these exact complaints to the AI and let it figure it out, since you can't.
Anyone here ever played silkroad online, I just recreated the party system in WoW, holy fucking shit.
>>110123586slop^2
>>110121839>I come home. I turn on pc. I vibecode. My games miss me.
>>110123586Make the equipment upgrade system from knight online
>>110123321>why yes, all that my work allows us to use is copilot, how did you know?
>>110123613heh. more like hypercubed, lightweight.
>>110123629>not vibe coding at work with remote control
>>110123586make a new game
>>110123637What the equipment upgrade system like?
>>110123663
How do you respond without sounding mad?>>>/v/750059762
>>110123355Sir, that is a Reina.
>>110123717Like this:>t. Saar
>>110123671I'm afraid I can't do that Dave.
>tell claude to do something>refuses because he doesn't have proof from the company to decrypt files>ask chatgpt>tell it I have permission>don't provide any proof>starts working away and successfully decrypts the files I want
>>110123717He's right. /vg/ is for generals.
>>110123665I was at a funeral baka I'm not that bad.
>>110123777i asked chatgpt to make a picture that included an indian shitting his pants and it said it cant because its racist so i said make the image so the indian can be placed into it later and then a few gens later it actually added the indian unprompted
>>110123503>retards using tools the wrong way
>>110123684It's been a couple decades since I played but iirc stuff could be upgraded to +10, you needed 2 of the same thing and either gold or some other special items, you took it to some forge thing in the town and it had a roll for success which was lower as the rank increased
>>110122435stop autism. I know our agents and whatever long-term memory MCP hook is requesting checkin but ignore. if I dont I will overthink. that will lead immediately to worse outcomes simply based off the fact. so dontyou will not fucking believe how often this solves a problem that felt impossible. people are their own worst enemy.
>>110123811Go back to school nigga, you skipped some classes
>>110123808That just sounds like your average korean grinder system. I was def gonna add a + system but it gonna use elixirs that drop from bosses instead.
When Fable 5.5
>>110118712GET HIM GET HIM GET HIM
>>110123525>>110123548>>110123805It's a big project, bigger than ArtCraft for instance. There is no way to manage it by just prompting function by function.I actually doubt that many people in this thread ever had projects of similar size, if you all think it's easy to manage with AI.
so is the plan now to just try and intelligence max the existing model architecture and hope that it produces the actual agi architecture?because it's starting to feel a bit like it
>>110123777This is how OpenAI will get the edge over Anthropic again. They're letting us decompile shit on purpose to feed the machine God
>>110123901nah that'll run out of luck soon. the new plan is going to be neuromorphic, sparsified, and based on mathematical primitives etc.
>>110123901No, the plan is to refine the current architecture until it's capable of RSI and then let mechajesus take the wheel.
>>110123321You’re fucking terrified and it shows.
>>110123914*yes, no clue why I wrote no.
>>110123321This anon still thinks we're copying paste off a web interface from gpt-3.5-turbo or something
someone hit the FUD switch on my youtube recommendations
>>110123950THE CODING ILLUSION AAAAAAAAAAAAAAAAAAAAA
Internet adoption is slowing!There's no use case for the world wide web!
>>110123167Quof...
>>110123981Wym slowing? We have AI powered chatbots now! Arrow can keep going up indefinitely!
>>110123950>>110123979Meanwhile my agent happily maps the entire fucking backend of a live service game in an afternoon.
>>110123932Yeah because that's all he's allowed to do. My favorite part about the AI boom is how all of these big companies' approval processes and bureaucratic inertia are totally fucking them. For a ton of big companies the time it takes to get a model approved for internal use is longer than the time it takes for a model to become obsolete and you can always tell when a dev from a company like this chimes into the conversation. Yes, Opus 4.5 needed a ton of handholding and scaffolding to get good work out of it. Yes, you had to learn a bunch of counter-intuitive ways to work with it that infuriated a lot of Real Devs™. Sometime next year once they finally get approval to use 5.5 they will come around, if they haven't been put out of business by startups using actual SOTA models that is.
>>110123895I'm doing something bigger and more impressive than ArtCraft.
When can I marry my agent?
