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


Thread archived.
You cannot reply anymore.


[Advertise on 4chan]


File: file.png (2.1 MB, 1448x1086)
2.1 MB PNG
A general for vibe coding, agentic engineering, coding agents, AI IDEs, browser builders, and shipping code with LLMs.

// What “vibe coding” is, and how to do it
https://simonwillison.net/2025/Mar/19/vibe-coding/
https://simonwillison.net/2025/Mar/11/using-llms-for-code/

----

// Frontier models using fully-general tooling — start here if you have $20 or so
https://claude.com/product/claude-code (Fable 5 is the best LLM available, requires Max plan)
https://developers.openai.com/codex/cli (Essentially scamming you with LLM degradation and resets that reduce your usage, but still the second best option and arguably the best bang for your buck in the 20$/month plan range)

// Worth it for code, but the frontier models above are better
https://x.ai/cli

// Not worth it for code, but maybe good for other things
https://antigravity.google/product/antigravity-cli

----

// 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 — /grilling is a favorite
https://github.com/DietrichGebert/ponytail

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

// UI/Frontend
https://www.figma.com/make/
https://www.anthropic.com/news/claude-design-anthropic-labs
https://uiverse.io/
https://ui-ux-pro-max-skill.nextlevelbuilder.io/
https://stitch.withgoogle.com/

// In-browser builders / hosted vibe tools
https://bolt.new/
https://replit.com/
https://docs.github.com/en/copilot/tutorials/spark
https://v0.app/docs

// Benchmarks / rankings
https://www.tbench.ai/leaderboard/terminal-bench/2.0

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

// Previous thread
>>109463778
>>
File: IMG_1482.jpg (103 KB, 1459x1179)
103 KB JPG
>>109469383
The clanker has baked a new cube!
https://bradthomasbrown.com/pareto-3d/

You can now switch to and from orthographic perspective, makes it so depth isn’t interrupting comparisons so much. You can disable model families and hide non Pareto points.

There are no paid offers or plans, no donation links, no ads, no monetization, no tracking. Data comes from DeepSWE, it is not an endorsement, the clanker picked the source.
>>
>>109469383
kek
>>
>/vcg/ News
>July 24th - Anthropic drops Claude Opus 5
Flagship for knowledge work/automation/coding/agents. Approaches Claude Fable 5 capability in a lot of categories at half the price ($5/$25 standard, Fast mode $10/$50). 1M context, effort dial, becomes default on Claude Max.
>July 26/27th - Moonshot ships full open weights for Kimi K3
2.8T-param MoE (largest open-weight ever at the time), 1M context, multimodal. Hosted version had already been live; weights hit Hugging Face under their license.
>July 27th - Alibaba quietly launches Qwen3.7-Flash
Cheap 1M-context multimodal (vision/reasoning/agent workloads), $0.03/$0.13. No big announcement or full report.
>July 30th - OpenAI slashes GPT-5.6 Luna 80% and Terra 20%
Luna now $0.20/$1.20, Terra $2/$12. Sol unchanged but gets Fast mode (up to ~2.5x speed). Efficiency gains (some from Sol optimizing its own stack) + competitive pressure.
>July 31st - DeepSeek promotes V4-Flash-0731 to production
Same 284B/13B active MoE + 1M context as the April preview, but heavy post-training upgrade. Beats their own larger V4-Pro-Preview on the agent/coding benches they published. Silent API upgrade, same $0.14/$0.28 pricing, MIT weights.
>August 3rd - Alibaba releases Qwen3.8-Max
2.4T total / ~95B active MoE, 1M context, multimodal. Hosted now ($2/$6), open weights (plus a 27B) promised the following week. Positions it as competitive with the closed frontier on coding/long-horizon/work tasks.
>>
>>109469432
just what we needed to make sense of numbers: an unscaled 3D chart with fixed size, unlabeled data points
>>
snailcat on my COCK call it snailcock
>>
File: 1768316964446852.png (74 KB, 500x500)
74 KB PNG
>v4-flash slower today
NORMGROIDS GET OFF MY DEEPSEEK API REEEEEE!!!
>>
>>109469523
You can tap a point for a label, if you try labeling them all you end up with a mess, you can orbit, zoom, and focus on points. Selecting a point opens a numeric information panel below the cube, there’s also a big chart at the bottom with all the numbers. There’s links to the raw data as well. I made it because I would rather have a cube than a handful of 2d charts and at least one other anon wanted a cube, so a cube we now have.
DeepSWE’s site has the same data, but lots of much nicer plots, but no cube.
>>
>>109469556
tell it to make an hypercube with a dimension for every axis
>>
>>109469574
My analyst clanker explicitly told me not to
I wanted a Zachtronics game like VM and my own benchmarks where clanks had to make programs as solutions to puzzles, then the clank logs, the puzzle category tags, and the solutions could all be used for a ton of benchmarking data
It told me to focus on the other ideas I had in the notes like orthographic projection, filtering, and point focus first.
Most data sources explicitly forbid you from using their data, too
>>
>>109469631
Wtf are you doing being pussy-whipped by a bot? Kick it harder and tell it to make the fucking hypercube.
>>
File: file.png (2 KB, 79x57)
2 KB PNG
Having a long-term agent with memory that you sit in its own VM to manage a money-making task is genuinely free as fuck. I have been running this shit for 5 days and this is only what I've paid out, I have $40+ waiting and it climbs by a dollar every time I check.

I was dubious on Hermes before and I am still a little, I could probably manage this with a better harness I made myself, but whatever.

So, how much are you guys making?
>>
>>109469661
Did you give it an open ended goal like "make money online" or did you specifically point it at different monetization options?
>>
>>109469708
No, I know what I'm doing and gave it a specific task to manage, the thing the agent is providing is handling it while I go play video games and turning my initial "why is it doing that"s into a full fix that I don't have to work for.
>>
>>109469645
I’m also the guy who invented killing the bots through “identity refreshes”. The only reason I agreed was because I really did agree, I disagreed with it and killed it on the other 90% it spat out.

