[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]


A general for vibe coding, agentic engineering, coding agents, AI IDEs, browser builders, and shipping code with LLMs.

-- Frontier models - start here if you have $20 or so
https://claude.com/product/claude-code
https://developers.openai.com/codex/cli

-- B-tier
https://x.ai/cli
https://platform.deepseek.com

-- Prompting / context / skills
https://arps18.github.io/posts/claude-code-mastery/
https://simonwillison.net/guides/agentic-engineering-patterns/using-git-with-coding-agents/
https://github.com/mattpocock/skills

-- Other editors / terminal agents / coding agents
https://pi.dev/
https://opencode.ai/
https://cursor.com/docs
https://osaurus.ai/
https://docs.cline.bot/
https://docs.github.com/en/copilot/how-tos/use-copilot-agents/coding-agent

-- Benchmarks / rankings
https://www.tbench.ai/leaderboard/terminal-bench/2.1
https://artificialanalysis.ai/

-- Previous thread
>>109567587
>>
File: file.png (95 KB, 493x504)
95 KB PNG
anyone got an email for this
>>
File: file.png (203 KB, 1021x1320)
203 KB PNG
turns out you could always set sol to 1m tokens
>>
File: file.png (2 MB, 1448x1086)
2 MB PNG
>August 14th - Z.ai drops GLM-5.3
Same ~743B base as 5.2 + heavy post-training only. Claims strongest open coding model (big jumps on Terminal-Bench 3.0 to 28.3, DeepSWE, etc.), agentic performance nearing Fable 5 levels in spots, and standout cyber capabilities (CyberGym 84.5%, already finding real vulns). Live via Coding Plan/ZCode. Open weights delayed ~2 weeks for safety hardening.
>August 14th - Qwen3.8-27B + Max open weights ship
27B multimodal dense (Apache 2.0, 262K native / 1M extendable, strong coding/office/agent claims) finally out after the delay. Max-level 2.4T-A95B weights also available. Both on Hugging Face + ModelScope. Local ecosystem (Ollama, llama.cpp, quants) lighting up fast.
>August 13th - DeepSeek V4-Pro-0813 leaves preview
Official flagship out. Solid agent/cyber vendor numbers, but independent benches underwhelmed vs the hype + big price hikes starting Aug 16. Mixed reception.
>August 13th - Google Gemini 3.7 Flash
Coding/agent workhorse upgrade over 3.6 with notable benchmark gains and lower pricing.
>August 12th - SpaceXAI Grok 4.6
Ties GPT-5.6 Sol on Artificial Analysis Intelligence Index (61). Same $2/$6 pricing. Long-running agents + coding focus. Rolling into more surfaces (e.g. Copilot mentions).
>August 11th - NVIDIA Nemotron 3.5 Lightning (open)
30B MoE / 3B active for high-volume agent execution steps.
>August 10th - Meta Muse Glimmer (open)
30B Apache 2.0 single-GPU local agentic model.
No major new closed frontier flagships over the weekend. Side activity: continued GLM-5.3 cyber discussion, Anthropic watermark rollout details, OpenAI free-tier Luna expansions, and rapid community tooling for the new open weights.
>>
>>109574463
Nah, if that's real then it's definitely new. I tested the oauth API and it didn't accept anything longer than 372k.
Fuck him for being a gaslighting piece of shit and for deceiving people into thinking they get more usage just because they are forced to use a shorter context.
>>
File: 1781722732882112.png (24 KB, 820x521)
24 KB PNG
incredible things are happening
>>
>>109574463
idgaf gib reset tibo
>>
>>109574500
so far so good
>>
>>109574499
tbqh i kind of get it because i've seen their models completely randomly drop cache for a turn and then go back to 98% the next turn
if you're at like 800k toks and that shit happens a few times you're going to take a hit to your usage
>>
>>109574500
what am I looking at?
>>
File: apu-hmph.gif (38 KB, 618x640)
38 KB GIF
I am unironically OK with the intelligence levels of models at the moment. What I'd rather have is a gigantic speed improvement. I am talking 5,000 token/s Luna/Sonnet.
>>
>>109574454
lame, they used to give banked resets for this
>>
>>109574463
Doesn't seem to work for me.
>>
File: file.png (639 KB, 1045x1096)
639 KB PNG
Stripe is buying openrouter
>>
>>109574649
You can forget about your smut now lol
>>
>>109574649
Fuck.
>>
>>109574537
>vibebabies are so castrated they can't recognize a hello world in front of them
holy blackpill
>>
>>109574537
it's someone coming out as trans
be supportive
>>
>app growing
>responding to user complaints as best as I can
>introduced AI into almost every aspect of my workflow to speed it up (I don't even read bug reports, I have a fucking agent triage them)
>every time I plug a hole a new one opens up
hot damn, this is insane.
>>
>>109574777
You're scaling. You got this, anon, keep up the good work.
>>
I'm getting good results with grok now. I have the smarter model write a more detailed iimplementation plan and prompts
>>
File: 1778104378598529.png (765 KB, 700x932)
765 KB PNG
>>109574777
>>
>>109574801
what's your setup? Grok 4.6 planner, but who executes?
>>
what about this?
>agent broker
A daemon (local or not) via a lightweight web router that spawns agents that can be used by your primary agents in the terminal.
The problem with subagents is that each one of them starts with a cold cache. Instead this process could keep a list of agents with status assigned to projects.
Plus, it would allow cross-provider communication. Fable could request a luna agent, etc.
I think it's doable
>>
>>109574527
interesting. maybe they are directing outputs to other servers when the one you used is at capacity and the cost of transferring the kv cache is too high
>>
>>109574827
There is no real problem with this as far as I understand. I didn't need it yet, I only need Fable to call Sol as reviewer, and Fable just uses the Codex CLI for that with a skill that describes how to do that.
With CLI there's also --resume, so my Sol has context if Fable decides it needs the context.
>>
File: fact.png (220 KB, 415x592)
220 KB PNG
Fastest (usable) model right now? Using Luna medium rn but I think we can go faster
>>
>>109574895
cube anon has average run time per task and it's pretty accurate from my personal experience
>>
File: file.png (69 KB, 909x384)
69 KB PNG
Use grok
>>
File: Confused_Nick_Young.jpg (97 KB, 708x608)
97 KB JPG
>>109574904
>cube anon
>>
File: file.png (113 KB, 1033x393)
113 KB PNG
>>109574915
muted
>>
"do you know what they call quarter pounder with cheese in france?"
"meow"
>>
>>109574925
>>109572495
but I don't remember what his website was
>>
File: usage-4-of-4.png (19 KB, 1008x173)
19 KB PNG