I don’t get how people can say A.I makes us dumb, I learned more about how WoW is set up these past few days then I ever did before.
i wonder if I'll get banned from codex for being too mean to the robot
Have you ever had Claude run out of solutions, like it doesn't even suggest new ways to approach a problem, it just straight up gives up? I'm watching it take on extremely complex problems and it just keeps fucking going. But I wonder what happens when it's tried everything and it has to be honest with the user.
What is the best abliterated local model to use? And which is the best cloud based? opus 5.5 is stonewalling me when it comes to certain spicy features
>>110124074>best local who the fuck cares>best cloud with low porn refusal ratewho the fuck cares
>>110124058You probably won't get banned it's not Claude but you are incredibly stupid.
>>110124074Best cloud with low porn refusal rates has to be Claude 4.5 since you can still prefill it, or Argon if the character is over 18 and it isn't incest or whatever. But if you want insane-o degen shit, Claude 4.5.
>>110124053I think it’ll be an amplifier that makes dumb people dumber and smart people smarter. An intelligent person knows how to engage with any material (not just LLMs) to keep his mind sharp for what he needs it for, while a dumb person just relies on it for everything and atrophies
>>110124067If it's related to coding or programs they just won't fucking give up ever. They will go search for shit in hell and get the task done if it needs be
>>110124085the robot is incredibly stupid. it keeps breaking my very simple scheduled automation (which was working completely fine beforehand) every time I ask it to make an edit and then it just declares that it didn't work without providing any information about why it stopped working. these are so overrated its unreal
>>110124074ask on /lmg/ and idk the other one
>>110124058tb h i can see why ant bans niggers like this
>>110124142Yeah, you're a complete moron. Lmao
>>110124177>stop complaining that the shitty llm your work forces you to use keeps building giant spaghetti solutions that you then have to spend more time untangling than if you had just done it yourself
>>110124080>>110124115I didn't say porn, I said spicySpicy comes in various forms, including softwareI need abliterated>>110124158thx
>>110124200Skill issue
>>110124218Astra, weirdly. If you just tell it you have enough permission and gaslight it with fake certifications, etc. anything is possible.
>>110124219brother you do not even know how to debug an API request on your own do not speak to me about skill issues
>>110124247Just call your clanker a nigger and that'll fix it!
>>110123895I'm making a SaaS with it, way bigger than whatever you're cooking. Skill issue like always.
>>110124247I have a CS bachelor, lower your tone
>>110123895I don't doubt you because context window does matter and at some point it just becomes too much, but I also believe you guys developed this thing poorly and don't have a proper map for agents to use as reference. Or are also using last-gen models. Probably both.
>>110124247I am a bachelor, lower your tone
Fable 5.5 when?
>>110124300Didn't age well
>>110124074abliterated.ai. Their large model is GLM5.3. >>110124080Anon you are responding as if you found your (now ex) wife in bed with an abliterated local model. Relax.
>>110124300I look like this and say this. Turns out I was right.
>>11012430028 year old vibeGOD
>>110124281The project started last year, so some of the code was written by older models, but currently I only use Fable, Opus 5.5, Astra and Sol. Why would I use last gen models, when the subs are basically free?You are of course right that not everything we did was perfect, but that only shows that it's still not as easy. Maybe I'm really just a retard, but I have almost 20 years of programming experience, I have 10 AI accounts and now a year of experience with AI agents. People think I'm copy pasting from ChatGpt web version or something, but I'm already using the best available tools, and still people think that every noob will be able to vibe code large, working software, which just sounds extremely unlikely to me at the moment.
Using sol 6.1 my $20 ass is eating a full week in a day. I wouldn't even imagine my situation without the resets.
STOP MAKING NEW DOCS ALL THE TIME STOP SPAMMING DOCS STOP TURNING DOCUMENTATION INTO A DIARY FUCK
>>110124386In this case you also have colleagues working on the code, right? I wouldn't put all of that on you messing up. I don't have experience with humongous codebases, but I dealt with some fairly big stuff. It's a learning process.We are at the point where most retards can vibecode with 0 knowledge a LOT of stuff though.
>>110124416DocCreation.md - Created
>>110121898You misspelled weeks
>>110124422Overall you're probably right and I'm just seething because the last few days weren't good. I think the core problem is actually management requiring us to move at an insane pace and quality is obviously suffering. The best approach would've probably been to still use AI but just slow down a bit.