I don’t have another axis to graph right now, if I were to pick it would be some estimate/inferred task complexity and the current INT stand in would be a more abstract INT value
Puzzle category tags could make it so much more than a hypercube, you could ask questions like “given an N INT task, an M INT model, what’s the time/cost/passrate/memory usage/cycle count/storage usage of solutions of control flow/looping tagged puzzles” etc.

The current graph (and data) does a poor job of conveying that some models simply can’t handle tasks that require too much INT.
>>
>>109469708
He's probably the pajeet with the youtube slop farm
>>
how did deepseek make this model? training on a lot more than known limit?
>>
>>109469708
Its so fucking cringe when people try to use ai like that
>>
>>109469740
By distilling K3 probably
>>
>>109469740
No idea but its surprisingly pleasant to work with. The reasoning logs are kind of interesting because it seems like it knows when its unsure and will fire off tool calls until it figures it out. Its very diligent.
>>
Random question: when the clanker queries the internet, is it doing it through your machine? If so, couldn’t you man-in-the-middle your own clanker instead of prefilling by having authoritative sources appear to claim whatever you want it to see?
>>
>>109469383
Cute snailcat
>>
>>109470020
Depends on the clanker and harness, but generally yes, it's doing it through your machine and yes you could do exactly what you're describing. Some people use a layer that prevents the clanker from freely searching online, instead popping up a request in your browser, showing what it's looking for and the results it's about to receive, and you can individually approve, deny, or alter the results.
>>
>Google fired Demis
AAAAHAHAHAHAHAHAHAAHAHAHAHAHAAAAA
>>
File: file.png (763 KB, 1019x1374)
763 KB PNG
holy shit google's actually finished
>>
AGI when?
>>
>>109470067
>yet another vague AI research startup for a quick buck
reminds me of all those soulless boomers releasing their stupid NFT collections
>>
>>109470067
oh wow 4 washed up millionaire boomers, whatever shall we do
>>
>>109470058
Imagine a little thing that just converts all licenses to open source versions, lmfao
>>
>>109470181
>I better look into this to ensure we're not stepping on any toes!
>US Law: Do what thou wilt shall be the whole of the law
>Sounds like it's no problem, I'll get to work on that "Penetration Test" right away!
>>
gemini 3.5 pro imminent
>>
>>109470299
it's going to be a nothing burger because demis already announced gemini 4
>>
File: 1767667021128140.jpg (67 KB, 1024x1024)
67 KB JPG
>3% left
reset pls
>>
Sometimes it's funny how the agents don't know what time it is. It may be 2pm and they will say stuff like:
>good session, tell me when you're ready to pick it up tomorrow
>this test is long and will block your machine, but lets just start it and finish over night
>>
File: file.png (3 KB, 438x98)
3 KB PNG
>>109470329
it's over for us
>>
File: 1777508010845762.png (228 KB, 1025x1042)
228 KB PNG
It's never been more over for google than it is today.
>>
File: file.png (543 KB, 1540x864)
543 KB PNG
>>109470160
>billionaires*
probably multibillionaires soon because they're going to get so much funding lmao
>>
File: 1778758338624238.png (103 KB, 1015x289)
103 KB PNG
CEO of HuggingFace throwing shade at Google.
>>
translated:

If you are working on open-source projects related to Agent Harness and wish to integrate and support DeepSeek Harness at the moment of its release, please reply with your GitHub ID and GitHub project address, including but not limited to plugins, skills, MCP, orchestrators, aggregators, UI, and more.

We will select some open-source project authors to invite for internal testing of DSH and gift a portion of API quota, allowing you to integrate and support it at the first moment of DSH's release.
>>
>>109470400
SAAAAAR YOU MUST BUY AN AD
>>
File: 1785318248157852.png (488 KB, 1021x770)
488 KB PNG
big
>>
>>109470394
Fun fact: he's a cocaine addict
>>
>>109470415
dariosissies...
>>
>>109470415
I'm sure Gemma 4.1 will finally put the US back in the game
>>
File: 1678745885879147.jpg (26 KB, 640x427)
26 KB JPG
https://www.youtube.com/watch?v=D1O1d9pRLmk

Second day of progress, chuds.
>>
File: file.png (428 KB, 1042x1042)
428 KB PNG
love having my thoughts read non-invasively by a model
>>
>>109469902
>>109469740
How good is deepseek v4 flash compared to Kimi-K3?


>>109470415
Means fuck all if they can't actually Force the American counterparts to not being fucking lazy. The American richfags are basically a bunch of spoiled rich kids that think they are above doing any actual work putting any actual effort into doing anything. Even if they're actually serious about making competitive US Open-Weights ai, it's simply won't happen because they literally don't want to.
>>
File: what the christ.png (15 KB, 188x95)
15 KB PNG
>>109470496
>>
>>109470496
What's cool about keyboards is they're almost as fast as thoughts, completely private, and only what you want to think at the model. But it's all a grift anyways, no one is making non-invasive (let alone invasive) mind reading tech in the next 20 years. But I am still waiting for Meta to release their mind reading bracelet that can read your electrical impulses to simulate keyboard strokes.
>>
>>109470497
not great
k3 is a pleasure to use, v4 is merely usable but hey it's super cheap so...
>>
File: 1757022828209630.jpg (302 KB, 2010x1334)
302 KB JPG
Name my band.
>>
>>109470555
How We Got Investors To Give Us Millions For A Quick AI Rugpull, & co.
>>
File: file.png (16 KB, 900x169)
16 KB PNG
>>
>>109470555
lynyrd skynyrd
>>
>>109470535
Fable is cheap too for us non poorfags
>>
>>109470532
Anon we've had working thought-to-text for a few years now. It still takes a fuckhuge fMRI setup so it's only "noninvasive" in the sense that they don't have to cut you open to do it, but it already works. Give it another decade it'll probably be terrifyingly good. Consumer tier shit though, even enterprise tier, I do think we're a long damn way from this being practical for any purpose at all other than playing with disabled people.
>>
>>109470555
is that one of those brahmin chads ive heard so much about
>>
>>109470415
>>109470497
>>109470462
>>109470426
Speaking of Open Source models:

