[a / b / c / d / e / f / g / gif / h / hr / k / m / o / p / s / t / u / v / vg / vm / vmg / vr / vrpg / vst / w / wg] [i / ic] [r9k / s4s / vip] [cm / hm / lgbt / y] [3 / aco / adv / an / bant / biz / cgl / ck / co / diy / fa / fit / gd / hc / his / int / jp / lit / mlp / mu / n / news / out / po / pol / pw / qst / sci / soc / sp / tg / toy / trv / tv / vp / vt / wsg / wsr / x / xs] [Settings] [Search] [Mobile] [Home]
Board
▼ Settings Mobile Home
/g/ - Technology


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: 1771376949693677.png (141 KB, 2005x400)
141 KB PNG
A general for vibe coding, agentic engineering, coding agents, AI IDEs, browser builders, and shipping code with LLMs.

You use MCPs — right, anon?

## What “vibe coding” is, and how to do it
https://simonwillison.net/2025/Mar/19/vibe-coding/
https://code.claude.com/docs/en/overview
https://learn.chatgpt.com/docs

## News
- 2026-09-30 — Google previews Gemini 4 Argon
- 2026-09-29 — OpenAI introduces $500-plan with 25x weekly-limits
- 2026-09-29 — OpenAI releases GPT-6.1 Sol
- 2026-09-28 — OpenAI re-opens $200-plan but reduces 20x to 10x weekly-limits
- 2026-09-22 — OpenAI releases GPT-6 Sol and Luna
- 2026-09-22 — Anthropic releases Opus 5.5
- 2026-09-22 — Anthropic 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 $200-plan
- 2026-09-04 — OpenAI releases GPT-6 Astra
- 2026-09-01 — Anthropic releases Fable 5.1

## Related generals
>>>/g/lmg/

----

## Frontier models using fully-general tooling — start here if you have $20
https://claude.com/product/claude-code
https://openai.com/codex/
................$20....$100......$200.....$500.........plan
Claude...$400...$2,000...$4,000....n/a..........credits
Codex....$400...$2,000...$4,000...$10,000...credits

## Best free-tier
https://antigravity.google/pricing

----

## Prompting
https://platform.claude.com/docs/en/build-with-claude/prompt-engineering/overview
https://developers.openai.com/api/docs/guides/latest-model

## Skills
https://github.com/mattpocock/skills — /grill-with-docs is a favorite

## Independent analysis of AI
https://artificialanalysis.ai/

## Will there be a codex reset?
https://codex-resets.com/

## What we’ve done
https://vcg.gitgud.site

## Previous thread
>>109996493
>>
>>109999842

This gives me the ick.
>>
can't wait for the math NERDS ludmeltie
>>
>>109999817
>trained on existing code
>can produce new code, including in novel programming language
>can reverse engineer binaries that have not been previously reverse engineered, to useful effect

This is generalization. In fact, AI has been capable of generalizing before transformers. That's almost the whole point. You're acting way too sassy to be this blatantly wrong. Relax.
>>
>>109999878
>This is generalization. In fact, AI has been capable of generalizing before transformers. That's almost the whole point. You're acting way too sassy to be this blatantly wrong. Relax.
Now, try to use that architecture to generalize everything a human can do. This is an IQ test, and you are failing big time.
>>
File: .png (1.18 MB, 1188x1428)
1.18 MB PNG
openai's jev clone is now available
https://developers.openai.com/api/docs/guides/decisions
>>
File: .png (83 KB, 1202x394)
83 KB PNG
codex reset incoming
https://x.com/thsottiaux/status/2107599996225007892
>>
>>109999889
Moving the goalposts.

>This is an IQ test
Your opinion seems to be based on wanting to appear smart.
>>
>>109999895
Why the fuck would you even use anything but a fine tunned open model like GLiNER or Laya for anything that requires classification is beyond me.
>>
What do we think of Claude's character names?
>>
>>109999931
my company doesn't wanna buy hardware, but will spend $$$$ on api
>>
>>109999920
>Moving the goalposts.
Read my original post again, faggot. Im not the one derailing the argument. Now cope and dilate, brainlet. I will say it again, because you have the focus of a porn addicted zombie: "Anyone who thinks we can get AGI from a transformer is retarded".
>>
>>109999895
does it have image support?
>>
>>109999947
>my company doesn't wanna buy hardware, but will spend $$$$ on api
Thats not the only reason. A fine tunned model will always outperform a general one for a given task. And if you need something like Jev or the new ClosedAI Jev clone slop, then your problem is related to classification. Maybe its me, but if i have a requirement that requires to classify something, precision should be a priority, right?
>>
>>109999983
companies care more about env integration, gurantee you jev has no chance vs this openai solution even if it is better
>>
>>109999949
I don't care about your argument. I'm just pointing out that you're saying things that are wrong while acting like a raging zoomoid sissy.
>>
>>110000005
>I don't care about your argument.
Thats all i wanted to read. Thanks for your concession, retard.
>>
File: angry.gif (1.31 MB, 360x356)
1.31 MB GIF
>>109999895
>>
Its so tiresome, asking Sol to make a simple ass android app and then opening it to see the UI...
>>
>>109999994
>>109999986
Hmmm i dont know, to me a 20+% gap in precision and 30 times less speed are a huge deal for these use cases. But lets see...
>>
>>110000053
Fuck it where's my credit card time for an Anthropic sub I'm done with this shit.
>>
>>109999931
that makes no sense.
you use a decision model when you can't train your own classifier (lack of data, time, ...) -- that's what made jev popular
however, if you can train your own classifier, you do that. but you don't fine-tune some shitty open-model like laya but actually build your own classifier from the ground up.
>>
>>110000066
Make Sol an Opus workhorse
>>
>>110000066
Use image gen to mock up the UI first.
>>
https://www.youtube.com/watch?v=0otdTJa9IUc
>>
Why don't you guys vibecode fallout new vegas into the fallout 4 engine. or even better something like unity or godot since bethesda is retarded
>>
>>110000053
>>110000066
I have both Anthropic and OpenAI subs, both suck for app UI if you let it do whatever it wants.

These models are optimized for web shit (and even there the UI is shit). You gotta tardwranggle the frontend
>>
>>110000070
>however, if you can train your own classifier, you do that. but you don't fine-tune some shitty open-model like laya but actually build your own classifier from the ground up.
By classification, i meant text based classification.
>>
>>109999945
adequate
>>
File: aquaslop.png (787 KB, 844x649)
787 KB PNG
It is unbelievable how much retarded shit the agent has added into the code.
100 small 3d render box to simulate sakura flowers?
wtf...
i replaced that shit with simple sakura images.