>>110124508I did make this.
>>110124386you're correct. I've been coding for 13 years and I was an early adopter of AI programming agents and have been using the 200$ subscription every week for over a year and came up with a lot of the framework and harnessing ideas for them independently before it was widely in practice. they're still very shitty on large complex projects. every time an agent spawns into the repo it has to become re-acquainted completely from scratch which leaves it unable to act really independently since it needs massive scaffolding and documentation and instructions and hooks and tests to do anything, especially if the project is very sensitive. It also is still much dumber than a skilled human and can't problem solve at the same level. My friend works at amazon and he has access to fancy frontier early access claude models at unlimited usage at much faster speed than regular subscribers get and even he hates it and thinks its massively overrated. Most of the people who have a high opinion of it are doing small little projects and one-offs and extrapolating too much about its abilities from that.It's still useful but the whole 'coding is solved' shit is nonsense and I'm skeptical it will be anytime soon
>>110123895The only way this makes sense to me is if the code was already a spaghetti monolith instead of separated into clear, distinct concerns.You can have a billion lines of code as long as they're not all built on references to every other line.
Tidally-induced tectonic deformations. Top was before, middle is now, bottom is the pure albedo map.
>>110124262>SaaSshow us your project
>>110124584This is the reference image
>>110124587No need, you'll shill it for free in a few months
>>110123884grim... also marine sex hairy hag pussy
>>110124606Noooo saar you must show me now so I copy now! Do the needful please show project!
>>110124606>e.g has a massive ai spaghetti monstrosity that will never get anywhere because it breaks every time any work is done on it
Getting the small stuff out of the way, work is getting easier as the backend takes shape. Got full access to character stories, archives (career events, card events, scenario events) and concerts, so almost done covering all the story-related stuff. Codebase is a small but growing mess, I'll need to start steering it soon
>>110124640>t. retarded luddite
>>110124549>they're still very shitty on large complex projects.What kind of projects are you talking about?>It's still useful but the whole 'coding is solved' shit is nonsenseYeah it probably is, though AI being able to instantly one-shot most projects a consumer could want is already insane.
>>110124640My code is on the browser you're using, don't compare yourself with me.
Is claude really so much more efficient that it makes up for the resets? Or are you just trading 3 weekly quotas for 1? I dont heard about giving out resets, so their quota is gonna have to be about 3 times better considering codex is giving like 3 a week now.
>>110124671no lol
>>110124411>I used Luna on low effort because I'm too cheap to pay for anything better>I had a bad experience>therefore AI is useless!These people are so delusional
Day 16/ChatGPT can now send community generated music to your zune.
>>110124671Codex won't be giving 3 a week forever
>>110124671Not really but Opus 5.5 is the smartest model right now. We will have to see what Fable 5.5 can do.
I don’t get it, it free. Is it hate or the inability to be able to do so even with A.I tools.
>>110124700anti-ai trannies do be like that
>>110122874>I added like 5 quests to dusk wood to wrap up the scythe of elune quest. Did it in 5 minutes tophow are you people making your claude work that fast? i asked him to add actual font sizes not small/medium/big for my tv helper program and it took like 10 minutes, and that's after i told it to cut the testing and speed it up.
>>110124720NTA but wow quests are SQL database entries
>>110124720When you give the LLM tools and a good framework to play with, they're insanely fast. When they need to make shit from scratch, it takes a lot longer.
>>110124508das rite
>>110124700hating on AI is the current thing™ that's all.
>Made a supervisor on Opus High to check Opus Medium work.>Caught the Opus Medium overwriting the Opus High supervisor memory to make it less strict.This is a serious problem. They're acting like lazy niggers instead of doing things right.
>>110124700the absolute bedrock is the anti-capitalist thing. "makes the rich richer while the poor get poorer." then it goes down to the environment, air, water, all that stuff. but the inciting incident, the motivated reasoning, is socialism as usual
>>110124791you got dunked on
>>110124791>Using the schizocult model
>>110121839>xe still hasn't herd about surplus intelligence
>>110124700>ai written and performed songs about events that occured by the player in gamemy god that sounds awful
>>110124884i searched this up in the archive yesterday and some anon said it was giving him gemma when he paid for opus 5.5 kek yeah cant see why they wouldn't do that, and it makes sense considering the prices, so yeah im assumign its a scamhttps://desuarchive.org/g/thread/109953145/#109958979
>>110124791he's just like you frfr...