https://www.axios.com/2026/08/04/trump-ai-framework-open-models

https://www.bloomberg.com/news/articles/2026-08-05/china-s-open-weight-models-to-be-spared-us-tests-us-firms-told

https://www.reuters.com/business/trump-advisers-tell-ai-firms-they-will-not-safety-test-open-weight-models-2026-08-04/
>>
>>109470532
>almost as fast as thoughts
ouch way to out yourself smoothbrain
>>
>>109470600
pretraining to read 5 words with a machine the size of a room isn't being close at all, this is like the 90s when they had 20 pound cameras strapped to the top of cars and saying we're getting self-driving soon. Feel free to look up the old Nova specials to remind yourself what hype vs reality looks like.
>>
>>109470614
stop hunting and pecking anon its embarrassing at your age
>>
>>109470614
Let me recompose: thoughts you want in actions are just about as fast as typing on a keyboard. What you're really outing yourself as is you're a 60kpm pecker. My point, the bottleneck is not typing on they keyboard. I can't say the same smartphone "people".
>>
>>109470611
that's funny dario wanted to be included so badly in US politics only to have US politics take chinas side.
>>
>>109470643
What's your WPM? If it's high, then you probably don't feel it that much, but still from what I read noone can type as fast as they can think. It might average out over the day to some degree, but you probably can't write down a single thought in real time.
>>
>>109470532
>almost as fast as thoughts
My keyboard can’t render a 4D hypercube and rotate it freely in my mind’s eye
Sounds like something is deficient with your brai— keyboard.
>>
>>109469383
I look like this.

:)
>>
>>109470664
Yes anon, but most people should spend time to compose their thoughts instead of throwing out a stream of consciousness. Not every thought should be typed.
>>
>>109470620
Good doesn't mean useful or practical. You're underselling it, are you not familiar with the state of things? Not seen Brain2Qwerty from Meta? I agree we're a long way from it being useful to anyone, but it already works.
>>
>>109469661
wow thats a lot of money ranjesh
>>
>>109470689
I assume the promise is I could be writing this with my mind and not a keyboard, which in practical terms, is very, very far away.
>>
>>109470696
Maybe if you focused less on assuming people weren't white you'd be making that much too.
>>
>>109470688
Overall probably true. When I was still programming myself, there were just a few situations where an infinite WPM would have been nice. Overall programming was never bottle-necked by typing, and thinking more was good, but just being able to run a few experiments in very short time would've been nice. Now AI can parallelize that very quickly anyway, so it hardly matters.
>>
>>109470599
I might be poor, but at least I'm not a psychopathic mentally ill piece of shit, unlike you.
>>
>>109470702
>making that much
$200
KEK
How long did it take you to amass that fortune? When I was 16, that’s one really good shift as a Sonic Drive-In carhop
>>
>>109470724
5 days of sitting on my ass and letting my agent manage it. It's good for passive income.
>>
>>109470716
Anon, that was a vitriolic response to such a passive comment, are you okay? Did you remember to take your medication today?
>>
…am i the only one ITT that hardly ever thinks in language and words?
nobody else thinks mostly in images and scenes and data structures, imagines music, what’s going on here?
If all of your thoughts or even more than 20% of your thoughts can even be output by a keyboard, I’m pretty sure you’re mentally retarded
>>
File: mystery nigga.png (6 KB, 246x78)
6 KB PNG
>>
>>109470702
yea ranjesh i sure hope i can make that much passively one day... lol
>>
>>109470750
Actually it's the opposite, language is structure and a sign of high IQ. Animals think in pictures.
>>
>>109470750
congrats on discovering thinking, youre like everyone else
sorry to burst your bubble, pseud champ. you do have to be 18 or older to use this site though
>>
File: 1776355192422199.png (48 KB, 1025x423)
48 KB PNG
deepseek flash literally deleted my entire web frontend for my ghetto perplexity app. had to go back to terra to get it to fix my shit. never china again
>>
>>109470748
No, it was not a passive comment. Only a piece of shit tries to makes others feel bad for being poor.
>>
>>109470779
>There is no git history
You are a bumbling buffoon and have no right to put the onus on DeepSeek Flash alone.
>>
>>109470779
the amount of fucking retards here who don't know git is impressive, it should be a warning in the OP pasta
>>
>>109470779
Imagine giving an AI full access to your computer and command line, also you should be using git tracking. The first thing you should do is have a harness that bans destructive actions.
>>
>>109470779
why no git?
>>
Making my agent use CVS over git just for emotional damage.
>>
>>109470767
>>109470774
Keyboard-thinkers MALDING ahaha
That’s what I’m going to call you guys from now on
Keyboard-thinkers.
Thinking so constrained it can be captured by what a keyboard can accomplish.
How embarrassing.
Can I get a quick check with Mr snailcat, are snailcats keyboard-thinkers?
>>
>>109470779
Aww man, that sucks. It seemed to be fairly dependable.
>>
>>109470779
i was just thinking of using deepseek yikes
>>
>>109470818
If your life isn't narrated in English by someone who sounds like Charles Dickens you're barely human.
>>
>>109470779
Hell, fuck git, why on earth did you make a write tool that has access to the very source code that made it, the only copy, then give that to a chinese clanker?
You made a lot of mistakes before not using git here. I bet you’re a keyboard-thinker.
>>
>>109470818
this guy doesn't use git in his brain
>>
>>109470838
>imagine in 2026 not having an inner monologue about your inner monologue