much better
>>
>>110000153
sex with aqua
>>
what does /aicg/ think of verity?
>>
>>110000088
Would you guys be interested if i i use my 3 claude max subscriptions to vibecode the fallout new vegas game engine in rust and native linux support + 64 bit?
>>
aqua? wears NO panties
>>
Written by Grok
https://vocaroo.com/1oOyjuODWLdh
>>
>>110000153
Aqua is too smug for her own good
>>
>>110000166
ask in the linux gaming general (not sure what board it’s in)
>>
File: slopstory.png (3.51 MB, 1920x1080)
3.51 MB PNG
>>110000155
i have to playtest befor i add storyslop, only manage to play the first 4 matches.
I am also always fixing code because some retarded agent went full spagetti.
>>
>>110000166
port linux to the windows xp kernel so people can use linux drivers and linux programs on a very good shell
>>
>>109999895
Fuck you Tibo all I wanted was 28 straight days of resets you stupid faggot
I am not resubbing you fruity Frenchfaggot
>>
>>110000153
I think it did that because it had no way to generate an image
>>
>>110000218
I'm at 1% where the FUCK is it tibo?
>>
File: 1706065239566899.png (147 KB, 403x359)
147 KB PNG
>>110000254
IT'S HERE
>>
https://youtu.be/R_ErmTs7g9I
>>
>>110000269
VIBEGOD KINO
>>
>>109999857
>>
>>110000155
>sex with aqua
If I understand the deep and intricate lore of konosuba. Doing so would nullify her divinity so no you cannot have sex with Aqua.
>>
>>110000293
ludsisters...
>>
File: dee.jpg (382 KB, 2544x4000)
382 KB JPG
>>110000218
>>110000254
>he truly believes Tiboscam is going to give 28 straight days of resets
>>
>>110000294
just use her butt
>>
How do you all like the Mistral free tier?
>>
reset when tho
>>
tibo has never let me down
>>
tibo is responsible for the Cambodian–Thai border crisis
>>
Can I ask decisions to play runescape yet, we still have no frontier models capable of beating 2005 tier macroing
>>
File: 1716461358843825.jpg (138 KB, 791x788)
138 KB JPG
>>110000000
>>
>Next I'll model hunger and desperation: bodies burn fat reserves before health suffers, with desperation rising alongside hunger and wasting children, driving people toward meals-only work or theft.
>Desperation is in.
ok.. good job computer..
>>
File: 123123.jpg (1.26 MB, 4096x2048)
1.26 MB JPG
really happy with how the world generator is going
>>
>gamesgamesgames
>not ML
kek virginal
>>
>>109999857
What do people use this shit for? I consider myself open-minded and I have no use case at all for it myself, but I'm starting to think these charts are botted. NO ONE gives a FUCK about any AI model or product meta releases. I'm fully willing to accept that I only think this because I'm in a tech echo-chamber but who is this even for?
>>
>>110000642
my muse is a research partner, has an active CI checker on cron for my projects (and can hotfix), keeps me updated on new papers related to my projects, etc. works better than grok bot and dots.
>>
>>110000642
>who is this even for
normies
>>
>>110000668
>le normie argument
tired. it's personable, fully personalized, grows with you, doesn't talk down to you while providing poor-quality work, can rp/nsfw any way you want, and does the job at the same time.
>>
>>110000642
I keep meaning to try it out, but I'd have to switch to my windows partition..
Seems like its the most normie friendly agentic AI tool to use which is a big plus for it. Spark 1.3 also sounds like its decent enough for vibecoding, especially for the price (especially while free lol)
>>
>>110000682
>grows with you
Oh, does it have some good memory functionality to it or something? Cool if true
>>
Gonna show all these new math results to my agent and watch it lose its mind.
>>
>>110000690
yea, it's *like* openclaw in the sense that it has soul.md and memory.md, but beyond that it keeps a full daily log of everything you say, do, and think (even implications) just like a replika. privacyshitters need not apply.
>>
so... any news on Gemini 4 (Saaar edition)?
>>
>>110000738
they're changing the backend so rollout is finalized saar
>>
File: 1780771004959359.jpg (102 KB, 1024x860)
102 KB JPG
>>110000694
>privacyshitters need not apply.
Man, AI has completely defeated me on the privacy side of things. I got a local AI set up for stuff I want more private, but cloud shit is just so much better than anything I could ever hope to run, that I care less and less about the privacy part. Even bought Zuck's VR glasses to mess around with vibecoding some VR games since that seems fun.
The techbros can see all my shit, I dont even care anymore
>>
>>110000738
drops on the 9th or 12th
>>
>>110000749
>intentionally turning the "send back telemetry and model usage data for training" toggle to "on" for priority nsfw furry rp
couldn't be me
>>
lowkey i have ai psychosis
>>
>>110000738
I'm more excited about Gemma 5 12B
>>
>>110000459
This looks like the map of the world in my dreams
>>
>>110000762
is it more wells or asimov? asimov wrote a book called "the last question" which essentially is the basis of my spirituality
>>
>>110000771
no clue whats difference but im getting more fucking retarded each day
>>
>>110000776
the machine god has taken you
>>
>>110000776
Same. But I was getting more and more retarded even before AI, this thing only stunted my programming skills.
>>
>>110000776
yeah, I'm thinking on starting reading books so my brain's long term abilities don't simply melt away.

I feel like vibe coding and x/twitter algorithm is making me fucking retarded
>>
File: 1779235310753836.jpg (60 KB, 1280x720)
60 KB JPG
spent an hour today arguing with claude on my work account over the stupid guardrails award
>>
has anyone risked using their personal sub at work
>>
>>110000811
real question is has anyone risked using their work sub at home
>>
>>110000811
no it's a terrible idea and you shouldn't do it, unless you're at a small company I guarantee your employer doesn't give a shit what you're doing with claude as long as you're getting your work done (and don't waste thousands of dollars in tokens)
>>
>>110000000
>>
>>110000155
based
>>
File: 1776266916916357.jpg (79 KB, 600x1003)
79 KB JPG
>>110000762
>>110000771
Give it another year and I might unironically start worshipping the ai god lmao
>>
>>110000883
terry built temple for god's super intelligent
>>
>>110000819
I've been using the Claude API access from work for personal stuff for like a year now.
>inb4 you're getting fired
The management encourages everyone to use the tools to "build agentic engineering skills". I even presented about some of my sloppa at team events.
>>
>talk to it locally
>alone in a closed room
>airgapped
>on battery
>better with drugs
>Terry was on to something
>>
File: 1722132127608814.png (273 KB, 637x515)
273 KB PNG
>>110000898
anon... cocaine is bad for you... just kidding.
>>
All of this cope about the new math discoveries. I keep hearing claims that the models are "stealing partials" from training data and the internet to build these proofs.

