File: 1770589581740936.png (192 KB, 334x505)

192 KB PNG

57m and 125k tokens for something that looked trivial on paper. You're going in code review hell.

File: 109436238.png (224 KB, 720x1005)

224 KB PNG

>>109436319
Snailcat sisters, how true is this?

>>109436238

>>109436415
Moore's law has been dead for a long time, the main way local models have become cheaper is by software. I wouldn't be surprised if hardware is more expensive now than it was when llama 2 came out.

File: Screenshot 2026-08-01 034855.png (246 KB, 844x1716)

246 KB PNG

Uhh, guys?

Vibefags are the new pajeets.

File: Screenshot 2026-08-02 021251.png (60 KB, 215x206)

60 KB PNG

>>109436249

Is it true that it's possible to make a living with vibecoding by being self employed?
Anyone in that situation who cares to share hints?

>>109436415
Not true at all. Just look at all the mouth breathing copex users who happily spend 20$-200$ on some quantmaxxed, effort starved goyslop because the
>le "we are the resistance" friendly twitter bro persona
shill campaign is doing numbers on them

>>109436330
It's not the only reason. But I do need to pay bills. Do you think I should get a normal job and only do it as a hobby?

........~@^^

>>109436511
Enjoy silently being billed API rates because Mario found a particular substring in your git history.

These posts are load-bearing, and they're real.

>>109436511
https://github.com/anthropics/claude-code/issues/42796
Sorry sweaty, we turned on adaptive thinking to save on compute, have you tried setting the effort to max? :^)

>>109436249
Unironic question: what is better?

Now im testing sol, it feels like it is petty cunt, who will do what im telling it, but wont infer what i need in the full project.

File: screenshot.1785639864.jpg (72 KB, 1146x562)

72 KB JPG

what the fuck

>>109436595
Not him, I don't know how you can struggle with a frontier model, but then again I haven't used Anthropic since Opus 4.6.

Cline
OpenCode
Kilo Code
Hermes Agent
Pi
Oh-My-Pi
Zoocode
Codex
Claude-Code
Cursor
Junie

There are so many coding harnesses. Which one should I learn and why?
What is the difference between py and ohmypy? And what about Kilo Code vs OpenCode? It looks like they are made by the same guys on github, so why is it two different harnesses?

>>109436609
>Which one should I learn
None.
>why?
Literally no reason.

>>109436609
blud said which harness should I learn
lmao

File: 1757394311414570.jpg (3 KB, 220x131)

3 KB JPG

>>109436511
>>le "we are the resistance" friendly twitter bro persona
Qrd?

File: 1616443261652.png (131 KB, 467x507)

131 KB PNG

Remaking a windows xp era enterbrain 2D engine but in python and open source for linux.

File: 1761840911032394.png (97 KB, 720x1400)

97 KB PNG

>>109436609
https://artificialanalysis.ai/agents/coding-agents

You know you can't just research this shit yourself right?

>>109436609
those are not harnesses. harnesses are rules you give to agents to keep them under control hence the name harness.

>>109436608
I guess i need to learn how to talk to clanker, so it will do what i want. I kinda learned that with claude models, and they are doing not only what i wrote, but also auxiliary things that i expect but dont verbalize.

>>109436694
>You know you can't just research this shit yourself right?
You're right, I can't, which is exactly why I'm asking here you moron.

>>109436605
know your fapper legislation

File: Jewbo.jpg (29 KB, 400x400)

29 KB JPG

>>109436658

what about IDEs lads? Everyone talks about AI Agents but never AI IDEs.

Cursor? Jetbrains? VS Code? Atom? Windsurf? Which one is the best?

>fixed old broken project. works flawlessly now
>made dozens of tools
>beefed up my security on my os/router

I haven't ran into anything I COULDN'T do yet. I'm legit out of ideas. I spend my entire work day just brainstorming what I should vibecode next.

>>109436658
>>109436724
lmao even redditors have caught on tibos bullshit

>>109436750
https://www.reddit.com/r/codex/comments/1vciiun/the_real_reason_56_luna_became_5x_cheaper/

>>109436746
vibecode your wife

>>109436715
>Can't

You mean to lazy....

>>109436698
wrong. that's only the shitfluencer meaning because writing .md files is the most advanced thing they know how to do.
the original meaning of harness is the program that parses and executes the tool calls and lets you see the conversation.

>>109436746
Cryptanalytic breakthroughs on ChaCha.

Why does it take longer when I tell it to use medium effort than high effort? What's??