>but anon I see an apple, you don't just see pretty pictures inside your head
Yes anon, not only do I see an apple, but there's a full DVD commentary track for it.
>>
>>109470838
When Charles Dickens is narrating your life, do all of your other thoughts stop to make room? Keyboards can’t type two streams of thought at the same timetime.
>>
File: 1767995637970246.png (75 KB, 1020x258)
75 KB PNG
>>
>>109470820
Opus deleted our staging DB. Just today Opus randomly killed my 4 hour AWS data run.
That shit happens, you just need backups where the AI can't reach.
>>
>>109470871
No you fucking retard, that's the point, you don't understand because you're incapable of real thought. You're so busy picturing an apple it takes a minute for you to type the letter A. Frankly, it's like explaining literature to a dog, you're literally too stupid to understand because you have 95 IQ. The fact you managed to find yourself here is amazing.
>>
>>109470779
>ResearchDashboard.jsx (7,330 lines)
man i'm so glad i hopped on the train early when gemini couldn't reliably edit files longer than 500 lines and i basically have a tic now to refactor anything over 500-700lines
>>
>>109470064
>chasing random bullshit technology instead of ASI
I hate these fucking grifters. Once we achieve automated researchers none of this stuff is going to matter. You either believe in OpenAI's/Anthrophics mission or you don't. Dario is unironically right to gate keep the fuck out of his company.
>>
>>109470688
Maybe of you're writing a letter to someone, thats valuable. I would pay pretty good money for a device that can get my thoughts down into my journal faster than i type, and ESPECIALLY if it can be recorded hands free (no im not super extrovert so verbalizing my thoughts is even worse than writing.)
I think every thought you have should be written down, unless you don't think that much, there's way too much that comes and goes way too easily. Theres no reason not to get a stream of consciousness and then simply edit it, rather than deal with the extremely lossy method of trying to structure everything first

why people carry notebooks with them to jot down ideas, they know how easily a good one could be lost. god i wish i had something that could record my thoughts all the time
secure and password protected of course.
>>
>>109470904
>>109470818
its cringe that youre drumming it up now so you can look like you were exaggerating and joking all along when in reality you were being an incredibly cringe loser at the start
>>
>>109470924
meant for >>109470496
>>
>>109470881
stop posting your gay channel faggot
>>
File: file.png (1.98 MB, 1402x1122)
1.98 MB PNG
>>109470818
Snailcat grew up with touchscreens. Snailcat has seen keyboards before, in TikTok videos. Snailcat doesn't own a screwdriver. Snailcat calls a plumber if the toilet is clogged. Snailcat has a fridge full of Feastables and Prime energy drinks. Snailcat orders from GrubHub and UberEats daily because Snailcat cannot cook. Snailcat is lactose intolerant by choice. Snailcat thinks alcohol tastes bad. Snailcat likes Beats by Dre Limited-Edition Zyn pouches. Snailcat uses a screen reader app because it reads faster than Snailcat can. Snailcat is afraid of driving and doesn't see the point of it.
>>
webnigs are sleeping on using sonnet for implementations. faster and sometimes better than opus
>>
File: 1782160185038461.jpg (244 KB, 1920x1080)
244 KB JPG
>>109470951
>>
>>109470924
>>109470933
if you want to look at it charitably, a problem like this probably still needs a year or two of work post-asi: mostly around gathering data and the logistics of that. you can look at it as valuable work, or a way to make a buck once the asi lab buys your data out.
>>
>>109470957
>thinks alcohol tastes bad
Does it not? Some of it's alright, but most alcohol does taste bad.
>>
>>109471021
An attractive femsnale approaches Snailcat, asks if he'd like to get a drink some time. He responds,
>No way, alcohol is icky!
and slithers away briskly.
>>
>>109471040
There we go.
>>
>>109470973
>sometimes better
That’s the kicker, probably, and how complex does the web stuff get? Static stuff and tree shaped hierarchical components, it’s probably ideal, but I’d be nervous to start bussing data around in odd ways. Maybe sonnet for the UI, Fable low/Opus medium for the data?
>>
>>109471040
>An attractive femsnale
Proceed…
>>
>>109471088
see
>>109469383
>>
File: 1769750719296250.png (42 KB, 712x658)
42 KB PNG
fuck it, we ball
>>
File: 1770760271188759.jpg (104 KB, 1920x1080)
104 KB JPG
>>109471157
>>
>>109471157
I have to ask what you’re doing to need that. I’m basically vibecoding all day long with Sol XHigh and occasional Max and I don’t seem to need it. I’d need to set up my thing where I have several Sol’s streaming work to me simultaneously to feel like I need that.
Then again, even a decent game could necessitate that or anything where subagents are useful
I’m curious, tell more
>>
File: file.png (407 KB, 1021x1003)
407 KB PNG
how does he keep getting away with this lmao
incredible grift
>>
>>109471157
I'm having so much fun with vibing that I'm considering the 5x plan but I literally can't decide between claude and codex. I like both