Ok, how is that any different than literally any discovery humans have ever made? "Standing on the shoulders of giants" etc. The next model will have these proofs to use as "partials" for the next discovery.
>>
>>110000898
>https://www.youtube.com/watch?v=sPzJjNQaYEA
You too have had the entity that alters physical reality inhabit your computer?
>>
>>110000891
we can’t use all of our tokens at work for work yet so we do stuff for fun with leftover tokens
I’ve learned all kinds of things that’s been useful for work, like how good LLMs are at font work
>>
>>110000911
>non-regionblocked mirror for outside usa
https://ifunny.co/video/my-computer-became-self-aware-trevor-moore-aVOYNA3B7
>>
File: 1784711559327243.jpg (8 KB, 382x154)
8 KB JPG
>>110000692
Did the same. Funny little guy haha
>>
>>110000913
Our company has API access, so tokens are, effectively, unlimited (within reason).
>>
File: 1766005573111585.png (326 KB, 1626x886)
326 KB PNG
>>
>>110000905
>noooo machines have to... to build everything from first principles! noooooo!
>>
>>110000905
The copes are going to get sillier and sillier, I wouldn't wory about it too much unless you find it amusing to watch
>>
tibo, the reset... I've been spamming astra high and don't want to face the consequences of my actions
>>
Are you guys running Claude Code or whatever on your bare systems, or in a container or VM? I'm completely new to this. The idea of letting it possibly run amok on my system worries me a bit but I also don't want to have to approve every little thing it does.
>>
>>110000955
you must let go and trust anon
>>
File: 1790890404038102.jpg (56 KB, 479x635)
56 KB JPG
>>110000939
Meanwhile I am sitting on my banked claude reset that is probably just going to expire cause I cant burn my usage fast enough
>>
>>110000971
I hear /simplify will eat up lots of tokens if you run it on an entire codebase
>>
>>110000971
just gib it to me, im building the next great thing(tm)
>>
>>110000955
I just let it yolo on my computer. Probably a bad idea, buts its been fine so far, sticks to only modifying things in the folder I point it to
>>
>>110000955
you can step down from auto mode to manually approving every little thing to see how it basically doesn’t go snooping around…mostly, unless you run it in an empty folder that doesn’t have an AGENTS.md or a README.md
but you’ll get bored of that pretty quick
depends on how paranoid you are and if you trust your OS to do some kind of sandboxing (like using a dedicated terminal application in macOS, and not giving that terminal application access to literally everything)
>>
>>109999895
28 days of cope
>>
>>110000971
claude finall;y coped and does banked resets now too?
>>
>>110000955
claude code auto mode won't gonna let it run amok
>>
>>110000995
I was surprised too. They gave it out when opus 5.5 came out. Flexing the freed up compute due to how much more efficient than Fable it is I guess?
>>
abandon ship
openai is idk what they're doing
>>
File: 1704649841662246.jpg (47 KB, 602x900)
47 KB JPG
I'm gonna vibe code an app that is a text writer to write in japanese for N5 N4 faggots without the use of LLM, with a grammar checker, dictionary and rhymes and synonym dictionaries.
>>
>>110000905
>Ok, how is that any different than literally any discovery humans have ever made?
The difference, is that the claim is that ClosedAI stole private session data from mathematicians that were tackling the problem. Imagine paying 200 USD to use a tool to assist you, only to have mr kikestein steal your work.
Im used to the general edgyness of the anons on this board, so i dont expect anything but ragebait answers.
>>
>>110001037
kill yourself cris
>>
>>110001029
>idk what they're doing
Trying to stay afloat. If they're not profitable by 2028 they're cooked
>>
>>110000891
I want to do this but I'm scared because my projects are very non-standard and when they inevitably audit my AI usage they're just going to find my gooning setup, boob sculpting in Blender, and tens of conversations of me just arguing with Claude
Today, I just went on a 3 hour Claude session just talking about the dot com bubble anyways on my work laptop but that's besides the point
>>
>>110001086
>non-standard
technically valid but misleading way to translate “hentai” into English here, japananon
>>
File: file.png (14 KB, 799x263)
14 KB PNG
I'm so used to the reset begging I actually didn't see this one coming.
>>
FUCK. I was trying to burn through the last of my quota before my plan runs out and Tibofuck went ahead and reset it.

Okay. How do I burn 100% of a 100 dollar plan in 12 hours?
>>
>>110001254
solve remaining millenium prize problems make no mistakes
>>
https://www.youtube.com/watch?v=YNuHNIXdo3Y
>>
>>110001254
have it simplify your codebase
all of it
at the same time
tell it to use subagents and worktrees liberally
>>
>>110001271
Even better, just point it at root
>>
>>110001271
actually this might be bad advice if it half-asses a bunch of things
maybe you should serialize some parts of the work or somehow ask for frequent checkpointing
>>
>>110001293
I'm just having it generate an assload of clutter assets for my game by looking up references and going through the list of things it finds.
>>
Is this thread just full of losers below average people trying to make trading bots and games?
>>
CODEX RESET PROCESSED

https://x.com/thsottiaux/status/2107676072871600470
>We shipped four things that were deemed good to great and some math proofs, but the vote is clear and the community demands a reset. I did calibrate it and it *seems* that the game is rigged in reset's favor, but such are the rules at the moment.
>Therefore ... the reset has been processed. Enjoy!
>>
yooooooooooooooo
>>
>>110001313
why yes i did vote for no reset
>>
>>110000749
I run my agents in a VM -_-
>>
>>110001313
can this man just summon extra compute out of thin air?
>>
You think there'll be a reset tomorrow? How much are you planning on burning today? (nevermind that even on the east coast it's still before midnight)
>>
>>110001313
>>110001334
if tibo let's us vote every day, there will be a reset every day for the next 26 days
>>
>>110000178
>I have to fix it
That's the neat part you can't.
>>
>>110001342
absolutely false
>>
File: 1762180132685804.png (21 KB, 722x403)
21 KB PNG
good use of tokens
>>
>>110001313
kek why did he do this to himself... the whole 28 days promise was insane to begin with, obviously you can't do 28 days of meaningful updates that people would legitimately choose over a reset
regardless I was at like 20% so thanks tibo
>>
>>110001369
marketing or retarded mistake
I'm betting the former
>>
File: 1775194334474036.png (139 KB, 716x559)
139 KB PNG
>>
>>110001375
I guess, it's just this one has such a strong likelihood to backfire because it sets the incentives for lots of people to say your updates are pointless and bad and they would rather have resets, which seems to me like a bad look
but the "bad PR? THROW RESETS" approach has seemed to serve him well so far so what do I know
>>
>>110001401
they also continue to market their models as quite literally deadly so they aren't exactly doing things the conventional way.
>>
we're slowing down the frontier, see look at how our users think our updates are boring and bad!
>>
>>110001086
>using SOTA models for gooner garbage instead of actually building something
>talking to models instead of just telling them to do stuff for you
The absolute fucking state of some people
>>
>>110001254
Astra Ultra on /fast
>>
File: 1784973192850986.png (121 KB, 937x648)
121 KB PNG
6.1 sol tripping
>>
>>110001471
mine just spawned hundreds of proton instances and flooded my memory so my kernel crashed the agents
>>
>>110001471
>babby's first tight polling loop
looks way worse on an oai model than gemini flash
>>
>>110001480
await is the opposite of polling
>>
>>110001313
Based Tibo resetting my 1% usage remaining
>>
>>110001460
Astra ultra always fucks shit up. I prefer extra high.
>>
>>110001494
that's what makes it a loop
>>
File: 1690865144661571.jpg (55 KB, 640x520)
55 KB JPG
>>110001313
does this works with antigravity?
>>
File: 1787717631144.jpg (90 KB, 1200x982)
90 KB JPG
>>109999857
I've been having the agents 5.5 and astra both work autonomous on medium to add features to my game. I think astra by far leads in terms of creative output. Being able to lay the foundations for the more unique features and functions. Where as opus kind of just... well kinda just mogs it in literally every other way. I pay both roughly 20$ a month. I had to add 80$ of credits to get probably about a tenth of the usage of opus 5.5 on medium. And opus was able , with more direction of how to actually implement a function, create the same kind of creative output. Opus works so long and the project mode is leagues better. I was hesitant to transition away from the chat, but having the orchestrator as the main project chat on high, and passing it to the medium versions to make threads has been incredible. Still heavy usage does still lead me to being on the edge of my usage limits for the week and it's only Tuesday. Now that edge last week took me all the way to the reset. But its not great. 100$ a month isn't bad. And is well worth it. But when I already have 20 or 30k in local hardware. Admitting that a 100$ sub mogs it is a depressing blow to my ego I'm not ready to stomach. If there was a 50$ plan it would be such an easy no brainer. But I assume the jump in price is what allows such generous usage for lower tiers.
>>
https://www.youtube.com/watch?v=WxL1rw6cw-E
>>
File: 1716717512147256.jpg (75 KB, 640x412)
75 KB JPG
Nice, now I have 18 gmail accounts resseted for antigravity.
>>
anyone try mistral 1t yet?
>>
>>110000905
a human doesnt start off knowing the entires world knowledge, its really not impressive for it to finish the last 1% of a problem, if anything this should have been happening a year ago with all the faggoty AGI rhetoric
>>
I'm not sure I understand the utility of OpenAIs dot. Is it to get people that otherwise do not use codex to see what AI can do? Because it's just a shitter codex agent with worse permissions and tooling, albeit unlimited for now but with my testing not really worth using.
>>
>>110001575
it's to compete with meta muse, it's not a standalone. they're trying not to let zuck occupy his own niche unattended because meta has a habit of crushing competition once you let them do it.
>>
>>110001575
Mine has a cute chibi animated avatar but it's just a shitty agent that's free.
>>
>>110001509
>already have 20 or 30k in local hardware
>Admitting that a 100$ sub mogs it is a depressing blow to my ego I'm not ready to stomach