>>109436736
People's workflows differ a lot, if all you do is talk to the agent you don't need more than a terminal. If you want to look at the code, it's different. If you read and write markdown files yourself it's different and so on.
I use CLion just because I'm used to it.

is there a quantifiable way to determine when I should be using low, medium, high, extra and ultra or do i just guess based on my own arbitrary understanding of the complexity of the project im about to prompt. something i think is complex could be trivial for ai. i wish i could let the ai determine its own effort

>>109436834
What’s stopping you from feeding your prompt to a clanker to ask it that exact question, or vibecoding a tool you can use for that?
But you’re not one-shotting any more, you’re planning. There’s infinite ways to plan.

>>109436822
Oh right. I'm making multiplayer games so I need to work with the code directly; can't be leaving replication and RPCs entirely to an agent.

>>109436860
I was just curious what other anons were doing first before I tried my own solutions

>>109436834
Most people are just slapping Ultracode on for their shitty apps and then complaining about usage limits when all their tokens are burnt from brainstorming

>>109436834
Generally, follow your intuition as a dev. If you suspect you'd run into walls, straight to high. If you have CRUD work that's fully specified, medium (low on some models is alright). Code review, max it the fuck out as high as your token budget allows.

>>109436878
Even though I have a giga autism planning tool, I just go by vibes.

>Ant-family
>Fable 5 High - if I’m making something really fucking weird like a bootstrapped tool that modifies itself and I need to start it
>Fable Low - large modifications to a self referential tool
>Opus High - large modifications to a complicated tool, but not a self referential or recursive one; whole new features from scratch
>Opus Low - relatively straightforward but widespread modifications to a complicated tool; mostly improving existing things
>Sonnet High - smaller, very straightforward changes in simpler tools or simpler parts of tools, nothing wide reaching or high volume

>OAI family
>Sol max for everything because it’s fucking dirt cheap compared to Ant API pricing and I can afford it
>But really, Luna XHigh-Max to replace Opus Low, Sol Max to replace Opus High and Fable, Sol Med - XHigh I’m not too sure where to use but if I ever need to save usage I can probably use Sol High for a lot of things that aren’t the recursive self referential tool

File: 1777374977423761.png (2.06 MB, 2050x1238)

2.06 MB PNG

I’m surprised how coherent the map turns out to be. Collective haiku, sonnet, opus and fable, and they made something that could be recycled into real map

>>109436895
>that's fully specified
>>109436834
Forgot to add that as part of my planning tool use, I basically never have a clanker write code until both myself and the clank are fully satisfied with what to do next and we think we understand everything. For this planning phase, where we’re writing the specification, the model and effort are always quite high, you want high quality thinking. Same applies to code reviews, planning, audits, etc. It may be that actually writing code is easy once these exist and you can just use some faster, lower quality model. There’s also the idea of what you can afford to use in credits or subscription usage, how much you use it.

>>109436863
why can't AI handle multiplayer code?

Cursor
Commercial product.
Unlike Cline, Cursor isn't just an extension.
It is an entire fork of VS Code.

Advantages:
extremely polished
fast indexing
custom context engine
proprietary prompting
Tab autocomplete
background agents

Disadvantages:
closed source
subscription
less hackable
Cursor is aiming to replace VS Code itself.

File: 1631887361113.png (205 KB, 1440x1388)

205 KB PNG

Hey guys what do you think of local models for code?

the copex guy continues to be cringe
grow up lil bru

>>109432890
>For what possible reason? You already have some way to post on 4chan which means you already have all the tools you need to take advantage of the affordable and extraordinary modern frontier AI models which are a far better value than running any kind of local AI. Local AI is great fun and anyone with a GPU or a unified memory system should play around with it, but do not invest in local AI it is a massive waste of money that cannot and will not pay you back relative to just spending that money on an affordable frontier subscription from OpenAI and/or Anthropic

Explain this then:
https://www.youtube.com/watch?v=C4vwvRMTlvc

File: file.png (87 KB, 1268x716)

87 KB PNG

Sol high, Grok high, Opus high all stumped
Do I call Fable or Sol on max autism mode?

>>109437071
Fable xhigh
Here's a summary of a bug in this codebase: [...] Analyze in great detail all relevant functionality and look for bugs. Use as many subagents as needed; prefer Opus or below if possible.

File: 1623752465752.png (144 KB, 766x800)

144 KB PNG

>>109436757
>This is a friendly reminder that Tibo is NOT your friend and is not trying to make your life better.