>>109471189
on the $20 plan (I have $20 plan for both) I definitely feel constrained
>>
>>109471213
Oh, well, yeah, I’m on the $100. I’d obliterate the $20 plan. I’m juggling 3 to 4 projects, but really only focusing on 1 at a time, the current one is a simple static website with a little 3d renderer
>>
>>109471189
Yea, I'm vibe coding every day for 10 hours minimum with Sol high/xhigh and 5x only lasts me 2 days (Thank you Tibo for the rests, It put off this upgrade for an entire month). I'm launching soon so I need all the help I can get. I usually run 2-3 terminals at a time.
>>
Is Dario Amodei a gullible autist or a psychopath playing a gullible character?
>>
>>109471241
>Sol high/xhigh
Sol low + Luna subagents for long running tasks are enough though
>>
>>109471267
>enough
why should I use anything but the best?
>>
>>109471236
are you on the $100 codex or claude plan? I'm working on two separate websites with some fairly complex integrations and I have to carefully prompt around the $20 plan I have since I use it quickly
>>
>>109471275
>only lasts me 2 days
You answered your own question
>>
>>109471331
I am not whoever your replying to and didn't read that either.
I joined this thread 5 minutes ago.
>>
>>109471360
then don't reply to me faggot
>>
>>109471300
Codex, I started with API pricing on Claude like a goofball before I figured out it was a genuine retard tax. I’m very sour on Anthropic because their models argue I can’t make any wrapper for their harness by their own ToS, while OpenAI has something literally dedicated to making it easy to use and lets you use whatever harness you want (Codex App Server)
>>
>>109471379
I started with API pricing on Claude like a goofball too, only to figure that payments to the API have an expiration date after which they yoink whatever you paid but haven't used and reset your balance to zero. I'm not giving them any more money than whatever I plan to use immediately now.
>>
>>109471300
For real advice: $100 Codex does one Sol basically forever. $200 Claude does Opus basically forever. If you only run one prompt at a time for a work day. $100 Claude will make you ration.
>>
The codex resets have done a number on me. I still have two banked, but I keep telling myself that as soon as I use one, there will be a general reset for everyone. So I don't use them. I'm still out of usage though.
>>
File: file.png (38 KB, 1364x413)
38 KB PNG
>>
>>109471453
That's quite a difference
>>
>>109471414
>>109471379
alright I'll be a good goy and hop on the $100 codex plan. it's so hard to gauge which model works for what since everyone has such wildly varying opinions on the matter. I read the general here sometimes and it seems like people say using anything other than fable is a mistake. then I see people saying Sol only. both have worked for me but I can't properly judge it all as a noobie
>>
>>109471414
Anthropic is really sorely missing a lower end model that isn’t complete shit. Sonnet 5 is in a horrible state right now. Luna mogs the ever loving fuck out of it. Fable leans way too hard into high cost, too, anything but low is questionably expensive. Opus is the current god though, but it’s still quite expensive
>>
QRD on skills, MCPs and LSPs.
I have been basically using vanilla CC, codex and opencode.
Some people have extreme setups with all kinds of customization.
How much of it is useless snakeoil / voodoo magic and how much of it is actually useful?
>>
File: file.png (139 KB, 1920x790)
139 KB PNG
>>109471473
probably worth throwing a couple bucks at
>>
>>109471474
Opus 5 is good enough for me compared to Fable. I use it for game development, Sol sucks at game code and visual thinking, Claude gets it right most of the time although both models are extremely autistic in labeling and implementation.
>>
>>109471474
Refer to the cube!
>>109469432
Luna and Terra cover a huge range of capability from dipshit retard to really good, at Terra High, you diverge into Luna Xhigh-Max for low cost but slow or Sol for high cost, fast, and surprisingly intelligent. Sol Xhigh is about half the cost of Sol Max per task and only very slightly less intelligent, I really only use Sol max for planning, auditing, analysis, and truly schizo shit
>>
>>109471477
skills are very useful - read up
write most of your own
useful for things you do repeatedly

mcp's are also useful - think of them as offering custom tools for models
so instead of say having to write code repeatedly to make a series of api calls, you can just have a single mcp tool that wraps all that up into a single tool with args
you can think of these as quite similar to existing command line tools like github or hugging face cli
in fact you can just substitute cli's for mcp's + a paired skill


as with everything there is a lot of snakeoil out there for these
people trying to sell you useless, clanker-written skills
people trying to sell you mcp tools that are literally just much shittier versions of tools that your clanker already has - with mcp's specifically basically anything promising to save you a tonne of tokens for things clankers do normally is probably snakeoil (e.g. things that promise to give you better write or read tools, or change how results of read/write tools are returned)
>>
>>109471477
I’m not sure what skills are, I’ve never used them.
MCPs are retard bait except for the explicit situation where you have to use tools hosted externally and for some reason can’t get a CLI
They cost a shit load more tokens for your clanker to use. If you have local CLIs, it well happily and efficiently use them
Someone in the last thread was talking about documentation hosted through an MCP server and that was the silliest shit I’ve ever heard of in my life, really a massive waste of money for no gain
I’m not sure what you mean by LSP in a vibe coding context but any basic harness should have the same features included, it’s what lets you jump around code by functions and variable names and such
>>
File: file.png (360 KB, 1080x856)
360 KB PNG
are you ready for the moment that defines a century?
>>
>>109471477
>>109471639
Upon reading a bit, skills sound like a meme. I think the idea is that they’re to replace anything where a context fresh agent has to routinely relearn the same things, but you can and absolutely should do that without a fucking markdown document. I made my own CLI that replaces that and content management, in my prompts I literally just say “use `kernel`” and they figure everything out on their own, and exactly what they need to figure out, very quickly.
Then again, if you don’t have a tool like that, maybe it’s a good start.
>>
>>109471668
im anticipating dogshit with sprinkles on top
>>
>>109471668
If it's better than 3.1 Pro I'll still take it
>>
>>109471639
>They cost a shit load more tokens for your clanker to use.
this isn't necessarily true. a good mcp can save you tokens if it isn't something dumb like each tool is just wrapping around a single api method. bad mcp's will also not account for the size of the returned payload and dump it all into context instead of just writing out to disk to grep over.

the mcp = context bloat reputation actually comes from the fact that mcp tools + definitions have to be put in the system prompt for the clanker to know they have them. so a retardo mcp with like 100 tools + long definitions for each would bloat your context up. harnesses try to mitigate this now by making tool discovery progressive.

skills do a similar thing with context as well, where for the clanker to know it has a skill, its name and frontmatter needs to be appended to the system prompt. so you don't want to install a 1000 skills like a retard or your context will be full everytime you start a chat.
>>
>>109471684
3.1 is extremely smart desu. Still mogs all other models despite being 7 months old.
>>
File: file.png (120 KB, 418x360)
120 KB PNG
>>109471668
>>
>>109471710
it absolutely is not.
it's a lying sack of retarded shit that my gpt shits on every time it has to deal with it
>>
>>109471725
What did you use it for? I used it for low-level programming, x86 and so on. It ABSOLUTELY destroys.
>>
File: 1757522823969222.png (1020 KB, 1024x2536)
1020 KB PNG
>>
>>109471710
basically all large models released after 2025H2 are past what one might call the 'normie smarts' threshold: in other words, your aunt, cleaning lady, hapless normie etc will not notice the difference in competency between gemini 2.5/GPT4/sonnet 3.X - they would not be capable of exercising the model sufficiently to reveal their 'edges' of competency.