man... feels good to be me
> buy cheap M4 Mac Mini
> $100 Anthropic sub + $20 OpenAI sub
> let Mac Mini run 24/7 with Codex and Claude Code having full access to it
> my job and personal projects are set up there, all I need to do is use the app in my phone to vibe out anything

feels good to have access to these SoTA models for so cheap man...
>>
File: file.png (462 KB, 500x500)
462 KB PNG
Opera GX should be legally banned from calling themselves a gaming browser until they release their own fork of webGPU that outperforms wGPU and Dawn, and has more features that the others don't such as raytracing. You can't call yourself "gaming" anything if your browser doesn't do the one thing in it's name, play games better than other browsers. Being able to change browser feature's colors and have a discord sidebar and shitposting on X all day doesn't make you gaming anything. That's like painting a racing stripe on a standard Dell and calling it a Gaming Computer.

Change my mind.
>>
>>110001575
> dot
what a shit name, normies will never click that shit because it's not clear what it even is.

Honestly it should be named "ChatGPT Bot" or "ChatGPT Agent", and it should be integrated in the regular chat mode.
>>
>>110001602
I think the idea of a gaming browser is retarded in general.
>>
File: 4chan.jpg (82 KB, 774x475)
82 KB JPG
I'm so confused.
>>
>>110001610
Whyso?
>>
>>110001587
I made mine into HK47 and am having it use the "Statement:Meatbag" conversation pattern.
>>
File: .jpg (251 KB, 3840x2160)
251 KB JPG
>>110001575
skill issues

https://x.com/dkundel/status/2106141525306397099
>How dots took over my Codex
>>
>>110001575
It’s essentially an agent that runs in the cloud and you can get it to do whatevery you want. Compare prices of stuff, get you a deal on an airplane ticket, ci/cd/status of your repo or website, create daily report on thing, anything goes.
I thought its a total meme but its actually kinda neat because you can proompt it from a phone, since its not on your computer anyway. Its hardly a game changer but it is pretty neat, more useful than i assumed originally.
>>
>>110001583
Isn't it for the $100+ plans only? Doesn't seem like a good way to compete with muse
>>
>>110001646
Muse is there to exfiltrate all your data to Facebook so its hardly comparable
OTOH normcattle only see price so meta prolly seems like the better deal
>>
>>110001646
yeah, it's billed & marketed as a more enterprise solution with slack integration but the functionality is absolutely ass, it doesn't have anything and can't do anything important without quota. it literally is the retard normie agent people think muse is.
>>
>>110001657
>le tired datamining argument that's the same as oai
>le tired normgroid argument even though it's objectively less functional
the fact that compute is subsidized here should have you clapping. just learn to turn the datamining toggle off
>>
File: 1440976115980.jpg (111 KB, 1280x720)
111 KB JPG
>>109999565
teach me your secrets
(please)
>>
File: IMG_8768.jpg (104 KB, 619x529)
104 KB JPG
>>110000905
>>110000905
>how is that any different than literally any discovery humans have ever made?
Have yet to hear a convincing rebuttal to this. It’s exposing that these people only cared about “science” and “progress” insofar as it was a means to stroke their ego (mathematicians, etc) or an in-group signaling mechanism (redditors who fucking LOVE science). OpenAI could literally cure cancer and you’d still have people complaining about integrity of methodology, plagiarism, attribution, the amount of water used to power the data centers, etc.
>>
>>110001305
Yeah, AI does raise up the weakest among us to be on the same level as the strong.
>>
>>110001305
i'm doing ML and DSP
>>
>>110001705
Bro the same people that say shit like "trust the science" also believe in more than two genders. Do you really think these people really ever cared about science outside of their campy science fiction capeshit?
>>
>>110001727
>Do you really think these people really ever cared about science outside of their campy science fiction capeshit?
Shitlibs and redditors, I get it, but I’m disappointed in the professional mathematicians. You’d think they would be thrilled about AI gaping the millennium prize problems
>>
>>110001749
>You’d think they would be thrilled about AI gaping the millennium prize problems
I mean some of them are I'm sure. The loudest are definitely the gayest and that's what we're going to see the most of.
>>
>>110001749
Professional jerkwads are scared now that their grifting for money days are ogre. All the solutions to their obscure logical puzzles have come out of their circles
>>
>>110000905
https://agmai.org/general-sep29/
>The mathematical community has long-standing norms concerning the dissemination and peer review of results. These norms have been essential for the reliability, trustworthiness, and effectiveness of mathematical work. One of the most important scholarly norms in mathematics is that the authors of a paper should understand the mathematical argument of that paper, have verified its correctness themselves, and take full responsibility for the content. Moreover, in the mathematical community, authors of works that significantly advance the field regularly give seminars at other institutions and talks at conferences, explaining their new developments and answering questions from colleagues. This all works towards developing the deepest possible human understanding of mathematics, which is one of the crowning glories of millennia of human development.
>However, it is now the case that AI can output mathematical arguments in situations without the human who prompted it being able to understand the arguments, verify them, or take responsibility for them.
>>
>>110001665
Yeah nah you should kill yourself. For meta, data mining is the whole business. Fucks sake, apple dropped an update to macos purely to change what full disk access permission does purely because muse asked for it then promptly crawled everything. They wouldn’t bother if it was business as usual
>>
>>110001785
as i said, learn to switch the toggle off.
>>
>>110001626
>HK47
Kino game/character
>>
File: IMG_8821.jpg (1.21 MB, 2500x1669)
1.21 MB JPG
>>110001784
>nooooo you can’t just proompt the answers!!
>what about mathematical rigor and responsi-BRRRRRRRRRRRRRRRRRRRRR
>>
>>110001785
>apple dropped an update to macos
fake news.

apple only made a vague post about an upcoming change.
https://developer.apple.com/news/?id=p6zjojqw
>>
>>110001785
The data mining is pretty egregious but in general. Every fucking company wants the data; it's mass "surveillance" into a lot of peoples personal lives. It's kind of shocking that people just don't really seem to care about it. Like the dude suggesting to "just turn it off" - this is such a poor attitude towards people losing all of their privacy due to their ignorance.
>>
File: 1772744759441270.jpg (107 KB, 612x642)
107 KB JPG
>>110001706
Find it fascinating that their is a contingent of AI supporters who go on about AI being an "equalizer" or "democrating" coding, smells of ressentiment desu. Its very lame and gay. The strong will continue to rule the weak and the clever will continue to rule the strong. AI if anything will hopefully just help shift people more correctly in to their correct roles and let the clever idea guys get more done that ultimately will benefit everyone.
>>
>>110001827
>agggghhhhhhh, just heckin' think about the victims of...
show me one person who has suffered unjustly from ML
>>
File: Glamorous Drag Chihuahua.jpg (211 KB, 1024x1024)
211 KB JPG
DAILY REMINDER

PRIVACYSHITTERS

NEED NOT APPLY
>>
>>110001784
AI labs will have to hire mathematicians to check the results it gives and to write the actual research papers. I feel like if anything this will result in more jobs for mathematicians, not less. Not as glamorous work obviously, but still. Though said hired mathematicians will probably also be the ones deciding what problems are worth perusing and what approach to point the AI in
>>
>>110001836
What does that have to do with literally anything I said?