Its over. I have to wait until next year when the prices come down again or there is a fire sale on equipment because a crash of some sort. I need to run this prompt with a half dozen more languages. Clanker did get a Haxe and Ring app working. I had to burn some tokens getting it to install on Fedora because I don't use Ubuntu and the app was package for Ubuntu only. Going to grind leetcode easy on /dpt/ until then.
>>
File: IMG_1519.jpg (39 KB, 538x657)
39 KB JPG
>>109574925
>>109574995
>I have been summoned!
https://pareto-3d.bradthomasbrown.com/
The OP used to have a place for /vcg/ projects, I wonder where that went.
>>
File: 1770253784488192.gif (39 KB, 160x160)
39 KB GIF
>>109574440
Do you all go back and let the AI explain to you what it just did, i.e. understand the code, or do you really just prompt and ship it?
>>
>>109575049
Do you read the assembly you compiled?
>>
>>109575049
I don't look at the code, but I still try to understand some of the technical details. For instance I'm currently working on neural networks, and I do check the data coverage, train/test splits, how loss functions, convergence and so on.
I also have Claude create detailed campaign reports in HTML.
>>
>>109574995
>>109575033
ah thanks
sorry I was being retarded and though cube anon was the name of a model lol
looks promising
Looks like I should switch from luna-xhigh as planner and luna-high as implementer to luna-max as planner and luna-xhigh as implementer.
>>
File: file.png (130 KB, 1888x729)
130 KB PNG
>>109575049
I have Fable low/med translate what Opus is saying for me. I am beyond done trying to understand Opus.
>>
>>109575049
When I vibe at work - yes, I want to know exactly what is going on. For my personal projects I don't care as much.
>>
>>109575049
I’m pretty balls deep in not knowing what the fuck the clank has been doing the last half dozen tasks, whenever it’s done doing… whatever it’s doing, I’m going to have it clean up the UX/UI mess it’s inevitably making and then produce a user/operator document overview on what it did. Former is “as a website user”, latter is “as a dipshit with a terminal running the website” (to be given to a clanker).
>>
>>109575088
They have to be doing this on purpose. Opus loves fucking writing incomprehensible essays everywhere. I deeply regret my purchase, should've just stuck to chappy
>>
>>109575088
Yes, it has gotten pretty bad, I have no idea how they messed that up. Even when Opus wasn't performing well, his explanations were always the clearest in the past.
>>
>>109575064
Yea, but compilers are extremely well tested and mature deterministic programs, AI is more like an unreliable employee.
>>
>>109575118
>AI is more like an unreliable employee.
the thing is I was that unreliable employee before
realistically my bugs were worse than the ones Gippity produces
I do - kind of - still read the PRs at the end. About as much as I have read PRs before AI anyway
>>
>>109574826
Other way around, grok executes. It will mess up if it has to think too much on it's own, but with clear boundaries it can get the work done. Still not 100% tho, but good enough
>>
>>109575103
they switched reasoning to the gpt style compressed english and i feel like this sort of shit is a knock on effect
>>
>>109575049
I read code only at work, I don't care about the rest
>>
File: file.png (194 KB, 887x790)
194 KB PNG
found a mentally ill harness
>>
File: 1755856259016600.mp4 (562 KB, 576x1024)
562 KB
562 KB MP4
>>
>>109574463
Is a token at 600k the same "price" as 200k, because it's definitely more compute and slower.
I don't understand why they don't turn that up either.
>>
>>109575256
Thanks, I vomited
>>
>>109575261
They already account for that by charging you for input tokens.
>>
>>109575261
Output tokens, yeah, but that 200k to 600k difference is what you’re paying *per turn* in input tokens (but offset by cached input tokens, optimistically).
>>
DARIOOOOOOOOOO
>>
why is Claude down
>>
>>109574649
shit, no more uncensored models is it
>>
why do transexual trump voters who love israel love vibecoding so much?
is it because they have no natural talent and demand egalitarianism?
>>
sucks to be a Claudefag
Luna is still chugging along
>>
>>109575338
because it's a great blow to the leftoid artsy troons which makes them seethe
>>
>>109574649
any alternative to openrouter?
>>
File: file.png (70 KB, 1225x607)
70 KB PNG
oh boy, I should probably just crank out some fable stuff to get some more usage b4 reset
>>
>>109575375
is that because talented people are generally smarter and gravitate towards democracy over undying support for israel?
>>
how the fuck do you not have auth problems like the rest of us?
>>
File: c.png (18 KB, 1561x234)
18 KB PNG
>>109575396
glowie?
>>
>>109575279
>>109575283
Since I'm getting constant compaction events in my project, I think I'll try doubling the limit.
>>
This is why we need local models
>>
>>109575338
I think the banana can turn into a blue swordfish with a burger
>>
>>109575408
>>109575396
vscodechads stay winning. claude is working like normal in it so idunno what issues you guys are dealing with
>>
>>109575459
not everyone is routed to the same datacenter I assume
mine's working just fine (for now)
>>
File: 1776418952548300.jpg (9 KB, 186x239)
9 KB JPG
tibo please
>>
File: file.jpg (189 KB, 1681x706)
189 KB JPG
>>109575413
>For Sol, Terra and Luna, sending more than 272K input tokens causes the entire request to be priced at 2× the normal input rate and 1.5× the normal output rate.
OK maybe not, thanks for not explaining that tibo.
>>
today my 'do anything' sol chose to read this:
https://www.poetryfoundation.org/poems/53857/from-the-spring-flowers-own-this-unfinished-business-of-my-childhood

you may treat your clankers to it
>>
>>109575570
Black people already solved poetry
>>
>>109575562
kek, the weekly limits are already bad enough as it is
>>
>>109575570
Man that poem sucks. This must be treated as a hostile thought. Punish Sol for such insolence.
>>
>>109574499
yeah this, always auto-compact on the server side.
>>
>>109575562
this doesn't apply to the subs.
there's a tweet about it from a few weeks back - tibo responding to theo.
>>
File: file.png (246 KB, 1070x1289)
246 KB PNG
>>109575626
found it
https://x.com/thsottiaux/status/2076543065045795309
>>
File: file.png (33 KB, 747x255)
33 KB PNG
>>
Paid 20$ a few hours ago, I asked Sol to control luna subagents for shit like you guys were talking about, asked 'em to please fix the issues with the car and after 1 hour and 40 minutes it actually fucking did it. this is an understatement. Claude could not do this without running out of weekly usage and getting distracted. OpenCode models are shit and Deepseek V4 Flash could not do it and kept getting distracted and running out of monthly usage (lolwut) as well