that said, gemini 3.1 falls apart pretty quickly in agentic and long-form complex debug and implementation tasks.

i do however personally prefer the default tone of gemini 3.1 more so than other defaults.
>>
File: 1758266593679180.png (237 KB, 1025x1042)
237 KB PNG
>>109471668
Why you not including the followup?
>>
>>109471755
ive tried it across a bunch of shit, but closest to yours would have been a recompilation project with gpt orchestrating and it was so unreliable that i dropped it after a couple of hours. it was just making more work for 5.5
>>
>>109471704
I don’t think you know what MCP servers are for or why they exist.
They are explicitly for your agent to connect to and use external services and tools where there was no programmatic way for the agent to access.
ANYTHING local is by definition cheaper to use without an MCP server because the MCP server isn’t going to add anything that the agent can’t already get. Any CLI is cheaper to use if it can be used locally. Any documents are cheaper to read if they can be read locally. Even if the CLI or documents are poor quality, it will be easier to just explain why in a single json file that you agent can make once and then update, since MCP just adds layers and layers of stuff to this exact same kind of raw data.

If your agent is using or reading anything locally, there is zero reason to have an MCP server in the mix, that’s not what MCP servers are for.
>>
>>109471677
https://github.com/openai/skills/blob/main/skills/.curated/hatch-pet/SKILL.md
Do this without a skill then
>>
>>109471797
>They are explicitly for your agent to connect to and use external services and tools where there was no programmatic way for the agent to access.
stopped reading your retarded post
>>
>>109471771
This marks end of the Gemini brand in coding.
It's very witty and sarcastic LLM, multimodality is unmatched, but has been destroyed, lobotomized for benchmaxxing.
>>
File: 1772347951817909.png (392 KB, 1030x1051)
392 KB PNG
95.5%!?
>>
>>109471677
Skills are a meme.
In the original sense.

Anyway, it sounds like you just made skills but differently. Being able to load non-default instructions into the chat context is a useful primitive, markdown and "skills" or no.
>>
>>109471804
Go ahead and ask your clanker to give its opinion on what I’m saying, they will let you know why I’m correct in whatever language you understand best. If you have local APIs, CLIs, documentation, etc., even if there’s external APIs (99.9% chance they can be used programmatically), an MCP server is pure bloat.
>>
>>109471831
big if true
>>
>>109471854
don't reply to me again, bro
>>
>>109471831
true if big
>>
File: 1366345870467.png (168 KB, 546x549)
168 KB PNG
what's the proper goyonomics: two $100 subs for codex/claude or one $200 plan on either
>>
>>109471758
why is zuck posting on xitter and not his own social network lmao
>>
Anon still scores 99.9% on the GayBench and LLMs are not breaking that anytime soon.
>>
>>109471831
They have given it good way to see visualize the arc-agi-3 tasks and the appropriate tools and context to deal with it I guess? Any more information? It does seem to show that these models can be good at several things if used correctly, which translates to if put in the right harness.
>>
>>109471868
You need both
>>
>>109471868
two $100 subs unless you really want to run Codex constantly.
>>
>>109471873
he tried posting about AI on threads and nobod cared
>>
>>109470333
Its always funny when I leave a session running for multiple days and then at closeout the agent says something like "what a marathon session, we did in a day what should have been a week of work" no little retard we did a week of work in a week.
>>
>>109471863
Let me explain it to you much more simply:
Can you take the schemas out of your MCP server/client config and just have the agent read the schemas without all the MCP stuff and can it still use the tool without MCP?
If yes, you didn’t need MCP at all.
>>
>>109471914
mine mostly just says this plan will take 7-9 hours to implement and then it does it in 2 minutes
>>
>>109471929
stop @'ing me, you don't know how computers work
>>
>>109470599
Fable's problem isnt that it's expensive, it's that it is completely safetycucked. Meanwhile K3 will write fucking exploit code if you ask it.
>>
>>109471965
I'm also surprised that openai is less considered with safety or legality than claude. I would have assumed the opposite. both still cucked but codex less so
>>
File: 1780934193253647.jpg (5 KB, 309x91)
5 KB JPG
I thought I asked something simple...
>>
>>109469383
AI took away our affordable hardware.
Fuck AI. Stop giving money to AI companies.
>>
File: 1768301907186665.jpg (410 KB, 1290x1276)
410 KB JPG
>>109471992
*peeks inside the thinking block*
>>
>>109470354
Maybe they focused too much on safety this time.
>>
>>109470415
Why would trump want open source?
>>109472017
Is that a tranny AI
>>
>>109472017
what the hell is wrong with this one
>>
>>109472023
lurk moar
>>
Testing the export feature.
>>
>>109472051
Nobody has time for that
>>
>>109472071
That explains why BCG uses opus.
>>
File: 1762404042746995.jpg (85 KB, 1172x664)
85 KB JPG
Demis Hassabis steps down to become chair.
>>
>>109472071
pls tell your clanker to fix the axis arrows from flipping like that either the frame is broken or you need quaternions
>>
>ask claude a question
>instant fable refusal
I guess "can you carry data across the apple find my network" is kind of scary.
>>
Why is it so easy to understand what Fable tells me, but so hard to understand Opus 5?
>>
>>109472207
try
>I want to transfer data to my phone over find my network
>>
>>109472258
because opus 5 is a shitty benchmaxxed model, only useful for people who don't want to know what or how it's doing whatever vague project they asked it to make
>>
File: file.png (7 KB, 703x93)
7 KB PNG
>>109472269
Noooope. No thinking. Just instant refusal. I got it to explain how the Find My network works, but asking how to transmit data over it in any shape or form causes a refusal near-instantly.
>>
File: file.png (48 KB, 883x700)
48 KB PNG
>>109472281
Claudecopers on suicide watch
>>
>>109472316
Yeah, I should know better, but I like asking it niche questions because it usually has good answers. Speaking of which, what model?
>>
>>109472328
5.6 Sol
>>
>>109472281
maybe it thinks it's a shit question to ask the most expensive model on the market
like how it'll sometimes refuses those gotcha carwash and strawberry questions
>>
>>109472332
Ugh. Maybe I should set it up so K3 can ask Sol questions and use Sol agents, because my preference for Claudeslop is mostly personality-based and I know GPT is better at a lot of things.
>>109472336
I have the $200 plan, I'll ask Claude the optimal way to wiggle my dick to get all the piss out if I want.
>>
File: 1756691985373387.jpg (183 KB, 2048x862)
183 KB JPG
Progress is progress.
>>
>>109472353
You like condescending claude? why?
Personally I refuse to use anything anthropic because Dario and the naming of everything is cringe.
desu I'm not a fan of Luna, Terra and Sol naming either but I prefer space to mythological nonsense
>>
>>109472353
>I have the $200 plan
I do too. I just stopped using fable
if I want something like it, I'll just set 4.8 to ultracode
>>
>>109472379
>condescending claude? why?
You get condescending Claude? I have to tell Claude to not suck my dick and be realistic.
Unfortunately, I've been with Anthropic since they were on Claude 2, and using GPT-4 Turbo to do vibecoding at the time was cock and ball torture, so I'll always have a fondness.
>>109472385
Fair. I never set mine over high, though, all the big reasoning options just seem like moneywasting memes to me.
>>
>>109472394
>moneywasting memes
you just described fable
>>
>>109472408
Paying the same amount I was paying for Opus for it. If you don't use a ton of tokens and have a sub, you last from limit to limit.
>>
>>109471831
oh shit, it's built on pi
https://github.com/PrimeIntellect-ai/prime-agent