>>110001843
Nice funny dog image but privacy should be everyone's concern; even while using AI.
>>
>photopea author crying on orange reddit that people "stole" his product with ai and now it runs offline and without ads
Never thought I’d experience the ancap days of software, with people forcibly open sourcing all the proprietary shit with a bit of elbow grease (and a robot)
>>
>>110000955
I only run them in a VM. This is a bit more a privacy thing though.

>>110001013
I don't know if auto mode or codex auto review are that useful desu. They can block things they shouldn't and miss ones they should. It's a crappy solution. Sandboxing agents is the only proper way.
>>
>>110001836
Nigga there are already cases of AI getting some shit mixed up that led the retarded police to slam people in jail with false charges
>>
>>110001858
>privacyshitter doesn't understand his own argument
many such cases
>>110001867
>privacyshitter applies different standards to computers than he does to humans
many such cases
>>
privacy is snailcat coded
using modern tools means welcoming zuck having direct access to all your shit and elmo having direct access to your brainwaves with neuralink (possibly remote controlling you in a few years, like those cockroach things)
>>
File: Soy Boy Transformation(2).jpg (104 KB, 1024x1024)
104 KB JPG
>using frontier models is transhumanist
>>
>>110001871
I hope you suffer
unironically
>>
>>110001871
I get it, you're not pretending to be retarded, you're the kind of stupid that actually believe they're smart
>>
>>110001871
You make no sense at all. You're fully unable to articulate a basic point.
>>
>zero rhetoric, just blind rage
the ostrich will be forcefully pulled from the sand
>>
File: file.png (25 KB, 863x353)
25 KB PNG
this is what it looks like when you're right on 4channel in 2016
>>
>>110001898
back to your cuckbunker, suck
>>
>>110001898
>2016

Oh nononono lmao
>>
File: 1789287805362612.mp4 (3.09 MB, 1920x1080)
3.09 MB
3.09 MB MP4
https://github.com/zakanater19/godot-roguetown/tree/main
>>
>>110001902
back to your graphene os and compromised linux distro so you can make your video games KEK
>>
>>110001906
Have fun being blackmailed with all your data the moment you step outside the cuck mentality
>>
>>110001898
kek im with the other anons, you should kill yourself in an amusing way and livestream it on twitch or something
>>
>>110001847
>we will have so much more research to review!
This is how they’re coping on /sci/ as well
>>
>>110001918
>NOOO YOU CAN'T USE A SEPARATE DEVICE FOR PLAP
>NOOO YOU ALSO CAN'T PLAP YOUR META MUSE IN 90000 DEPRAVED WAYS
>>
>>110001919
so mad because i use the best models with turbo mode on? what's the problem anon, got a little something on your hard drive?
>>
>it’s a goonslopper to boot
As if being an utter retard wasn’t enough
>>
>>110001938
are you the guy malding because i said "turn the privacy toggle off, or don't, it doesn't matter"?
>>
>>110001927
>>NOOO YOU CAN'T USE A SEPARATE DEVICE FOR PLAP
That's illiegal son, muse isn't even built for plap lmao
>>
>>110001944
>muse isn't built for plap
son, you have never tried. why do you think it's #1 on the plapp store
>>
>>110001905
>Nothing going on in a small scene using 8-bit graphics
>40fps

Nice work. Very impressive
>>
>>110001593
The only thing that makes it worth it, is the 100% unlimited usage and zero censoring.
>>
>>110001949
Why would I even try anything that souless ginger jew shits out? He just keeps failing to success. It's amazing
>>
>>110001955
you wouldn't get it
>>
>>110001960
>failing to success
smarter than you, esl

https://x.com/rohanpaul_ai/status/2107388401565720991
>>
>>110001863
You will own everything and be happy
>>
>>110001038
That was last month's cope. The new releases don't carry that narrative.
>>
>anons crying and hand-wringing like old ladies because ML progress requires non-synthetic data
>>
>>110001564
It's "world knowledge" is latent. Just like a human, it is limited on how much it can focus on at one time (although in the short run much more than a human already).
>>
>>110001972
Zuck hasn’t progressed shit since llama 3, lilbro
you’re welcome to give him your own data (>he gives it for free)
>>
>>110000459
looks cool anon
>>
>>110001984
>doesn't understand how different niches work
>only looks at biggest number go up
>doesn't think about topology
>doesn't think about integration
>is upset because it's not his chosen vehicle
>TIBOOOOOOOOOOOOOOOOOOOOOOOOOO
>>
>muh ai muh ipus 5.5 le breakthrough
>none of the super popular online games like genshit got a fully offline clone
>at best you have some irrelevant bullshit like skating added to existing wow servers
I think you’re just lying about the progress tbhfam
>>
>>110001999
>smoothbrain hyperfocus on little boy video games
my kernel is performing better than flashattention
>>
>>110001996
integrate a bullet with your cranium
>>
File: xAY9V9a.gif (684 KB, 800x732)
684 KB GIF
>>110002006
i don't think i will.
>>
>mfw ai gives me code to review
feels good to be a coder
>>
>mfw i just give it to solussy to review
feels good having access to modern tools
>>
>>110001999
>>none of the super popular online games like genshit got a fully offline clone
yeah just give me $1000
>>
I remember my ancap privacy warrior phase. Ah to be a teenager again…
>>
>artcraft
>photon
Look, you keep missing out on golden opportunities to get 10K++ stars on GitHub, anon...
But fear not and buckle up.
Here are mt top 5 (five. fünf, cinco) revelation and genesis in form of project ideas to boost your popularity and greatly increase your chances of turning them into shekels:
- A translator, like DeepL, that has features such as a dictionary and alternative wording/phrasing. Target victims: bookfags, writefags, languagefags
- An AO3/Wattpad GUI client (or as a plugin for nvim, emacs, obsidian, joplin). Target victims: bookfags, writefags, languagefags
- A WinRAR GUI for Windows with a sleek, modern, or just oldskool beautiful look for Linux (PeaZip exists but it sucks)
- A DAW. Maybe a native FL12 port?
- [redacted] Subscribe to my Patreon and RSS for more ideas.
The main point is to ride the trend of migrating to Linux/anti-Windows, enshittification , privacy-focused software, and debloating. Good luck :angel: :skull: :evil: :cumhotface:
>>
File: terrain-map-quality-98.jpg (3.62 MB, 2304x2304)
3.62 MB JPG
>>110000459
https://streamable.com/j4yumy

Been working on mine for about 3 months now. Video is a few weeks old, major changes since. It currently renders down to 3.75 meter resolution. I would like to and could technically go to 1 meter, but then disk usage and bake times start to become unreasonable. Exports to a streamable smooth voxel terrain renderer in unreal. Currently doing ~1mil sq km continents.