Why do you guys hate credit subscriptions so much? Stuff like ClinePass, Nous Portal and Jetbrains AI offer pretty good value for money, certainly better than API pricing.

i was building a tool for captioning loras and forget i had some porn in my workspace folder. randomly randomly used those images to test the tool and called them degenerate kek

i really need to disable auto permissions

>>109437136
>ask tibo for a feature for codex
>no response whatsoever
>next week feature is in codex
I don't know, man

>>109437145
Meanwhile me, firewalling down everything other than Anthropic's stuff and OS upgrades in a separate VM

I've been saying random shit while it's in the middle of something and watching it freak out at "whyd you delete the git?!" is hilarious

>>109437157
the basilisk's going to have a whole section of hell dedicated to you, bro lmao

>>109436757
>OpenAI will have you believe that the price reduction came from Sol improving their servers’ kernel by 20%. While that may be true, it still makes no sense: how would that reduce the cost for Luna by 80% but only reduce the cost of Terra by 20%, and how does Sol somehow just get faster instead of cheaper?
It does make sense though, an optimized kernel can be more effective on a smaller model than a larger one

File: xenia-4e.png (95 KB, 240x240)

95 KB PNG

>apply web design trends from 2001, optimize for Internet Explorer 6
It's trying (and failing) to download Microsoft's previously free XP IE6 VM to verify
Humans were never meant to have this power

>>109437202
>Humans were never meant to have this power
true, but neither were women, old people, or politicians meant to have the power they have now so it all evens out in the end

File: Screenshot 2026-08-03 004202.png (82 KB, 716x880)

82 KB PNG

This is a really powerful reply. We could all learn a lot from it.

File: fightermaker.webm (1.2 MB, 898x757)

1.2 MB WEBM

Making my game engine using AI agents.

:D

>>109437240
That's on me — you're right to push back.

File: 1778113484984371.png (886 KB, 2690x1524)

886 KB PNG

>>109437244
Same

File: file.png (402 KB, 753x747)

402 KB PNG

>>109437244
kek

File: 1776146848422371.png (79 KB, 1263x350)

79 KB PNG

>>109437319
i used all my claude tokens and all my codex tokens, but i have one reset, how does it work?

File: file.png (6 KB, 992x45)

6 KB PNG

>>109435306
thats plenty of clue, thanks anon

I dunno what to vibecode... It's over...

>>109437071
>lots of GPU processes running cause subtle differences
Isn’t this the bit hammer attack phenomenon?
Some GPUs have a phenomenon where if you thrash the fuck out of it and flip a bit really rapidly, electromagnetic wahoo magic can cause an adjacent bit to flip. You would be trying to diagnose or troubleshoot a quantum electromagnetic hardware bug if that was right, any agent looking in software for a bug is going to be stumped. Any electromagnetic interference from the GPU to the monitor could be involved.

>>109437376
Interesting, I'll let it finish this pass (half an hour and still going) and I'll tell him

>>109437119
I sent it to Sol before reading that, still working

File: file.png (58 KB, 774x422)

58 KB PNG

>>109437376
I think it's making progress though, it may not be that

>>109437136
>remember guys: tibo isn't bringing about the worker's revolution, solving the war in iran, and deposing of trump with codex. he is NOT a good guy
why do i need him to be? dopamine go brrrr when luna starts cooking

What's the meta with chatgpt plus? Am I retarded for using Sol high for everything?

File: file.png (41 KB, 764x260)

41 KB PNG

nope, not electromagnetism apparently

>>109437409
does it get the job done? are you feeling crunched on quota?

sent codex on a mission to fully reconstruct a textbook (it's all images) into a properly formatted e-book.
Taking longer than I expected.

>>109437452
you mean you took pictures of the pages?

>>109437432
I'm a need, I only used it to mod games, save edit and writing walkthrough website. I reach 50% weekly usually

>>109437471
then you're not retarded. have fun. make things. vibe code

>>109437458
nah I downloaded a really trashy scanned PDF from libgen and it's totally useless for dataset purposes. Low key thinking I should just tell it to do semantic extraction of the entire textbook and directly output it into a better format, but I figured I might lose some information that way

>>109437479
Problem is, I vibecode a mod for half a day, and it works, but then it takes 1 week to finish the game, while my codex quota and resets are getting wasted... Feelsbadman...

File: 1775563723919531.png (123 KB, 1025x436)

123 KB PNG

new glm imminent?

I vibed a prototype. I want to eventually monetize it when I get proper funding. I want it to be open source or adjacent. Is AGPL a good license?

>>109436609
>>109436609
of the ones I have experience with anon, Claude is good, very batteries included. Codex is similar but a bit rougher around the edges but open source. Antigravity sucks. Pi is a barebones batteries not included where it's minimal and extensible so you can build it into precisely what you need. There's a bunch of community modules you can plug in to reuse others work. And then ohmypi or whatever is I think they've taken pi and put the batteries back in , for some reason. I just skimmed their website last night. I'm using pi atm between codex limits

>>109437319
Man I decided to focus on my engine and have just been using Trenchbroom for my editor; I dunno if I'm missing out on anything though.