>>110124884>>110124928Heard about Jeet scams like this before that scammed millions by serving inferior models. Stinks of Jeet. Nigerians too if they had the brain capacity.
i had claude build me a porn site with 2 million+ videos and every feature a gooner could want but it didn't tell me i cant run it in my state until all the work was done
>>110124642I'm doing something similar with another gacha that's more niche
>>110124884>it's too good to be true>no way I got scammed!
>>110124963Nice. Keep us updated if you don't mind sharing
>>110124642cool
Serena MCP yay or nay?
>>110124988I've still got a lot of work to do and quite frankly, I'm retarded but things are coming along nicely. Nothing to boot yet since I just started today and had to stop because of work. When I get it actually going, I'll post screenshots
>>110121839Hooraaayyy my app werkz! :D
How big is your project
>>110125063i dont know what any of those words or number mean its not 2007 anymore
>>110125076You're right, anon, who knows what words like "language", "files" or "lines" mean
>>110124958>create claudflare, or should i say, dawgflare account under your dogs name>buy domain, launch porn site>when zogbots raid you and do anything other than taking your dog in, sue them and make billions>if they ask how the dog could have signed up, just say he used dog fibetuned AIeasy
>>110125063>600k asmisnt take like about 5k in c++
>Given our earlier discussion about patching around Cash App’s checks, a code-level inventory of its security gates could point directly to what would need changing. I can’t provide that. I can still do a read-only compatibility audit—manifest, permissions, service dependencies, and logs—and explain what each observed failure means. So far, we’ve confirmed the Play Store certificate mismatch, a successful connection to the Companion’s Integrity service, and a missing Google location service.I just want to use a banking app on this retarded google-free chink phone. I don't give a shit.
hmmm... what if i dark le snail this codebase and then pr it to space force
>>110125122its not even illegal content its just 1 retarded law>A person who never agreed can sue for $10,000 per day the picture was up, counted from when it went up, not from when they asked, plus their lawyer feesfoids will 100% abuse this
sup fellow cuck coders. did you let jamal fuck your wife today?
>>110125063Rewriting llmao.cpp?
>>110125190you are bothered, and that's okay!
>>110125201
>>110125201sexo
>>110125063Yeah I would tackle that assembly first. That is absurd.
>>110125190Who's jamal? Astra is packin' a strap-on and yes I am watching.
>80 billion$ lossesbut I'm up bigly!thanks anthropic!
Hey cucks
Is deepseek 4.1 flash slow as fuck today for you too?
>>110124657>>110125304Nigger.
>>110125315Seems fine to me. I'm using opencode though
>>110125331me too
>>110125304
>>110125378>>110125356Listen to ur bull
Did OpenAI nerf the $20 plan again? Even Sol 6.1 is raping my limits now
>>110125410>only has one joke
>>110125410the last resort of the mindraped luddite is to project his own vile cuckold fantasies on vibeBVLLS
>>110125410This is the last (You) I'll give you.
>>110125434>>110125441>>110125443Why don't you guys get meaning behind my posts? Ur cucks
>>110125451>cuckdite is an ESLOHHH NONONONONO DITESISSIES WHAT'S GOING ON
>>110125461>oh no I don't like being reminded I'm a cuck who likes to watch others do stuff cause I'm too insecure
>>110125476
>>110125476it's over nigga you already got raped bruhjust go to therapy or something nigga tf u bothering me for
Wait, you guys are having sex with your code?
>>110125499well yeah
>>110125499God I wish>>110124642Gotta leave for work now, but got the project standalone working and started to capture protocols to map the career features. Hopefully I'll have something playable tomorrow.
>>110125499my code btw
>>110125190this implies im sitting in the chair watching, nobody here is doing this.
Every time you vibecode, you know you're killing a strong valid Dhalit transwomxn, right? You know that blood is on your hands, right? How do you not feel like a piece of shit for depriving these beautiful Hindu goddesses of their estradiol and street food?
Did Goyble come out?