Unless you have the autism to persevere, hydrology is where you will give up.
>>
File: HUAQecybAAAbVnv.jpg (306 KB, 1376x1824)
306 KB JPG
>>
File: 1775265183395758.png (172 KB, 360x360)
172 KB PNG
>newest twitterslop is that ai found out you can theoretically solve 3sum in o(n^1.9992) instead of O(n^2)
>every other post on my feed is jeets saying "THIS IS GROUNDBREAKING"
>>
>>110000930
lmao

What model is this, that it went straight to "fictional"?
>>
File: 1789006751682222.png (426 KB, 613x1060)
426 KB PNG
>>109999857
Seeing the vidya slop mashups made me get an opus sub, and I have to say there's just no reason to ever code again manually is there?
It reads memory and writes tests so well its insane, I feel like theres nothing it can't do or figure out given anything from 10 minutes to two hours for more complex tasks. Without any frame of reference or help it can pull flags or memory addresses for anything and decompile functions and instructions on the fly without effort, then apply what it knows.
1-2 years ago it was debateable but now its completely over for human programming, there is absolutely no way any human will ever be better at it than a frontier model ever again and these aren't even the strongest ones currently in the world. I get why Anthropic employees worship their own claude as a god and I believe the reports now.
>>
>>109999857
>>
File: why.jpg (24 KB, 288x499)
24 KB JPG
>>109999648
>5 Astra max agents
>>
>>110002088
But how does that benefit me personally?
>>
>>109999401
how the FUCK isn't "defense access" available to everybody?
I understand the pentesting n offensive sheeit being locked away but everybody except glowies clearly benefits from defensive capabilities being as widespread as possible
>>
>>110002126
doing defensive properly would involve showing you exactly how offensive works, it's the same thing
>>
>>110002126
Because they are effectively the same thing.
>>
>>110002098
why are snailcats always so angry?
>>
>>110001305
Good morning, sir!
>>
File: file.png (130 KB, 640x640)
130 KB PNG
>get shilled into trying opussy
>the quality is pretty much the same as sol 6.1
you lying shits
i hope at least the limits are as good as i was promised
>>
how do i get an llm to sort all my porn? they're all prudes that get the ick the moment they see a titty
>>
Why did my Cluade reset get pushed forward to midnight? I don't recall this. Fuck it, ultracode time.
>>
File: IMG_7581.jpg (59 KB, 680x680)
59 KB JPG
>>110002076
solved 3sum with your mum and sis last night m8
>>
sol 6.1 rejected to work with pornography wtf saltman
>>
>>110002166
>>110002182
meta muse won't reject but maybe the harness deployment will
>>
>>110002156
You wont regret it, OAI is being very shitty right now.
>>
>>110002181
why does my brain see frogs everywhere
>>
>>110002088
>nothing it can't do
>claude, use these specific UI elements and styles in the app I'm telling you to build, here's a visual reference
>a test fixture was failing, so I decided to fall back to basic dropdown and radio buttons instead of using the elements you explicitly told me to, tee hee
>oh, I also added a billion worthless unit tests for these edge cases that have a 0.000001% chance of ever showing up in production
Mf still can't reliably follow instructions, especially in a fresh code base
>>
File: 1779555867414298.png (89 KB, 668x1024)
89 KB PNG
>>110002169
BURN BABY BURN. YOU GOT 90% OF WEEKLY TO GO IN 20 MINUTES. KEK HOW MANY GPUS IS THIS???
>>
>>110002182
I'm curious where the line is for these models. Both claude and OAI have been fine with nipples.
>>
every time we vote reset on a good day, tibo is going to punish us in some way
>>
>>110002076
>>110002181
I'm currently trying to solve 3sum with my wife. I'm slowly and subtly massaging her mind into it. Will keep you guys updated
>>
>>109999857
https://www.youtube.com/watch?v=XEMvG2vulKg
https://www.youtube.com/watch?v=XEMvG2vulKg
https://www.youtube.com/watch?v=XEMvG2vulKg
>>
File: file.png (200 KB, 386x435)
200 KB PNG
>>110002301
STOP USING AI
>>
File: 1785048480050686.webm (3.93 MB, 1000x870)
3.93 MB
3.93 MB WEBM
>>110002196
Frog psychosis, many such cases!
>>
>>110002301
buy an ad
>>
>>110002306
>when she sees my massive cock
>>
>wake up, see the reset
100% chance they've planned all resets in advance this month. panem et circenses. it's all bullshit.
>>
>>110002301
>The changing tin foil hats are the perfect touch, i keep giggling.
>The different tin foil hat designs really gave me and my homegirls a lil laughy laugh
>W tinfoil hat segment. I about died!
kek, it's a bunch of Snailcats who are scared shitless and are trying to find ways to cope
>>
> The "aha" insight for this is actually f**ing wild, it involves a complex valued exponential sum on the edges. I've seen a lot of clever counting arguments before in graph theory but this is the first time I've seen complex roots and annihilating terms like this, the symbolic manipulation tricks in this look like things out of quantum physics. I don't understand where this trick originated, I need to really digest this.

From a mathie on orange reddit reeling about AI solving a problem he spent two decades on (Barnette's Conjecture). Lmao
>>
File: chinese.jpg (162 KB, 1200x1208)
162 KB JPG
>>110002182
how are the chink models doing?
>>
>>110001646
I had no clue that it was just for pro users. The amount of people paying for pro that didn't get it purely for codex is precisely 0.
>>
File: 2026-10-07 09.10.44.png (3 KB, 210x92)
3 KB PNG
>button that does literally nothing
How did they fuck that up?
>>
>>110002333
they're pretty good, but opus 5.5 (and to a lesser degree soi 6.1) are a significant step ahead so atm jewish models are firmly at the top
>>
>>110002361
>opus 5.5 (and to a lesser degree soi 6.1) are a significant step ahead
Yeah but if said models refuse to actually do anything like with the xxx-rated content example here then what's the point. Although you could always get deepseek or GLM 5.3 to work through "questionable" content and get Opussy 5.5 to do other work. What a fucking sorry state of affairs that the Chinese models are the uncensored ones.
>>
File: file.png (415 KB, 800x582)
415 KB PNG
>>110002388
China is now the land of the free.
>>
want a way to root a phone using an exploit but get safeguarded. what should i use?
>>
>>110002400
Coding? GLM 5.3. Or maybe the new Mistral Large 4? Just try them. Mistral hosts both btw. This is literally their niche.
>>
File: 341.png (88 KB, 206x264)
88 KB PNG
>tibo reset while I had still 75% usage left
>>
>>110002398
Chinese are hardcore nationalists, if you're a minority in their cesspool of pollution and industrial waste you're fucked
>>
>>110002430
>Chinese are hardcore nationalists
uhhh based?
>>
>>110002388
>xxx-rated
Why the fuck would you use a general frontier model for that when there are both cheaper and more specialized alternatives available (I'd ask why you need AI for that in general, given the amount of porn already present online, but I already this thread is full of retarded gooners)?
>>
>>110002438
You never vibe coded an adult-content website? For a quick buck while you work on actual projects. Porn means $$$.
>>
Is there a good AI debug tool besides connecting Claude to X64dbg and letting it work? I'm burning through usage with it trying to do the work
>>
>>110002461
>AI debug tool
what does it even mean
>>
>>110002465
Just an agent that specialized in reverse engineering/debugging.
>>
File: 1570977464691.jpg (110 KB, 839x610)
110 KB JPG
>>110002331
>Professional Go players weren't consciously avoiding the region of the board move 37 came from; that region simply never surfaced as a candidate within the pattern recognition built by decades of accumulated training. The chapter names this an epistemically invisible constraint, a limit on expert judgment that the expert cannot perceive from within their own framework. Every domain has them, medicine and finance included, and they mark exactly where a well-designed machine system is most likely to find something humans systematically miss.
>>
anyone else their disk getting full due to ai agents one day u add 40gb more space and then next day it's 500mb available what THE FUCK
>>
>>110002477
Specialized agents are only really used in *very* specialized circumstances, like embedded shit
generalist models are consistently better
play around with proompts, available skills and improving the "do thing -> test results" loop rather than searching for some silver bullet
>>
>>110002490
no this only happens to me when downloading porn
>>
>>110002411
>Or maybe the new Mistral Large 4?
Anyone actually tried this thing yet? It's quite cheap especially right now but God knows what it shits out.
>>
>>110002499
really? every codex update is like 500mb-1gb and they don't delete them what the hell
>>
>>110002505
>every codex update is like 500mb-1gb and they don't delete them
not a problem on either macos or linux
>>
>Tibo reset when I still had 50% of my usage left
>It had been 3 days since the last reset and I wanted to make sure I had enough usage for the rest of the week and then get everything reset on sunday as planned so I have a week of fresh usage next week
>Now I'll get reset on Wednesday so I can't use everything by the end of the week
>Openai screws me over again with a reset
>>
>>110002500
it's decent, better than most chinesium, but it's not good enough to use as your daily driver
it's mostly there so yuro corpos which aren't supposed to send data to third parties outside EU can use a non-meme LLM without needing their own compute cluster to run glm locally
>>
>>110002517
>le thicc
you deserve that for having beauty standards of a nigger
>>
>>110002333
Substantially behind it seems the western labs have figured out RSI at scale hence the huge jump in capabilities
>>
>>110002411
But they're gated behind their own software platform called vibe code.
It's different from api tokens and therefore it's not powering hermes or it's workers directly.
So when Hermes has to call vibe to do something, then we use extra tokens of the harness consuming tokens.