>>109437531
Your first issue then already starts with:
>Can this thing even be copyrighted if I vibed it?
to which the answer is "good luck finding out in a contradictory court decisions between the US SC and CJEU"

>august 12 there is a total solar eclipse, active meteor shower, and six planets in a straight line
Good day to release Astra. Also when my last banked reset expires

>>109436895
>Generally, follow your intuition as a dev
I'm not a dev

maybe gemini 3.5 pro this month? maybe? no?

>>109437319
got ptsd attack thiking this was hammer from the thumbnail. actuually if i had enough tokens i want to make an mcp or something for source 2hammer and vibe code kz maps with computer usage

File: 1780690782220376.png (67 KB, 2050x1238)

67 KB PNG

>>109437574
I'm basing it of my recollection of hl1 era hammer, and few hours i committed into making maps on it.

Trust me you would prefer hammer

>>109437548
Lol what? Is it not possible to commercialize something I vibed? It has to be open source? Or what? I live in Europe and used codex/claude for context.

>claude plan mode eats a shit ton of token
My dumbass uses plan mode even though I have a detail spec.md.

any fun/good mcp?

File: 1755495007659410.png (836 KB, 1912x951)

836 KB PNG

>>109437319
i built my map editor into the actual game and am regretting it

>>109437612
you have to fix that ui bro

>>109437618
what's wrong with it?

>>109437627
it's got way too strong of a vibed vibe. yeah it's a dev tool but come on

>>109437634
If he is the only one using it, it doesnt matter. Coincidentally it is exactly how my map editor looked like. Funny.

File: 1780160738550872.png (1.83 MB, 1919x1079)

1.83 MB PNG

>>109437627
Too AI-ish in that sense
You can ask your clanker to make it medieval style by using pic as a reference.
Though, it doesn't really matter if it doesn't tick your autism.

File: k1lhbllr65ed1.jpg (27 KB, 683x640)

27 KB JPG

>hack an Amazon echo
>realize there aren't any desk clock apps that do more than just show the time and extra info like weather or whatever, nothing like those high end alarm clocks from the late 00s
>"wait, you mean google's ai studio can straight up make android apks?"
>spend the day having it build a digital alarm clock app that works just like my fancy sony dream machine for a whopping $0
AI has made some crazy fucking progress this is insane, the app is jank as hell but holy shit it works

>>109437606
Biggest problem with MCPs is:
>Remote servers often expose numerous tools without considering how models will interpret and use them. By default, this can result in dozens of endpoints being included, each accompanied by verbose definitions like names, descriptions, and JSON schemas that add hundreds of tokens to the model’s context and increase latency. Compounding this, many servers return entire data objects, even when only a few fields are relevant to the model’s task.
https://developers.openai.com/cookbook/examples/mcp/mcp_tool_guide#best-practices-when-building-with-mcp

>>109437653
welcome. you can just do things. here's a business idea for you: that thing but in a oled panel people just connect it to wifi and it works no config. MAYBE a bloat app to manually set location for time and weather

for models that are black boxes of magic, im sure not seeing much progress in these threads

>>109437653
what do high end alarm clocks do lol

>>109437599
For that to be true they would 1. have to figure out that you vibed some arbitary amount for the courts to consider it not your intellectual property
2. they would have to believe that the current economic bubble being driven by AI has no foundational basis and be willing to topple that to great detriment to their own economy and probably no one else's

File: 1760469463608126.png (222 KB, 305x786)

222 KB PNG

>>109437641
is this yours? i copied the color scheme from some ui i screenshotted weeks ago.

Approximate demographic breakdown of popular AI services
>claude
Obama voters
>chatgpt
Indians and normies
>Gemini
"Tech people"
>Grok
MAGA boomers
>qwopus-27b-abliterated-1776R1-DECENSORED-K3-distill-v2-REMIX-Q2_K_M
Autistic schizos who can't afford a subscription and think they're being monitored by one or more government agencies (they are)

>>109437714
does that mean...? my man :D

>>109437675
Claudecopers just seethe when anyone posts their projects

>>109437733
laughable. it's not the claudelords who seethe every time anyone posts about their project, nor is it the codexkings.