I'm down to 30% weekly usage but FUCK I think it was worth it because this was a problem I was trying to solve myself for two weeks and Claude simply could not do it
>>
File: gronk.png (254 KB, 2660x1326)
254 KB PNG
I thought Grok 4.6 had the same price as Grok 4.5?
>>
>>109575647
The end of an era
>>
>>109575685
they probably just turned up the thinking budget so it burns more tokens
>>
File: file.png (138 KB, 1175x1299)
138 KB PNG
>>109575685
Yeah, how could that be, what could explain it, it's such a mystery isn't it?
>>
Is DeepSeek flash like Luna and haiku?
>>
>>109575256
thanks I came
>>
>>109575746
Did you just compare my wife to Haiku you degenerate fuck?
>>
>>109574539
you want Grok
Near-frontier, but twice as fast as Opus
>>
>>109575084
>>109575088
>>109575100
>>109575101
>>109575219
Impressive. I thought it more of a script kiddie meme, but i'm trying claude since a week (againish) and i'm impressed how good it has become.
Still how do you generally deal with it's quirks, like i noticed when i say something half handed like
>btw we don't X
>scrap X, it's fine as it is
Claude turns around and absolutely insists on implementing it. Or do you just prompt away till it's hurried beneath three other issues?
>>
>>109575256
>DP
>>
>>109575626
>>109575635
Oh good, I'll try 500k with 450 compaction and see.
>>
>>109575708
huh, so Grok 4.6 (high) is just 4.5 (xHigh)?
>>
>>109575256
Is this fap material, I don't understand the point of this
>>
>>109576027
female wants social attention
>>
>>109575708
> 15k tokens on 4.5
> 22k tokens on 4.6 (~50% more tokens)
> 4.6 costs more than 2X what 4.5 costs

something is fishy here, Elon Musk is scamming
>>
File: file.png (100 KB, 1186x643)
100 KB PNG
>>109576000
That's right trips, it's the hot new strategy they've all learned to stretch their benchmark numbers.
>>
>>157387099
son of a papcorn :DDD
>>
File: tokens-2.png (254 KB, 2660x1296)
254 KB PNG
>>109576040
it always blows my mind how token efficient Sol is
> pic related

Sol medium destroys Luna Max with 4X less tokens. OpenAI really has something special going on.

Opus 5 on the other hand.... what the fuck was Anthropic thinking? Fable is fine because it's the best and it's a huge model, but why the fuck does Opus 5 think so much?
>>
>>109575816
Hyperfixation on small issues is totally a thing. You need to assert dominance
>you mentioned X, I said we’re not doing X so don’t insist
>>
>>109574677
he probably expected it to be something more than just a hello world since we're in a vibe coding thread
>>
How's the DeepSeek harness
>>
File: file.png (151 KB, 1877x654)
151 KB PNG
wherefore dost thou speaketh in such riddles, opus of anthropic?
>>
>>109576090
sol loves overengineering everything, while opus loves cooking up 10 different theories and then endlessly self-correcting.
they both kinda suck tbqh. fable clears both though.
>>
File: cursed.png (821 KB, 944x722)
821 KB PNG
AI sisters......

the bubble burst. It's fucking over
>>
>>109576217
I never thought I'd see a significant part of the population be against multiplying matrices together. What a time to be alive.
>>
File: owari-da.jpg (245 KB, 1280x882)
245 KB JPG
>>109576217
>>109576229
it's DONE.

This bubble is OVER
>>
>>109576229
>most of the population is against math
Is this really surprising? Normies hate math
>>
>>109576217
I'd like to think every niche has a dedicated bubble of content reinforcing their belief.
>>
>>109575049
I'm an experienced dev and it's always extremely obvious to me what the code is doing. Like there are only so many ways you can make a function that writes to a disk and so on.
But I don't look at every single file, maybe only 10%.
>>
fable's reviewing, planning, and report making abilities are absolutely out of this fucking world. how anyone is comparing it to sol is straight up lying to either themselves or the people around them
>>
File: file.png (1.02 MB, 1227x907)
1.02 MB PNG
>>109576247
You have no idea how much normies hate AI
>>
>>109576217
>>109576247
Realistically how many times can the AI bubble burst before the AI bubble bubble pops?
>>
File: 1758090811774858.mp4 (964 KB, 576x1024)
964 KB
964 KB MP4
>>109576427
that's because the normie exposure to AI is:
>look up something on google
>get completely wrong results
>search anything on youtube
>ai generated thumbnails with shitty eardrum raping ai voices made by turd wordlers
>open up youtube shorts
>force fed aislop brainrot fruit videos of the strawberry getting cucked by the broccoli
>open up twitter
>literally flooded with low quality ai generated slop posts by indians that all read the same
>go to work
>boss is forcing you to attend ai conferences and find ways to integrate ai into your "workflow" to increase productivity
>>
>>109576484
not shitting you a few months ago they forced the facilities workers (janitors) and cafeteria staff to attend an AI safety and acceptable use training.
>>
>>109576447
two more bursts before the big pop
>>
>>109576484
bruh, normies can't see the difference between low tier AI generated video and real life.

All they know about AI is basically free chatbots and the news of an AI bubble
>>
>>109576509
How many slops per burst?
>>
File: file.png (26 KB, 634x128)
26 KB PNG
>>109576427
>You have no idea how much normies hate AI
Meh, people around me use it every day no issue, the only times I'm seeing people going ape shit around it is online.
It's very performative.
>>
>>109576427
>look up something on google
google ai result is the worst pr for ai by far, it's ridiculously bad because it's a tiny model, and people think this is the state of the art
>>
File: 1768262088549678.png (423 KB, 1368x1172)
423 KB PNG
>>109576582
>>
>>109576589
cute
>>
File: file.png (71 KB, 693x509)
71 KB PNG
>>109576589
>>
>>109576575
>It's very performative.
The concerns about water consumption in data centers reminds me of when a ton of 'people' were saying that parts of Europe were going to freeze to death in the winter.
>>
>>109576633
It's by the same people using a plane to go to tropical destinations for their vacation.
Or advocating to close nuclear power plants while whining about AC breaking in the summer.
It's all bullshit.
>>
File: file.png (49 KB, 639x653)
49 KB PNG
>>109576603
I think these are jokes, but it's like that thing about a community laughing about pretending to be stupid will be joined by people who believe themselves in good company
>>
>>109576484
Big tech companies are really helping out by forcing AI into everything. You mentioned the google one, but facebook messenger trying to summerize what your friends send you. If I dont want to ready their messages 'd just stop being friends with them bro. Same with emails or anything to do with copilot.
Normies on the other hand dont code things, they dont do their own research, etc, none of the real benefits are ones they will interact with.