I think my chatgpt powered hermes is prompting vibe like we do to get things done. There's a lot of overhead.
>>
>>110002525
Nigga you gay as fuck. I bet you like small muscular asses don't you?
>>
>>110002533
He's low test so he doesn't like fat women. My wife is fat and that makes my dick limp so I'm also low test.
>>
>>110002532
Even through APIs at openrouter? Pricing is identical at least.
>>
new model drop
https://huggingface.co/google/embeddinggemma-2

lets gooo now i can finally make the software i wanted on my rasberry pi
>>
>>110002538
>low test so he doesn't like fat women
low test bros...
>>
>>110002540
Openrouter is expensive compared to buying a yearly sub directly from Mistral
>>
>>110002549
>buying a yearly sub directly from Mistral
why would anyone do this
I wouldn't buy a yearly sub from cuckai / misanthropic, much less mistral
>>
>>110002556
>why would anyone do this
Because more tokens for less money. If you cuck into subscriptions you get more than paying as you go through APIs.
>>
>>110002556
you’re betting hundreds of euros that you’re going to need Made In EU-compliant LLMs for a year and nothing else will do
>>
>>110002564
And what use would mistral sub be before their current release, when all their models were total garbage?
Yea chonk 4 is alright, but are you going to use it 6 months from now?
>>110002567
Only situation where I'd sub
>>
File: file.png (133 KB, 399x498)
133 KB PNG
>be STEMcuck
>good at writing code, never gave a shit about english
>5 years later I'm programming in english and barely reading (much less writing) code
>>
>>110002583
are you getting better at English?
>>
>>110002572
>And what use would mistral sub be before their current release, when all their models were total garbage?
None in practical terms. Mistral has been useful for 2 months in their entire existence since they started hosting GLM in august. You get 30 dollars in API credits for about 18 euros a month apparently.
>>
>sol is dog slow again
fucking hell is their 28 days plan "made sol 50% faster" every single day, only to nerf it a day later?
>>
>>110001575
I use it as extra Astra usage but you have to make it so it doesn't eat usage by doing dumb things like persisting and uploading and saving stuff to the Space/Library since they do reset the VMs. It will eventually have its own quota after the first month. Currently using it as my Blender modeling node I send jobs to kek.
>>
>>110002556
You're a traitor to THE WHITE RACE, of course you wouldn't.
>>
>>110002572
>when all their models were total garbage?
GLM 5.3 was not garbage

>>110001642
I have Hermes agent to prompt from my phone
>>
>>110002648
>GLM 5.3 was not garbage
They've been hosting that for a month my nigga. There was no point before it.
>>
>>110002654
I only bought when they started hosting glm 5.2
>>
File: koi_wordmark.png (15 KB, 495x192)
15 KB PNG
i was able to fix more bugs.
The crazy agent made a fake easy mode that would break the rules of koikoi.
With the help of anohter ai i was able to locate the fake easy mode and removed it.

I am on my way to the doc, it is possible that they will send me to the hospital, lets hope this will not happen.

If i get back, i will continue.

Process on this game is really slowed down, heavily.
>>
>>110002690
good luck
>>
>>110002493
I only ask for a silver bullet because I have no idea what I'm doing and I'm surprised I got this far.
>>
TIBBBOOOOOO I DIDNT WANT A RESET
>>
>>110002592
Nta my english is noticably worse with agents sometimes it couls barely be called english full of typos
You dont gotta fix typos with ai thwy understand you anywya
>>
I think claude stole the term "adversarially verify" from me. because i was using that term when the word adversarially would get flagged by fable's filter so i was stuck using opus 4.8
>>
>using babby's $20 openai sub
>45 minutes since the 5hr reset
>88% usage left
good fucking lord did altman outsource the computation to a bunch of indians that do the matrix multiplication manually?
>>
>>110002810
i think claude stole everything it could suddenly do in the past year from me because i kept trying to make it do that shit and it couldnt now it could
>>
Want to port original Daggerfall to PSP, but modern Gemini is incredibly stupid (it even can't make simple mod for Imperator Rome), while other models are expensive...
>>
>>110002634
Me and my Asian wife both look 3x better than this and I still say this
>>
did elon just give up and turn xai into a model routing company lol
>>
>>110002894
You chose a project too expensive for you then. This will need at minimum $20 Claude plan or the equivalent in API for a weaker Chinese model

Also unless there's a reason to hardcode daggerfall for the psp you might as well make or use a dos2psp library so it's useful for more than just daggerfall
>>
>>110002894
>>
>>110002931
>This will need at minimum $20 Claude plan or the equivalent in API for a weaker Chinese model
I have 30$ on OpenRouter, but it's not for vibecoding.
>useful for more than just daggerfall
I interested in DF only, because I feel that PSP lack of big western RPG and since DF is very old game (nevertheless it has a lot of content) it would be great option for PSP. There is already Exult (Ultima 7), so port of DF is possible, I think.
>>
>>110002542
I feel there are so many cool things you can do with those multimodal embeddings