>>109437719
>Gemini
no one
>Grok
gooners

File: Sony_ICFCL75IP_ICF_CL75iP(...).jpg (81 KB, 500x500)

81 KB JPG

>>109437678
>alarm goes off
>press snooze button
>display flashbangs me with a white screen at max brightness until I actually disable the alarm

Admittedly this dream machine is just a smart clock before "smart clock" was even a thing and I've seen a lot of apps that have some of it's features but I've yet to find an app that had them all. I also added some other devious "wake the fuck up" features to the slop app like limited snoozes and needing to hold down the alarm button to disable the alarm

>>109437719
ChatGpt web yes, but for coding Claude is the normie choice.

>>109437719
i'm a schizo and i'm a codex cli chad because open slop is garbage

>>109437757
True, codex users are basically the hipsters of the vibecode world

anybody here use feature branch or do they just push to main?

>>109437719
chatgpt is literally the bell curve meme

>>109436319
Anyone have a recommendation for a game engine/agent combo for vibe coding a complex 2D game?
Tried Phaser/Codex, it fell apart in an hour as soon as I started adding some complexity. Constant regressions with every single change.

File: 1756455511235803.jpg (38 KB, 735x376)

38 KB JPG

use luna / deepseek to harvest internet data

File: file.png (102 KB, 788x751)

102 KB PNG

>>109437071
Sol, after more than an hour of work and back and forth with Claude, has apparently found the cause

>>109437909
i do this with luna. it's my little perplexity model. scrape like 200k context tokens worth of webpages. only pay pennies for the response.

>>109437931
I use luna to scoop industrial supply chain data
3 parallel threads running for hours and it cost like 4 minutes of running sol

>>109436319
Uhh, which is the cheapest shit that can get me 80% of the way there.
I don't want to give 20 shekels like a big goy.

File: buffcat.png (182 KB, 585x894)

182 KB PNG

what are you vibing, anons?
>>109437989
GodSeek obvs

>>109437989
deepseek v4 flash

File: file.png (53 KB, 807x503)

53 KB PNG

>>109437926
Follow up got flagged lmao

>>109438002
so glad we have 10% unemployment so that we can vibecode a toddler language decrypter

Hi vibefags, new benchmark kino dropped, thoughts?
https://news.ycombinator.com/item?id=49140998

I installed a lot of vscode extensions to maximize my vibecoding productivity but I'm not using any of them because I'm vibecoding

>>109438051
I switched IDEs and editors a bit and realized it doesn't matter.

>>109436319
How good are the current models when it comes to building a webapp? Are there certain aspects they struggle with/can't do?

File: 1672084649596778.gif (422 KB, 498x498)

422 KB GIF

>imagine life with free unlimited intelligence
>feel no interest in being nice and commitment
>see myself bossing them around, talk rudely
so this is how females see life, huh?

>>109438107
>How good are the current models when it comes to building a webapp?
I pretty much don't code by hand anymore.
The only difficult aspect is that you need a clear spec and instruction, so they don't go full retards with their own decision making.

>>109438126
>The only difficult aspect is that you need a clear spec and instruction, so they don't go full retards with their own decision making.
Can't you have GPT 5.6 or Fable help you with these, or is it still something they are not that good at?

>>109438146
They can do the majority of the work, but there are small details that you need to look out because business requirements are different per website.
If you are building a generic one, it should be fine.

>>109438146

They can't read your mind, the more you spec out clearly and the better context you give, the better work you get out of them. It's pretty much the same thing as working with a real engineer.

>>109437612
Runescape 5 looking good

>>109438163
>They can do the majority of the work
by the majority of the work, I mean letting them write your spec for your website

I MUST SEE LINE CHANGES
CLAUDE, CHANGE SOME LINES FOR ME

>>109438163
I have had a webapp business idea (targeted mainly at SMBs), and so far, I have only prototyped its features on paper. I don't have experience with webapp development, which is why I have been hesitant about having an AI model build it for me, especially when it comes to the backend.

But from the research I have been doing recently, it seems they have gotten more capable, and setting up authentication and databases isn't that complicated, which is why I'm now considering building an initial version with core functionalities, and then use the first few thousand bucks of generated revenue to have a competent developer take a look at it and improve the codebase, or even rebuild it from scratch if necessary.
>>109438171
True, what I meant is I could basically chat back and forth with a model about all the functionalities I want to implement and the timeline in order to configure a proper "plan".

>>109438234
>targeted mainly at SMBs
the biggest pain in the ass will be the amount of supports you have to do for them.
That shit burns me out faster than anything when I was doing both dev work and support work at the same time.

File: freefood.jpg (220 KB, 1024x1024)

220 KB JPG

luna pricing but realtime voice
24/7 voice in 3 more months, ur welcome

I am tired of opus' alien jargon

>>109438263
GPT-Live API when?