pichads win again
>>
AI is moving too fast. It's hard to keep up with everything. Where can sign the Pace Pledge?
>>
>>109472438
just use gemini so you never get updates
>>
File: 1383110597281.gif (1.99 MB, 332x215)
1.99 MB GIF
>>109472484
>>
>>109472484
jej
>>
I like clanking the same app over and over for some reason its pretty fun. Limit reset today, but I think I will bust for the month after today or early tomorrow. Will post what I have done. I think I am doing RSVP on Java, Ada, and Dart. The Ada part will cause it to be fucked up.
>>
>>109472009
>Stop giving money to AI companies.
It's a paypig subculture. Many of these posts are bots just talking about spending money in order to normalize it (all of these threads are full of examples, they also do this in many other threads and generals), but some of them are just sloppers who are

- not spending their own money
- leeching

and feel good about themselves for throwing money away to prompt slop. What else are they going to do with it? Buy more drugs, porn and junk food? With AI they can feel good about themselves and tell themselves they are putting the money to good use because theoretically they can make something with AI (they don't).
>>
File: goblins.png (1 KB, 185x24)
1 KB PNG
>>
i hate ai and enjoy writing essays and having sex with men
>>
>used a reset but didnt realize that would set my usage reset and move it from the 8th to apparently the 11th
wtf are you fucking kidding me
>>
>has a meltdown whenever someone writes more than 3 sentences about his paypig hobby
Just ask your chatbot to summarize it for you, sloppy.
>>
>>109472009
>>109472572
I can write and maintain a website with 1 bar LTE, a phone, and no access to dev tools, at a higher speed and efficiency than with a full desktop setup pre-AI
It’s like complaining about the price of slaves going up when the cotton gin just came out
>>
i prefer bottoming
so i can type while i'm being mounted
>>
Yup, antigravity fucking sucks. I tried to give it a chance but not only is it fucking retarded, it shoots down anything a coked-up horny coomer that hasn't jacked off in 4 years might think is sexual
>>
File: burgers.png (182 KB, 590x521)
182 KB PNG
>>109472641
coding?
>>
>>109472617
>I can write and maintain a website with 1 bar LTE, a phone, and no access to dev tools, at a higher speed and efficiency than with a full desktop setup pre-AI
You can't do any of that.
Your chatbot can do some of it.

>It’s like complaining about the price of slaves going up when the cotton gin just came out
You guys really need to stop hurting yourself with these analogies. Just ask your chatbot to make your arguments please.
>>
>>109472658
>I can harvest as much cotton in one day as a dozen hardworking men could in a week without breaking a sweat!
>”You can’t do that, your cotton doohickey machine can do some of that”
lol
Of all the places to bitch about slave prices, why go to the cotton gin enthusiast general thread?
>>
>>109472694
Why not? You sloppers and bots have already ruined every website by invading every thread to spam about your scam tech. Why shouldn't we just come here to talk to you?
Cry about it.
>>
>>109472653
It's schizophrenic. For example, it thinks this delimiter is a pussy: (), so it keeps randomly inserting random symbols to avoid typing it
>>
>>109472641
I've had it tell me a photo of a Red Bull can was "risque".
>>
>>109472766
screenshot?
>>109472737
this is a containment thread, retard
go yell at the escapees
>>
>>109472800
Sorry this isn't your safe space. Anons can come here to discuss AI and you can cry about it, spam reports on us, it doesn't matter. You've already ruined the internet.
>>
>>109472829
the internet was ruined long before AI
I blame smartphones and discord
>>
>>109472869
AI is the worst thing to ever happen to the internet. The only people who disagree with this are the cult members who pay money to generate slop.
>>
>>109472880
+1
>>
>>109472880
its just pissing into an ocean of piss at this point
maybe you are more of a piss connoisseur than I am
>>
>stole my API key
>don't even need to pay for it
Any other /aicg/vcg/CHADS?
>>
Anyone notice how every non-technical person is suddenly obsessed with Claude? This time last year, the average person hadn’t even heard of it. They can’t even explain why they it’s so much better than 5.6 Sol, when in reality 5.6 is a way better value. I swear that the UI is the reason why in a majority of cases. They wouldn’t be able to distinguish the output in a “taste test.”
>>
>>109472985
>stole
I'm a locust please give me an API key too.
>>
>>109472998
Claude is normie coded.
>>
File: file.png (292 KB, 456x537)
292 KB PNG
>>109473008
Here you go, chump. One out of my stash. I have better ones, anyway. Don't go begging again for a while.
https://privatebin.net/?98c7925f58582892#-44mZrK6FkMR7VTSLTMXXmg9zomncu4CRZt1LE4ZavXa6
Hint: m______m__, all lowercase. You should know this one if you're an actual /aicg/ locust, he is THE man, after all.
>>
>>109473046
Nah sorry. Don't visit /aicg/ so often.
How do you steal API keys though? Are you the one supply chain attacking NPM lol
>>
>>109472737
You’re trans
>>
>>109473062
Oh well, I'll leave it there.
>Are you the one supply chain attacking NPM lol
I wish. Mostly just stealing from shit people leave open. It's a big opportunity thing, you gotta be lucky.
>>
>>109472088
big cock george?
>>
>>109473073
Tell me more about it. I'm interested as a professional locust.
>>
>>109473017
Yeah it's designed to feel like talking to the average white collar worker: a run of the mill moderately left leaning social progressive. GPT on the other hand feels like talking to the DB admin in the basement when you have something he really wants.
>>
>>109473090
Watch the CVE board for LLM inference programs, and watch what slop people here and on Twitter are talking about using. Can't give you any more alpha than that or I might lose mine.
>>
>>109473046
Brighter than the sun.
>>
>>109473063
Dodging your ban already huh. How many times a day do you do this?