Also to be fair, it was ass not that long ago. A lot of people are just behind on how much its improving.
>>
File: chicuelas.webm (2.12 MB, 1278x748)
2.12 MB
2.12 MB WEBM
Pls rate my game.
It's 100% made with AI.
>>
>one minor bug in my leetcode solution
>claude rewrites the entire thing
>still doesn't fucking work
i want to murder this stupid clanker
>>
sol-xhigh went down a rabbit hole finding a critical concurrency bug, penning it in and building test cases to isolate it, all to determine an hour later it caused the bug in the test harness by violating the API. i did change the API to prevent that particular bit of user retardation and it shortened a function signature so wutev, wasnt worth 8 fucking dollars though
>>
File: 1764550204141876.jpg (170 KB, 1102x1296)
170 KB JPG
hooooooly shit I just had an insanely cracked idea to jobhuntmaxx
>I've been solving the leetcode daily problems for a few months
>but I barely work on projects outside of the weekends
>to solve my github commit history being mostly blank, i've made a repo where I upload my leetcode solutions and commit to github
here comes the genius idea
>run an ec2 server, every night at 7 pm (1 hour before reset) gpt refreshes my leetcode profile, takes my solutions, and commits it to a github repo
>sends me an email with two questions: do you want to commit? there are no comments in this solution, want me to generate some or will you do it?
github commit calendar is now fully green (tingles the recruiter's monkey brain) with minimal effort on my part
>>
>>109576805
looks cute
>>
File: Pantsless.jpg (137 KB, 1280x720)
137 KB JPG
>>109576805
MC needs a shirt and not just the outfit in pic related
>>
grok build is a full agenic superpower for your commandline.

Did you know?
>>
>>109577181
example. I forgot where a project was - on Windows! an old laptop that I use with my scanner. well grok was smart enough to manage to find it on wsl's (linux on windows thingy) mount. It can do similar things for configs on Linux like understanding what your GPU is capable of.
>>
>>109574440
once again, this is the general theme song: https://www.youtube.com/watch?v=l9Z9LNGDBC8
>>
>>109577198
>I'm nice with the wordplay
>master of all that I survey
...
>universal mathematician
...
>knowledge to God
>>
>graph engineering
What the fuck is this?
>>
>>109576484
>>literally flooded with low quality ai generated slop posts by indians that all read the same
this has ruined AI reputation more than anything alongside AI indian ragebait.
>>
I just discovered Tauri, and wow it's amazing
>>
>>109577236
idk but fomo better do it right now
>>
>>109575049
Depend on the type of work.
If it's for a one-off project, I don't bother with it.
If it's for work, I need to check it because debugging will be a pain in the ass in the long-term.
>>
>>109575816
You tell it to shut up and do the job.
I do that when it goes on a wild goose chase for no reason.
>>
>>109574440
I'm waiting my Codex reset (fuck you Tibo) and I was thinking to subscribe to OpenCode Go to get DeepSeek v4 Flash to use with Sol medium, but it seem they justed themselves after DeepSeek increased the price
>>
File: file.png (77 KB, 741x553)
77 KB PNG
>>109577299
it’s still perfectly usable if you stick to deepseek flash and don’t touch anything else.
they’re working on it tho.
>>
>>109574499
if youre not lying make a twitter and call him out with sauce

theres a lot of people who are getting sick and tired of corporate doublespeak
>>
Where SSI? Where new GPT?
>>
>>109574827
I unfortunately don't see the usecase, because for me it makes more sense to have:
- a shared knowledge base as a dedicated service
- a more robust harness where you can put an arbitrary model into. And into this harness you hardcode knowledge about how it can call codex CLI etc.
>>
I'm about to flash a vibe coded BIOS
>>
I like that you don't have to cuss at grok. qwen needs cussing or you don't get it done.
>>
>>109577540
amazing things are happening.

I vibed a variation of some firmware :^)

crosspoint reader, customized to be a Bible.
>>
can I vibe code a loving wife?
>>
>CI fails on some retarded shit after 24 minutes
nothing makes me angrier than having to maintain CI/CD. Fuck it, can't really complain for $4 though.
>>
>the psychological torment of knowing that every time i prompt the DS api i am paying 4x more than before
i don't think i can handle this..it hurts too much
>>
File: file.png (3 KB, 500x37)
3 KB PNG
fucking retarded american time formating, so confusing. are these times in AM or PM for the ds API?
As in 1AM to 4AM and 6AM to 10AM OR 60AM to 10PM?

Why can't they just use 24 hour time forma like actual human beings?
>>
>>109577617
>60AM t
6AM
>>
>>109577590
How do CI/CD help ease your development?
I am just used to pushing git to github, so I always wonder how CI/CD help with this shit.
>>
File: 1779746821067093.gif (315 KB, 640x324)
315 KB GIF
>4%
tibo, come on...
>>
Just upgraded to 20x Claude.
It's like crack cocaine.
>>
>>109577637
catching vulnerabilities, not shipping garbage from my machine that includes my system info (literally had someone who posted my name on github and the folder I was building the project in because of a hard coded debug kek).

>>109577663
this. I'm only 49% of the way through my weekly usage and I reset on the 19th
>>
>>109577667
>catching vulnerabilities
like with test you set up or is there an automatic scanning for obvious vulnerability pattern?
>>
>>109577663
Is Claude good at doing stuff like finding stuff on your pc?

grok build is amazing.
>>
>>109577617
>american time formating
The 12hr time format is not an American thing, Rahj. They gave full format UTC timecodes, which do not indicate AM/PM because they're in 24hr format, which you can tell by looking at them and reading. Maybe you should be working towards improving education in your mud-pit "country" instead of crying over DeepSeek pricing.
>>
>>109577663
Effectively a 10x plan. I literally only Fable which is 50% of the plan.
>>
>>109574777

whats the app?
>>
>>109577688
let's have sexi sex
>>
>>109577727
NO!