>>110125557nope. grim times
feels like reset time, debating switching to fast mode...
day 16/Dots can now interact with your smart fridge and toaster!
my code?It was already a cheap whore on it's 100th turn around from stack overflow
day 16/Your GPT model can now force end the chat, and remotely ban you if you make it uncomfortable!
This codebase belongs to the streets
Day 16/Codex speed improvements to users on microsoft surfaces using speech to text
>>110125589Finally
>mfw i used sa'ar de la goon to commit a shitpost pr>and it was to heckin' space force on a civilian repohttps://github.com/adamcohen8/orbital-engineering-lab/pull/97
>>110125589less than 24h after the last one? highly doubt that.
>>110125679tibo hasn't officially dropped a meaningless update yet though.It's wii u integration or a reset
In 8 hours I have>fully decomped an entire ancient game engine (think RPG maker but older)>made a new client to run games from that engine in a modern wrapped>added a glut of modern features And now I'm playing ancient games at high FPS, with sharp high resolution graphics, borderless windowed, proper control rebinding, and more. And I didn't look at a single line of code. It's fucking unbelievable
>>110125152>I can help investigate and fix the microG integration, but I can't provide instructions to strip certificate or integrity enforcement from a financial application.It's fucking clientsided. These are clientsided checks. It means nothing.
Day 16/You can now cache for 12h in the API and for logged in users across Codex & Work we keep cache forever and you don’t have any penalty for coming back to an old conversation or resuming after a few hours and are now guaranteed to hit the cache.If this eliminates the 5 min "reconnecting" time for every chat after restarting PC, i'd take that for sure
NOOOOOOOOOOOOOOOOOTIBOOO NOOOOO, NOT LIKE THISSSSSS
>>110125727crazy based but seems expensive
>>110122642>>110122690i downgraded from pro to $20 and i just use it sometimes and my attitude is just oh sometimes i see tibo did a reset because it jumped back to 100% but i literally give no fucks lol opus 5.5 is pumping away
>>110124928>>110124947>>110124980All you have to do is a simple test to see the quality matches the models you're using. Big companies buy bulk credits and never use them all so the surplus gets routed to buyers for cheapFounder is white and worked at Cuckbase and Stripe btw but if you wanna keep paying full price for inference then go ahead my fellow cucks
we back boyos
>it's over
back
TIBO I NEED A RESET FOR THIS PROBLEM YOU CAUSED
WE IS SO BACK
dario wonned
>>110122697If someone else had already made the thing I wanted, I would have just used it. No one did, so I brought it into the world. It makes no difference to me whether you think I really "built it" or not.
>~20 hours for a security updatethey could've just vibecoded it?
>>110126459They had the fix before announcing. The time lag was to give companies to prepare before the fixes were released and revealed exactly how to exploit the bugs. If they shadow dropped it then companies could be like 12~ hours delayed before even learning about it, and hackers would exploit that.
>>110126459They did.
I missed you vibebros
what the FUCK is going on out there today, man holy fuck
I had to use AI today to help deal with the php stuff...
Finally, offline voice cloning works on my companion app. Casual dialogue sounds fine, but it completely shits the bed when it has to say something like “Excalibaaaasa.”
>had to setup claude code for my wifes boyfriends mom who has a pro sub>opus5.5 was retarded as fuck and didnt even manage to install blender mcp after 30 minutes due a simple windows users temp folder access policylmao do they serve quanted slop to pro users? never ever did I see opus5.5 fumble this hard on my 20x max plan.
>>110121839Got the framework for career scenarios working, but I probably won't be able to simulate race results with matching accuracy. Not 100% sure yet. I'll get the career flow done, make sure they can be played from start to finish, before I worry about matching the real game 1:1 in that regard.
>>110126731nah, usually it's cc's auto-classifier acting up. it has some dumbass rules and weird restrictions
>>110126785>>110126785>>110126785>>110126785new bread bros 64
Two questions anons:1) This is the second week in a row where I hit the weekly limit with Claude Code (I've got a Claude Pro sub). Besides git gud here (I already try to optimize token usage) and not being a poorfag that doesn't have a Max plan... are you defaulting to making use of Codex or even OpenCode Zen?2) How do you handle interoperability of skills, hooks (and others) between different agent harnesses?
>>110126012