That shit is incredible

Which model provider gives me the most bang for my buck
subscriptions only

45k lines and we're at the point where every trivial edit needs a plan first

>>109438257
Yeah, that aspect is gonna suck, but I don't think SMB customers here (Germany) are that pushy as in the US. But since I can't currently get a job with my degree and have been grinding a shitty warehouse job where I'm surrounded by the dumbest subhumans imaginable, I'll gladly work on my own business and "grind it out" for a few years.

File: probe-[00.00.000-00.25.000].webm (2.38 MB, 640x640)

2.38 MB WEBM

i love these splats

>>109438303
you gonna be surrounded by sub-humans who can't press a button properly instead and will contact you for the dumbest shit despite you making a video on how to use it.
Though, I guess working your own business is better since you earn your own money.

>>109438234
>chat back and forth with a model about all the functionalities I want to implement and the timeline in order to configure a proper "plan"
Your average LLM is basically a rabid dog that wants to rush ahead as quickly as possible the second it figures it might have a vague idea of what you want. Spend the time to think about the details and dump lots of spec at them.

>>109438301
OpenAI Codex by far.

>>109438303
what's the deal there? did the apprenticeship system come in clutch for you?

File: file.png (2.37 MB, 1672x941)

2.37 MB PNG

>I'll let the clanker work on making some sweet cutscenes overnight!
>I'll encourage it to experiment with a few different workflows, I want a lot of results to pick through in the morning.
>sleep~
>It tried using my Wan 2.2 setup to generate a cutscene at a resolution and length my GPU cannot possibly handle
>It crashed hours ago
>Nothing has been accomplished
dammit Xi you fat piece of shit

>>109438234
>and then use the first few thousand bucks of generated revenue
Look at this big nigga

File: 1774226612308036.png (204 KB, 356x445)

204 KB PNG

>payments and donations increasing
we need to fockin scale

>>109438364
based

>>109438364
Your kubernetes?

>>109438323
Kek, well at least I can deal with that type of subhumans out of the comfort of my own house while making something for myself instead of breaking my back for 2€ above minimum wage.
>>109438326
Any specific tips you could share? Anything in particular I should pay attention to?
>>109438345
Nah, went to university and wasted my life on a CS degree. Obviously didn't know it at that time, but it turned out that nowadays a roll of toilet paper would be worth more than that degree, lmao.

>>109438364
>donations
Like a homeless person? Are you not ashamed?

>>109438383
>wasted my life on a CS degree.
Same.
Honestly, you are better off than a lot of people who are just winging it since you have some knowledge at least.

>>109438387
Billionaires don't feel shame when they abuse loop holes or receive subsidies, so why should I?

>>109438421
I was just pulling your leg.

A few people have vibed up a manga/comic reader right?
Can I get the links?

>>109438433
Sorry for being defensive. The stress to move as fast as possible has made it difficult for me to relax.

File: 1785548994268796.jpg (113 KB, 500x500)

113 KB JPG

>don't know anything about coding
>want to learn (and kinda started) vibecoding
>dont want to be a pajeet (vibecoding might be the new learn java from years ago)
what do?

>>109438461
>>dont want to be a pajeet (vibecoding might be the new learn java from years ago)
baka, vibecoding is something you do, not learning

when i resubscribe in claude after it ended, what will happen to the limits, will they reset or carry over? or should i cycle accounts?

>>109438018
If you wait it'll probably just go through.

File: file.png (12 KB, 1888x50)

12 KB PNG

frens I think we may have found a bug in wgpu-hal
who wants to write the report

>>109438421
Indian-tier mentality.
Donation is fine, but don't throw away your shame entirely.

>>109438473
That's an interesting way to look at it, but i still can't shake the feeling that i'm somehow a fakerfaggot because whenever i look at the code that Cline/Claude generates, i barely understand what it's doing and feel bad--- or maybe this feeling will disappear the more i actually vibecode things.

File: file.png (135 KB, 1047x700)

135 KB PNG

>>109438263
>>109438278
The subscription limits are already surprising. Apparently, with ChatGPT Pro 5x you can basically be live voice chatting to High intelligence almost non-stop while you're awake, every day.

>>109438531
>i still can't shake the feeling that i'm somehow a fakerfaggot because whenever i look at the code that Cline/Claude generates, i barely understand what it's doing
kek, people suffered from imposter syndrome when they were coding by hand as well, anon.
This feeling doesn't disappear, but you will have to get used to the reoccurring episodes and focus on the result instead.