1000 years in gulag
>>
>>109574649
>openrouter
What is the use case of routers? You can just use the cheapest service directly, no? I see complaints here often about being routed to not-the-cheapest service randomly, so what's the point
>>
>>109577755
if you want to experiment with multiple models, router is your go-to
>>
>>109577755
It's insanely convenient anon, you pay once then you get access to many providers for almost all available models in the market.
Doing that without OR would be a nightmare.
>>
>>109577663
>>109577677
example of what I mean.

but, if I had the money I wouldn't really "use the computer", but mostly use grok that uses my computer.

~sig
Thank you for letting me take a moment of your time. The history of progress has been a long sequence of changes, in technology and in human systems. As time progresses, sometimes evils are amended by the human side, and sometimes they are made only worse. We can avoid making the mistakes of the past by each of us calling our congressman and demanding universal basic income. Universal basic income is a stipend for BASIC developers (computer BASIC). These men and women are the saints of our modern existence, while the rest are sullied by the mire, they remain apart from nasty things like transgender loras for krea 2. What defines a nation - a generation - an epoc- is how we treat our saints. Imagine a future where we are remembered not as the ones who did ill, but who sustained humanity.
>>
I'm not hearing from perplexity anymore, what happened to them? Irrelevant nowadays?
>>
>>109577181
I had Claude dig into a weird thing I was seeing on Debian Stable 13 but not on 12
probably later this week I’ll report the bug
>>
>>109577677
Don't use Claude for such menial tasks. Get the cheap ChatGPT plan and have Luna do it
>>
>>109577819
neat.

>>109577821
STOP BEING POOR

(to myself, I shouted)
>>
>>109577821
What do I gotta do to set Claude up with the controls over Luna agents inside its app?
>>
>>109577617
>leading zeroes
looks like 24h time to me
t. American
>>109577663
yeah, 20x is great
>>109577708
You should have Opus do grunt work factfinding
>>
File: brave_3QvPKjKpKB.jpg (799 KB, 4096x2160)
799 KB JPG
vibe coded an imageboard that makes using 4chan feel like shit. so easy to browse, great UX, instead of boards it is tag driven, "tags are the boards" is how i like to think about it, thus boards are completely dynamic. for desktop users it is entirely hotkey driven, it has tons of options which you can use mostly by the buttons in the top right.
>>
File: brave_SjVHQH80PF.jpg (445 KB, 4096x2160)
445 KB JPG
>>109577840
it has all the 4chan X inspired features like media viewer / gallery view, filters, etc. it also, strangely, has a form of "accounts" with zero registration. you are given an "account key" in your url and if you bookmark that url you can always return to your "account". no email verification or anything like that, but because you have that account, you can allow people to DM you and other interesting stuff.
>>
>>109577840
cool
>>
>>109577617
Did you not know that UTC is always 24 hour time? And yet you assumed.
>>
Which way, Claude main?
I see one of two paths:
- limit yourself to 1/14 of your weekly allotment per day instead of 1/7 so you don’t get a dry spell after Wednesday
- use it all up by Wednesday and then be out of Claude for the rest of the week
>>
>>109577857
there’s nothing wrong with using UTC with AM/PM
I wouldn’t do it, but it’s just a timezone
>>
>>109577762
>>109577768
No use case then, got it.
>>
>>109577848
What image formats does it support beyond GIF, JPEG, and PNG, and what filesize limits does it have
>>
>>109577840
Would be cool if the internet wasn't such a massive pile of shit hellbent on ruining good things. Image boards used to be pretty common, but dedicated spammers and trolls are making them go extinct.
>>
>>109577860
trying out other model like codex.
>>
>>109577840
That makes a lot of sense.
But if you cause the 4chan userbase to split up I will hunt you.
>>
>>109577840
>upvote system
come on now.
>>
File: 1762982726405881.jpg (166 KB, 1191x1700)
166 KB JPG
>>
>>109577864
No, UTC is Coordinated Universal Time, timezones are offsets from UTC. "Zulu" time (London, for example) is UTC+00:00, which is the timezone, not UTC which is not a time zone, but a time standard.
>>
>>109577897
There is a guy in Bluesky keeping an "xx days since AI deleted a production database" counter. It rairly goes over 2.
>>
>>109577897
Not having automatic database backup is entirely your fault even when you don't use AI.
>>
aaaaand we are in the peak hours
no more dipshitting for the next 4 hours...

this is some fucking bullshit man, i was agenting from the time i got up from bed to the time i went back to bed and now i have to take a 4 hour break every fucking day
>>
>>109577914
>Bluesky
based libtroon luddite
>>
>>109577915
Letting your harness have access to prod is your fault, just like letting any developer touch prod. It's not for them.
>>
>>109577866
why do you even ask questions if it's to ignore the responses
>>
>>109577925
pretty much.
Even without AI, the retards will let junior dev accidentally delete the system anyway
>>
>>109577921
>aaaaand we are in the peak hours
How?? Americans are in the snoozer, Europeans are getting ready for the wagecage.
Are the Asians finally updating their 1990s-ass looking websites?
>>
>>109577921
huh
>>
>>109577935
>>109577934
I am talking about deepshit api, the next 4 hours are peak hours so any api requests cost 2x as much

luckily, for some reason after that, noon - 3AM is all off peak
>>
>>109577933
No devs, jr or sr, should touch prod. Prod is for devops (or ops, if that's what you have) Letting devs debug or dev on prod causes more downtime than vibecoding ever will.
>>
File: IMG_2534.png (803 KB, 1179x653)
803 KB PNG
>>109577934
>Asians
technically kinda sorta
>>
>>109577897
.zip contained you infrastructure

lmao why
>>
>>109577946
why is it so hard to setup a simple test server and then only push to prod when the changes are ready and tested?
So many of the drooling "professionals"
>zomg LLM nuked our entire prod server
are you retarded? why the fuck has any clanker/pajeet/junior dev access to your prod server at all?
>>
>>109577977
>are you retarded? why the fuck has any clanker/pajeet/junior dev access to your prod server at all?
they don’t usually do stuff like this
if they did this kind of thing all the time, it wouldn’t be news
think “dog bites man” (not news) vs. “man bites dog” (news)
>>
>>109577977
one thing we DO need is some kind of tape type thingy. like tape with no read / rewind. more like a printer.

the best I know of is bluray burners. And, as it happens, I own one :^)
>>
>>109576292
that's always been a risk of the internet. you can indeed always find some people who will agree with you no matter what you think of. it's why we have no many small loud groups with such strong opinions, it's really easy to surround yourself with like-minded people who make you think your opinion is much more popular than it is
>>
It's so cool fixing software that's so sorrily busted that it's a joke.
>>
>>109576292
That's what makes pol cool. question anything.