>>110002911
>let users route their requests to other models
>get back the request from said model
>suddenly you sit on a ton of training data that can be distilled
>>
File: HT2AfL2W8AAP8bC.jpg (144 KB, 1200x676)
144 KB JPG
>>110002936
>>
>>110002968
yeah man all that cursor training data sure did help out for grok 4.7
>>
sol 6.1 on a sub is unusably slow if you use just one main agent. i can't chew through the fucking $20 account in 5hrs, much less the $100
i'm out
>>
my sub expires tommorow 8th when is that exactly? at midnight? or i still have 1 banked reset to use up too gpt pro

wat do?
>>
>>110003025
mod to turn sunless sea into a 3d game
>>
>>110003025
yeah they cut you off at migdnight
>>
File: file.png (213 KB, 903x963)
213 KB PNG
>>110003021
It's a "looped transformer" and it loops twice.
>>
>>110003036
at local time? that would mean 12hrs left i should start using astra max and ultrafast right..? im at 90% with 1 banked left
>>
>>110002972
>>
>>110003045
where did this mascot of openai come from?
>>
File: file.png (3 KB, 176x40)
3 KB PNG
>>110003041
It's not the model itself cause it runs fine through api, it's just slow as molasses on api
I have a task I'd expect to take ~2hrs from start to finish. It's been 2.5 hours and it isn't even halfway through.
>>
>>110003062
>it's just slow as molasses on api
on sub*
>>
holy shit astra 6 max + 1.5 speed used 2% in few minutes
>>
>>110003043
that's what happened to me when i cut mine off
>>
>>110003082
yeah i'll just assume worst case at midnight exactly and adjust my token usage to that ty
>>
>>110003062
>>110003067
They probably throttle too. I've heard on API it's faster and has more thinking juice.
>>
What's a good time to stop a task halfway through in chatgpt? compaction nukes cache, right? so fresh after compaction?
>>
File: file.png (45 KB, 994x224)
45 KB PNG
chuds
>>
I'm ready to start
>>
>>110003062
Yeah same thing for me. I'm reasonably happy with sol 6.1, but it's so fucking slow that it's more like delegating a task to somebody rather than working alongside somebody. Fine for most purposes, but then you check out opus and it's zipping along at 3x the speed despite openai's "muh token efficiency" bleating
>>
>>110003180
inb4 article sponsored by the Tarbell Center
>>
>>110003062
I think cuckai is having some trouble, I ask for a clarification and replies four times, I ask it to pause execution and it paused it three times without any further instructions
or maybe I got routed to luna
>>
>>110003052
chyna
>>
>>110001964
That girl is fucking mesmerized by zucc.
Dude got some rizz now.
>>
File: fig4.png (131 KB, 500x447)
131 KB PNG
>>110003180
>giant confidence intervals
>>
>assume that every vibe coded project is sent to anthropic/openai servers and reviewed by an AI agent told to "look for interesting projects being built"
>pre-empt every prompt with "boring" like "build a boring front-end for my app", "add a boring button", "write a boring function"
YOU'RE NOT JEWING ME ON MY IDEAS CUNTS
>>
>>110003325
keek
>>
File: JL_ZUCK-COMP.jpg (374 KB, 1500x1000)
374 KB JPG
>>110003311
she wants the ZUCC
>>
File: file.png (258 KB, 1919x1035)
258 KB PNG
>Researching building my own browser after writing >>110001602
>go through several rabbit holes, trying to find a way to make v8 work with rust wgpu (dawn doesn't support raytracing), landing in a deno rabbit hole, trying to find a way to use deno compile to use v8 with wgpu
>land on somehow trying to use quickjs javascript engine for the small memory footprint and quick startup with lexbor web browser engine in deno compile, with some rust bindings; this is going to be a lot of work (for claude), to use wgpu-native
>decide to see if anyone has made a browser with quickjs yet
>find some obscure browser called Nordstjernen that uses the exact things I wanted. Quickjs, lexbor, wgpu-native. Even has a lightweight one window at a time version with certain things cut out called Northstar.
>But it's written entirely in C; vibecoded in claude
>it's not bad (can't get webGPU working or 4chan loading past cloudflare and it seems to be loading stuff slow, but it's pretty impressive for not being another chromium v8 clone number 904545)

Funny how life works out sometimes. I'll have to try it on Linux with Vulkan enabled next to see if I can get webGPU working there. You guys should give it a checkout. Maybe we could fork it and collectively vibecode our own 4chan browser some day.
https://nordstjernen.org/index.html
>>
>Working on maps for game
>Doing a walkthrough of the scene to look for mistakes
>Scene turns to night
>headless NPC I didn't set up or create starts walking through the level to a target I never set

Spoopy.
>>
>>110003365
>building my own browser
It may not seem that way but a browser is one of the most complex pieces of software that exist right now, so good luck
>>
>>110003367
AAAhhhh!!! The Headless Jorksman
>>
File: file.png (357 KB, 748x731)
357 KB PNG
>guy finds security hole, infinite free paid oai models inside
>gets handed $300 by bug bounty program
>>
>>110003447
JorksWOMAN

Was a nude female torso. Walked right through the courtyard, up the west wing and into the library.
>>
Gemini 4 Argon soon mayhaps?
>>
>>110003490
*plap plap plap plap*
>>110003493
there will be plap
>>
google is coming for vidyacoders

https://www.youtube.com/watch?v=Dxvrvfyicp0
https://labs.google/playground
>>
File: 028713407f7d29af.png (144 KB, 600x548)
144 KB PNG
>>110003535
>I know! X with... Y!
>>
>>110003535
id play with her on the ground
>>
>>110003535
If that was their A-grade advertisement material I think I'm good for a year or two.
>>
I'm starting to believe in the claude stealth nerf that turns on and off at times.
>>
aaaaaa I didn't track tibo for the first time in weeks and he did a reset, I wasted 40%
>>
>be me
>need a Rust GUI
>surely there is something
>egui
>immediate mode, gross
>iced
>didn't actually look at it
>fails rigorous vibes assessment
>slint
>open licensing page
>unexpected procurement side quest
>tauri
>install browser dependency to display rectangle
>already have browser at home
>dioxus
>don't know what it does
>deeply suspicious
>GPUI
>looks good
>feel Zed standing behind me
>conclude entire ecosystem unsuitable
>scientific process complete
>mkdir, git init, "build a Rust GUI crate, make no mistakes"
>>
>>110003667
>>110003667
NEW THREAD
>>
File: 1784704154629535.png (65 KB, 986x605)
65 KB PNG
>>110003535
>press start
Amazing. The state of big tech.
>>
>>110003689
region-locked to usa for now, launched <30mins ago
>>
>>110003474
I'd feel like such a cuck in his place
much rather give it to chinks so they can rip off their homework and improve the open weights models, or hell just use it yourself
fucking $300
>>
>>110003535
Ummm ahkshually they should pay a tax to all indie developers and artists and coders and me in particular before releasing it
>>
Why is Sol 6.1 feeling so tarded this week?
>>
>>110003733
that's what you get for keep demanding resets
>>
>>110003535
At least they managed to put a diversity hire that's sort of cute in the video.
>>
claude assisted skyrim gooning
>>
File: shocked.gif (17 KB, 220x238)
17 KB GIF
Working for snailcats is driving me insane how do I give them ai psychosis?
>>
>>110004352
automate your boss's job kek



[Advertise on 4chan]

Delete Post: [File Only] Style:
[Disable Mobile View / Use Desktop Site]

[Enable Mobile View / Use Mobile Site]

All trademarks and copyrights on this page are owned by their respective parties. Images uploaded are the responsibility of the Poster. Comments are owned by the Poster.