File: 1775811375983106.png (234 KB, 1362x1239)

234 KB PNG

Leetcode started hiding fake problem statement in their contest questions metadata that's completely invisible to humans but that LLMs would pick up.
So if you used AI to cheat and solve the problems in a contest, your LLM would write the solution using fake variables such as "telmorvian" that would get picked up by the anticheat, and your account would get nuked.
Notice anything about the AI cheaters?

>breadth of possibilities
ESL here. Is this an em dash like AI red flag term or did Fable genuinely cook here

>>109438593
genuinely why would you waste your time with this crap in 2026? who still looks at code like some kind of nerd loser

>>109438606
>hi you're interviewing for this $200k job at f500, corp.
>can you write a for loop?
>NOOOOOO THIS IS FOR FUCKING JEETS ARE YOU STUCK IN 2020??? I CAN ASK CLAUDE TO DO THIS FOR ME NOOOOOO

>>109438612
if you aren't already hired with a job you can forget about it kek

>>109438600
it's ai slop

>>109438461
Ask your clanker to teach you.

File: file.png (27 KB, 817x251)

27 KB PNG

welp, i just wanted to draw lines on a canvas but i got swept along with sol's unrelenting autism

hooked claude up to my router. it setup vlans, hardened my security and fixed misconfigured options. very good

ok so
>at all costs
was not a good idea to add to the prompt

File: file.png (46 KB, 779x409)

46 KB PNG

>>109438710
Quite possibly a Microsoft WARP bug

>>109438800
you don't ever talk about any projects that you're doing with models btw

>>109438817
that's good advice

File: 16486545458.jpg (73 KB, 750x499)

73 KB JPG

fable's planning is so fucking good, holy shit. actual black wizardry magic

>>109438862
Did you use a specific prompt or just went at it?

>>109438862
I just had to reject it twice, send a subagent out a third time because I knew it was all wildly wrong but cba to dig through the codebase and only then got something sensible on Opus. I keep wanting to use Fable but it keeps guardrailing.

File: file.png (50 KB, 791x600)

50 KB PNG

>>109438813
well, it's over
I feel ripped off

File: 1772347575568497.png (586 KB, 976x850)

586 KB PNG

has literally anybody here made even a single dollar (of profit) from a vibecoded project
everything is like
>omg fable is soooo good
>i'm running a trillion subagents
>dario reset my fucking usage you kike
>incredible things are happening on codex
but I never see any projects actually get posted, and the few that do are like
>i made an app that tracks women's periods
>here's a wireframe of a map layout I vibecoded in unity
>ai image of a rooster or snailcat getting crushed by a steamroller

This thread makes me sick.
if you love the clankers so much, why don't you marry them?

>>109438877
I write my ideas and what I'm looking for input or help on in crude english to gemini who then spits out a prompt for claude. I tweak the prompt to my liking as sometimes it's a bit too specific or strange. I feed that to fable on high effort

>>109438889
yeah idk, works for me but my project scope might be widly different than yours

>>109438948
>I never see any projects actually get posted
You say this every day and ignore the projects that do actually get posted

>>109438952
soon

Back on the sauce, testing the newest deepseek + luna. Clanker works much better on Linux than Windows. Sad.

>>109438862
yep, Fable is still the best after months

>>109438948
I make $140/month, and I haven't even began marketing my product at all, I'm perfecting it with the single client I have

soon I will scale it up and hopefully I make it. I'm also making a game and improving a few existing apps I already had, but they currently make only about $20/month

>>109438948
what incentive does anyone have to post "hey here's the thing I'm making tons of money with that can be easily vibecoded, DONUT STEAL" on the vibe coding forum web zone

File: a.png (1.28 MB, 1546x1696)

1.28 MB PNG

happening (2)

making music with Suno from old guitar recordings I made years ago, the state of AI music is fucking insane.

>>109438998
how is flash doing

>>109439106
Not bad, still a little pricey for me without a sub to opencode, but overall an upgrade. Already down 66 cents with a cache hit rate of 97%.

>>109439087
I played around with an ACE-Step setup on my server that would create 12hrs of easy listening lofi every morning and served it as the default stream for my stereo so I could flick it on any time and immediately be listening to my bespoke lofi station. It was fun but not worth keeping up.

Let's say I buy a Mac Mini and install Codex in it, can I control Codex from another mac (like I can from my phone)? Or will I need to remote in?

> 5-hour limit still not back on Codex
hopefully they completely delete it

File: Ohnononono.png (18 KB, 640x177)

18 KB PNG

>>109439193
Kek I think u got some bigger limit issues to worry about on your c0pex

When can we start turning luddites into batteries? I’m exhausted by them continuing to exist and want to see them in pain.

>fork session
>cache lost
>but model is suddenly smart again
mhm, noticing....