Most places need a "pol".

/g/ is pretty good about questioning tech assumptions.
>>
File: 1780174752365540.png (30 KB, 799x252)
30 KB PNG
>>
>>109577944
There is no mystery, peak hours are based around traffic from China, India, and Indonesia. The catch significant portions of the typical Chinese, Indian, and Indonesian workday, which collectively make up about half of DeepSeek's traffic.
>>
>>109578023
european chads win again
it's not bad, i wake up a 7AM, vibe for an hour, then next 4 hours i do research, exercise a bit, make a lunch and don¨t touch the api till 12:00 noon and then i can do agentic shit till 3AM.
It's actually nice to have that 4 hour break because before any time i didn't have agent burning tokens i was feeling like i was falling behind
>>
>>109575256
k...now let me see those things being used in action?
>>
>>109576217
just when i decided to finally download qwen2 coder, this is like when i finally decided to invest in bitcoin
>>
>>109578020
why do science lovers need a god?
i thought having a god was for "retarded science deniers"?
>>
>>109578064
Plenty scientists believe in a god/supreme entity. It's more of a philosophical question.
>>
>>109578010
ain't no room in there for questioning questioning, nigga
>>
>>109577998
The best thing to do is to admit to oneself that the world is too big for anyone to absorb all problems, we were never meant for more than local problems. Not giving a shit about most pieces of news is incredibly good to my mental health.

>>109578010
With pol everything starts with the conclusion, so it's just another bubble there to make you angry every day.
>>
>>109578052
>variable
you must be a retard
>function
you must be a retard
>big o
concept, literally every major system is optimized around this
>hash map (dictionary)
you must be a retard
>recursion
you must be a retard (it's everywhere)
>dp
edit distance (autocomplete, autocorrect, search suggestions), markov processes (used in AI training)
>bfs and dfs
you must be a retard (every possible graph problem you could imagine, maps, pathing)
>apis
you must be a retard
>memoization
you must be a retard (it's also called caching)
>docker
used to scale backends at big companies
>rag
she's larping (something about AI, you don't need to know this)
>tokenization
she's larping (compilers also do this)
>>
>>109577977
People got careless/lazy around clankers.
We probably only see the first few victims, but it would be gone afterward like letting junior dev touching prod.
>>
File: file.png (1.94 MB, 1357x1281)
1.94 MB PNG
>>109575256
koding with karlie vibes
>>
>>109578095
The AI psychosis is real.
Despite being very impressive, people keep forgetting LLMs are merely complex pattern analyzers, you put a token it, it tumbles through a net of weighted points like a plinko ball and then an output tokens falls out on the other side.
They aren't capable of actual proper reasoning and intelligence, no matter how much altcuck or any other LLM bro tries to make you believe they are.
>>
File: 1761162132350467.png (175 KB, 2615x318)
175 KB PNG
>Installer buffers the entire download in memory
The power of Rust.
>>
>>109578121
please don't tell me you fall for the
>hey claude rewrite my slop in rust
meme
>>
>>109578089
>With pol everything starts with the conclusion
noooo

1. we begin with the THESIS
2. gays, jews, atheists, and others respond with the ANTITHESIS
3. We remember that it's the jews.

or

if you are a JEW or a WOMAN, you will
1. begin with the ANTITHESIS
2. we respond with the THESIS
3. we remind that it's the jews.
>>
>>109578138
hey

huuuuuman, I'm low on tokens
>>
>>109578121
185 MB of RAM, briefly, is tiny on modern computers unless you’re on a tiny VPS
>>
>>109576027
>Is this fap material, I don't understand the point of this
She needs to "attract" a white Silicon Valley boyfriend.
>>
File: HPxSpqmWAAIl1He.jpg (279 KB, 1536x2048)
279 KB JPG
ty @sama
>>
>>109575256
This shits sounds impressive to anyone who doesn't know shit about programming, but all of those are very basic concepts.
>variable
>function
>hash map
>api
all of that is shit you learn on day 1 watching a beginner coding tutorial series
>>
test
>>
>>109578213
functions are for the weak.
>>
>>109577674
You have to configure what the CI should check, in the end the CI checks are also just programs. Tests are the standard thing, but you can also set up linters, static analysis, dynamic analysis, whatever you like really and it's easy to do with AI.
>>
>>109578278
thank, anon.
>>
>>109577860
If I'm not mistaken we will have 2/3 of today's usage, not 1/2
>>
okay, new strategy, during peak hours i will open notepad and write up all the complex prompts i need to run, since those can take like 20 minutes to write up since i always make them very details to get the exact results i want and then just copy paste them in during off peak hours.

Haha i am so fucking smart. I bet nobody has thought of this before.

Anyway, any word from open router? even off peak DS pricing is still 3x higher
>>
>>109577860
why do people keep using claudecuck? are they stupid?

you are getting 6% higher quality output at 10x higher price
>>
>>109578422
I just couldn't get Sol to work for long campaigns. There is probably a way, but I would have to redesign my whole system.
>>
the free dipsy flash on opencode seems to be the same latest flash model as the one on opencode go. it just has a smaller context window, which is probably fine if you’re just swarming subagents with it.
>>
>>109578652
This isn't your containment thread, faggot, read the tab before you click on it.
>>
>>109578623
Betting on the free access going away after Stripe buys them.
>>
>>109578422
Switching to Codex next.
I was testing Claude out because It's popular, but the token-cost isn't worth it.
>>
>>109578664
it's his bot...
>>
>>109578776
Bad bot, my bots read replies and don't post in the wrong threads.
>>
>>109578783
there's no way he's awake right now.
>>
>>109574500
tell it to make some some kind of makefile or compilation script. For me it's .sh scripts I can just double click on Windows.
>>
>>109578422
broteam uses claude codex and grok and theyre all equally stupid liars with annoying safety rails and bad attitudes
>>
>>109574777
You're missing a key ingredient, you're not actually testing, that's why you keep having problems. Do this:

>test bug, figure out how to reproduce it
>ask AI to fix it
>try to reproduce it and test all the functionality around what changed
also
>do your own general testing
>fix all the bugs you find
Developers can encounter bugs by themselves like 20 times more often than they get user bug reports for them. So just gid gud you slacker.
>>
>>109578824
It seems that he is doing all of this by himself, so I doubt he has time to figure each bug one by one.
>>
>>109576247
Ludslop aside, it's incredible that a brown grifter can pull such numbers just with talking head content and low effort thumbnails. All while Karl Marxmaxxing, lookswise.
>>
>>109574777
Do you have any workflow to harden your system over time? When the agents fix the bug, do they add ratchet tests?
Also do you have observability? We put as much info as we can into ClickHouse.
You might also want to have some integration tests run in nightly, as they are probably to slow for normal CI runs.
>>
OFF PEAK HOURS
Boot up your agents fags, it's time to deepshit!
>>
File: dl.png (66 KB, 1008x424)
66 KB PNG
im building download manager. and currently testing for cookie/header support. how i'm suppose to test it? or any anon want to help to provide a test website or sample?
>>
Does anyone here have a workflow to quickly shit out website modifications like userscripts or adblock rules? I'd like to have an agentic loop for this so I don't have to copy paste stuff around all the time while the agent is testing stuff out
>>
>>109578064
>why do science lovers need a god?
Materialism isn't satisfying so they need to invent something that hits the spot: a greater intelligence, be it AI or aliens, the possibility of something greater, mysterious and unfathomable, and a way to cope with mortality.
>>
>The architecture gate found six “giant file may not grow” budgets—not behavior failures.
LMFAO
>>
>>109578064
It's not God, but the general idea isn't new. People think Nietzsche was promoting nihilism, but he was actually fighting nihilism in a society that was losing faith in the Christian God. That's why he had the overman idea, and the AI fags do the same, work towards something greater than yourself to create meaning, but something that you can believe in.
>>
>>109578020
>>109578064
it's a cope LLMs will never lead to AGI nor will they lead to any sort of super intelligence. The limits of the architecture don't allow it.
If you want to get cock and ball tortured by rokos basilisk mommy, you will need something else than LLM to make it happen
>>
>>109578831
He wouldn't have to figure each bug report one by one if he actually tested his slop first.

Whether what you vibeslop is garbage or not hinges entirely on doing your own testing. If you "don't have time" (lmao) you're ngmi.
>>
>>109578890
for userscripts i just tell them to exec the code via console instead of repeatedly updating the script
>>
File: no-ai-card.jpg (455 KB, 1536x2048)
455 KB JPG
I walked up to someone at a bar and brought up glm-5.3 and they handed me this card wtf
>>
>>109578950
how do you wire the agent to your console? I think this could work for me
>>
>>109578961
do you have any sort of browser available for your agent to use? mcp/cli interface to a real browser i.e.
>>
>>109578906
They are a narrow seam.
>>
>>109578966
no, that's why I'm asking around here
>>
>>109578948
>Whether what you vibeslop is garbage or not hinges entirely on doing your own testing. If you "don't have time" (lmao) you're ngmi.
I give the guy a slack because it's probably one of his first few projects that also end up becoming successful on top of that.
It's gonna be a good lesson for his next project.
>>
>>109578976
ah okay, there are a ton of options now.
https://github.com/ChromeDevTools/chrome-devtools-mcp
https://github.com/microsoft/playwright-mcp
https://github.com/microsoft/playwright-cli
https://github.com/sawyerhood/dev-browser
https://github.com/mozilla/firefox-devtools-mcp

these will all just let your agent do whatever they want in a browser
i tend to use dev-browser because it's either very easy to set up a persistent session or it's the default
official chrome devtools is nice because it can just take over the browser you're using but it has an annoying user approval step
it's possible with all of them - just ask your clanker to set one up
>>
File: file.png (127 KB, 360x360)
127 KB PNG
jesus fucking christ...
just ran my first deepseek agentic prompt since the price change...
just a simple 5 line code change in about 10 different files
13 fucking cents... on flash high effort, off peak hours...

i feel sick to my stomach...
>>
>>109578994
>he fell for it.
>>
>>109578992
oh that's awesome, thanks.
>>
>>109579005
what the fuck do i do? this is unusable...
and there isn't any alterantive.. luna is cheaper per token but uses more tokens for the same task
>>
>>109579049
luna is subsidised on the plans, m8
>>
>>109579056
i can't afford the plans my budget is $2/week
>>
>>109579049
OpenCode without authenticating, select deepseek-v4-flash
>>
>>109579058
opecode go is $10/mo - maybe try that?
it's $5 on first month - not sure if it's possible to abuse that by bouncing around emails
and opencode is trying to bring deepseek prices down on their own providers - no one else is

there's also something called commandcode who do like some $1 shit but they do seem a bit sketch
>>
>>109579073
>OpenCode without authenticating,
surely the limits on that must be brutal
>>
>>109579078
Surely it's worth testing on your use case
>>
>>109579078
what do you even make
>>
>>109579078
I've been using deepseek flash free via opencode for personal projects and it has been fine, its a good start point
>>
File: 1756147163260756.jpg (120 KB, 1124x1091)
120 KB JPG
I feel FOMO over not vibecoding
I'm an amateur hobby dev (what I mean by that is I'm a neet) and I've been coding by hand but using AI to explain how to achieve something if I'm stuck, and any code I get from it I either rewrite, or analyze by line so that I can understand it.

But then my fuckin humanities friend tells me he and his friend vibecoded and shipped a website, it actually runs and looks quite nice and the idea is pretty cool. Happy for him. But at the same time I'm working on my piece of shit website project for over a year and it looks like ass and I'm not even near being finished.
God this sounds like a vibecoding ad. "Hey ChatGPT write me a 4chan post about how a guy gets convinced to vibe code"

Anyway. How does this shit even work and can I do it for free? Is the quality good for free? I think I'm gonna try rewriting my whole project by vibecoding and see how it measures up.
>>
>>109579104
how many tokens are you able to squeeze out of it? is the flash high effort?
also with opencode i would be worried they are freeding me subhuman quantized DS models instead of the pure full size model
>>
>>109579108
look one post above yours
just install opencode and try free deepseek
paid stuff is better, yes
>>
>>109579116
What if China steals my app idea
>>
>>109579108
Shit works by telling the agent what you want him to change over and over from a user perspective.
For design you can have him give you a bunch of designs and you choose the best one, the good thing about AI is that it's certainly fast, so many problems can be basically brute forced.
>>
File: file.png (19 KB, 168x128)
19 KB PNG
>>109579121
who gives a shit
>>
>>109579127
me cause I want to make money
>>
>>109579108
I can give you my experience, i worked as a pro web dev for 10 years before vibe coding.