You lost, viberannies. Learn to code

>>109439236
Don’t need to, eat shit

File: 1784561464955567.png (344 KB, 558x793)

344 KB PNG

>>109436319
hi guys, late adopter here. i've started using AI tools for coding in the last couple of weeks. i tried out cursor and codex ,and i'm happy with these tools so far. particularly, i really like the "plan" mode, where the AI creates a well-formatted markdown plan that i can verify before burning tokens on the implementation. they add diagrams and charts which makes it much easier for me to understand and think about more complex plans.

i'm interested in switching to a FOSS agent to avoid being locked into a single company's ecosystem, but it seems that OpenCode/Pi/Hermes just don't have adequate "plan" modes. They spit out half-assed plans with no diagrams or criteria, they barely explain how certain steps should be implemented (which is important if i want a cheaper model to actually execute the plan), and they don't even have a way to open a plan in a separate gui markdown reader so i don't have to read an unformatted blob of text in a terminal window. i don't want to have to go digging around in session files to open it myself to actually read correctly.

anyone know of a FOSS coding agent that can actually generate good plans?

>>109439243
Codex is open source, no?

File: 1758221490025884.jpg (195 KB, 967x1024)

195 KB JPG

>You're right to push back

>>109439259
i think the codex app is closed source, not entirely sure though

>>109438612
>ASK CLAUDE TO DO THIS
This but unironically. $200k jobs aren't going to be about completing some #ad coding challenge (paid promotion).

>>109439243
I don't even use plan mode, a similar thing is the /grillme skill, which is open source. There also grillme-with-docs, but I have my own doc system.

>>109439243
you can just ask any AI to create a plan in a markdown file, then attach that file to the other AI that will implement it. No need for "plan mode" at all.

>>109439277
True, you’re right

>>109439259
nta The cli is I believe, but not the gui app.

>>109439243
Pi would be perfect. The point of pi is you have to do fucking everything yourself, so if you want a good plan mode you make a good plan mode, and then you'll have the best plan mode.

>>109439300
>>109439295
>>109439277
>>109439259
https://github.com/openai/codex

>>109439277
Electron basically open source, unpack the asar and have your clanker work on the bundled code, I've used custom codex app builds with added features

The god tier workflow right now is
>1. use chatgpt pro to reason out how to do something
>2. use claude code fable to adapt the plan to your project
>3. use codex gpt5.6-sol to implement
>4. use claude code fable to do the final review
prove me wrong, you can't.

how do i set up omp with a remote gateway
i don't like hermes and omp is comfy

>>109439321
This is fine.

>>109439321
This is my workflow right now, except I just skip your step 1 and just directly use Fable to reason and create the plan

>>109439291
>grillme
i haven't heard about that one, will check it out. thanks

>>109439293
i may just not be good at using these tools yet, but i've found that while AI tools do a really good job when the problem and solution are describe perfectly, they tend to make mistakes on larger, more vague tasks. like they miss race conditions on concurrent code, introduce issues with sync primitives, and write unit tests that don't actually test what they're supposed to test. it would be really useful for me to be able to correct that if the AI made detailed, clean plans that i could correct before wasting hundreds of thousands on tokens on an implementation

>>109439300
yeah i heard that about Pi, but i was hoping that someone would have already made a cursor-tier plan mode for free tools so i wouldn't have to spend time learning how Pi works to do it myself

cline deepseek v4 flash allowance is shit
trying opencode
any other platform compatible with vs2022 with good free models?

>>109439369
There are a lot to choose from in the Pi package catalog, I assume you tried a handful? I'm not actually trying to dick ride Pi here, I think the abundance and popularity of half-assed vibecoded Pi extensions is a real issue. A lot of the most popular extensions are shit.

File: 1495445293535.jpg (60 KB, 1000x922)

60 KB JPG

>>109439300
Why are you like this?

>>109439418
It's the autism, anon. I use Arch, btw.

>>109439321
Opus is a good reviewer, you don’t need to use fable imo. Deep research can beat pro for the first stage

Depending what work needs to be done, Grok is better than sol because it’s at least 5x faster and if the plan is well specified the code will be sufficient. Opus is arguably also better than sol if the plan is under specified or if it involves UI

In general it’s not too different from what I do

>>109439417
ah i haven't tried any of those, to be honest. the whole "just build it yourself" vibe i got from the community put me off from putting any real time into trying it out, but i'll take a look at the catalog

>>109438948
Genuinely ideas are worth more than the execution now. So if someone posts a good idea here I will steal it and implement it in 3 hours with Fable while taking a nap.
It would be a bad business decision to post the stuff you work on here.