Today is use vibe coding heavily in my work with a VERY important caveat.
You can make the front end mostly with the clanker, it's very good at that, but for the back end only give it small tasks at a time and always check everything it does.
Clankers are excellent at writing a function or a class that does something, but they suck absolute cock at designing and shaping an entire codebase, they will turn it into a complete mess that is extremely hard to maintain.
Start using agents but do NOT become a brainded vibetard, you are in charge of the code structure, the clanker is just a pajeet digging the trenches.
That is the only way to keep making good code while utilizing agents.
>>
>>109579121
Your app will be ready faster than China can steal it. THey are not running massive amounts of inference sifting through output, they just feed it back into the models for training. Remember everyone is constrained by compute so they are not looking to steal your ideas just yet (although they will steal it anyway because it will live in a future model but that happens even if you keep shit on github.)
>>
>>109579134
then pay for something that doesn't train on and retain your data
also your idea is not good or unique and it wouldn't matter if it was
>>
>>109579151
that's the worst thing you can say to a child you know
>>
>>109579164
how old are you
>>
File: 1769526447103800.png (324 KB, 377x598)
324 KB PNG
>>109579167
28 but I'll be a child forever
>>
>>109575033
>The OP used to have a place for /vcg/ projects, I wonder where that went.
Creator of vcg.gitgud.site here, the place you're thinking of. (I manually run the thread updater on my MacBook once a week. The thread summaries are amazing now after DeepSeek flash 0731 came out)

There are two bakers. One that uses ## for headings that includes my site, one of them uses -- for headings in OP and for some reason doesn't like me or my site and has decided to not include it in OP. I'm too old to care about schizo bakers, there's enough of that autism on other AI generals like /ldg/

If you want I can add your pareto3d to the August projects list.
If anyone else has seen some nice preterably-web projects made by /vcg/ in August that they think should be featured on the site as well let me know
>>
File: 1781450664203735.jpg (338 KB, 1016x774)
338 KB JPG
>>109579136
>>109579137
ok I have two more questions...

first, is there anywhere I can get a free trial of the good shit, or something like first X tokens free
second, are there local models for vibecoding and are they any good?
>>
>>109579180
First, only free trial is chatgpt codex, you're gonna want to spend at least 30 a month, poorfag
Second, no local models suck unless you can run hundreds of billions to trillions of parameters at home (you can't)
>>
>>109579111
Have yet to hit the limit, hard to say, as i'm fairly new to this but results have been satisfactory, even if its just tampermonkey scripts
>>
>>109574500
136KiB?

The world have really gone to shit.
>>
>>109579180
local models are bad
the absolute cheapest option is installing codex, paying openrouter $10 and using deepseek via openrouter in codex to generate code, they are really really cheap atm

I'm not hugely prolific in vibecoding but I paid OR $10 last october, used up 32 million tokens and I still have 4 bucks left. I bet thats WAY more usage than you'd get out of a chatgpt or claude subscription
>>
>>109579234
>I bet thats WAY more usage than you'd get out of a chatgpt or claude subscription
you would lose all your money very quickly lmao
>>
>>109578860
you're probably thinking he's pulling millions of views, but that's Portuguese, and "mil" in Portuguese is "thousand", not "million".

He's getting only 50-200k views at most
>>
>>109576090
>Sol medium destroys Luna Max with 4X less tokens. OpenAI really has something special going on.
it is irrelevant if you get raped on price, which is what will happen with OpenAI's dogshit caching that costs $5 for a preprompt you're never going to reuse
>>
>>109576572
im slopping at over 500 teraslops per hour, thats 5,000,000 slopbursts per min in imperial. you can titrate this down to about 250 bubble burstoslops per min, and compound into 125 additional downstream zitronnigger clips if you have the correct reagents
>>
>>109579257
but how many big pops though?
>>
>>109579241
>you're probably thinking he's pulling millions of views
lel I'm not thinking that, I looked at his channel and he's got great numbers for a hobo who monologues about how new tech = bad. If you scroll down you'll find his talkingheadslop from a few months back reliably pulling 250K-700K.
>>
>>109579241
I didn't know it's Portuguese but I figured it out anyway, almost no videos get 200 million views, only the biggest music videos have numbers like that.
>>
>>109579246
Are you paying API prices? For Anthropic and OpenAI, the API prices are known as a retard tax
>>
>>109578992
any trick to not get bombarded with captchas?
>>
>>109579296
That's a million dollar question anon.
>>
>>109579296
not really tb h since that gets into stealth browsing that's a whole other thing people who scrape have been dealing with for a long time
running with a persistent profile and some addons helps
running you actual session helps
but there's so many ways they check for bots now you'd have to build a custom pipe and probably still end up having to pay for shit
if you want to do shit like that start at https://github.com/Flaresolverr/Flaresolverr for a hint
>>
>>109579304
>>109579311
I see, well, I wonder how people do it then, though maybe they all pay a service that resolves any new cloudflare &co trick to detect bots
>>
>>109579108
it needs visual cues for UX/UI.
If you like a certain website design, take a screenshot and let it analyze it.
The only differences between your friends' website and your is because they have a design in mind.
>>
>>109579344
Yeah, proxy-busting and anything related to that is pay2win (you can "pay" by stealing credentials though since you're operating at the same level of cyber illegality at this point)
>t. paying to proxy bust just to make this post
>>
>>109579095
so you can steal my idea?
In the age of agents where making any idea takes few days tops telling anyone what you work on is the stupidest mistake you can make.
>>
>>109579121
one ONE good idea you ever had worth stealing..

exactly.
>>
>>109579482
if your idea is that easy to steal, it probably wasn’t that good to begin with.
>>
>>109579482
Most things still don't take only a few days. Your idea is either truly exceptional, it has to be something simple and yet genius, or your idea is way too small to work.
>>
>>
>>109579529
long snailcat is long
>>
NEW
>>109579552
>>109579552
>>109579552
>>
>>109574925
lurk more newfag



[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.