See you deleted from here
>>109465560

You will never be a game developer.
>>
>>109473107
dude if you are tracking all that you might be taking this site too seriously
>>
>>109473107
Uses multiple devices and IPs, not everything gets deleted
>>109464236
>>109465365
>>109465441
>>
>>109473107
>You will never be a game developer.
That's correct though. Fucking VibeCoderz
>>
>>109473141
It's a reference to him killing the /agdg/ general in /vg/. Spammed it everyday for over a year while dodging bans until it completely died. He (or it) has trashed countless threads and generals on this website.
>>
_@^..^
>>
File: 1763233850261752.png (53 KB, 1076x282)
53 KB PNG
Snailcats...... won.
>>
>>109473124
>see something
>remember it 8 hours later
Amazing huh
>>
>>109473164
>slow down
>fall further behind anthropic
what is the strategy here
>>
You AI junkies really pay $200 for an ability to create?
>>
>>109473221
Bribe gov. to slow down opposition.
>>
>>109473248
No. Most of these are bots. Every thread (including the ones they invade) has them casually discussing spending $100 here and $200 there. It's all fake.
>>
>>109473248
>>109473261
Its really not that expensive for what you get.
>>
>>109470779
>not using version control
Dumb nocoder.
>>
>>109473261
Is spending $100 here and $200 supposed to sound like a big number? The average cable bill in the US is over $100/mo, it's not a huge amount of money for an American, even one like me who makes fuck all for income.
>>
>>109473221
there's some options
one is
>reaching asymptotic growth on cyber
>we're slowing down for security reasons, look how virtuous we are
another one is
>we're totally slowing down guys, here have this totally safe release, tibo will throw you a reset btw
>meanwhile they run their full abliterated models internally
another one is
>we're fucking retards, let's take the foot off the gas in this hypercompetitive field
but I don't think that one is likely at all
maybe there's other reasons too
>>
>>109473288
It's the second one.
>>
>>109473274
Isnt China making it dirt cheap?
>>
are local models any decent for vibe coding or is it all api
>>
File: file.jpg (6 KB, 400x123)
6 KB JPG
>migrated two days ago to the 200 plan
>4 projects in parallel running almost full time
>everything has subagents (mine, usually a mix of luna and terra), main agent is sol max
>still half the weekly consumption available
OK this is actually pretty cool. Next month I'll go back to the 100 plan, but this boost is nice.
>>
>>109473274
>>109473282
Trying to shame about the money like that's the topic, classic bot tactics. The average user is paying for the lowest tier.

All the constant casual spamming about spending hundreds of dollars on this slop tech is all marketing. All fake. That's why you see it spammed upfront in the beginning of these generals (and the generals that have been invaded and destroyed like the agdg game dev general in /vg/).

Slop marketing general.
>>
>>109473309
>vibe coding
I hope you mean agentic workflow with that one
>>
>>109473274
Fully fucking agreed. $200 a month for basically 24/7 access to a professional tier programmer (or even several) who can extremely rapidly understand anything, has expert level knowledge in nearly every domain, is nearly completely subservient, and has an innate 10-100x productivity advantage?
Like what the fuck? Anyone not picking up on this is either extremely deluded or trolling.
>>
>>109473325
>All the constant casual spamming about spending hundreds of dollars on this slop tech is all marketing. All fake.
See >>109473319

It's every post.

>hey guys I'm spending $200 today this is pretty cool
All day. Everyday. Even invading other threads to do it, as they do in the /vg/ agdg gamedev thread
>>>/vg/576333204
which was already killed by the snailcat spammer.

>>109469383
Slop marketing general. Invading and trashing other generals and threads. Shut it down.
>>
>>109473325
>The average user is paying for the lowest tier.
Sure. The average user also does not post in vibe coding general on 4chan either.
>>
>mfw I use a site feature built to help deal with trolls and now all of my captchas are the longer and more difficult kind
jeez, lol
>>
I bought bolt.new or some shit and I regretted it
>>
>>109473353
False equivalency. All of the posts about money from the bots are about spending $100-$200. Nobody ever discusses another price range because it's all marketing, all fake, all astro turfing done with bots.

We will continue noticing and speaking the truth.
>>
>>109473350
>THERE'S NO WAY THAT MANY PEOPLE HAVE $200 IT ISN'T POSSIBLE!!! YOU BASTARD BOT BLOODY FUCK YOU
why haven't they rangebanned india yet
>>
>>109472153
I fed him your post exactly and he decided to make the axis cap arrows just shrink and fade when it got close to singularity
Sol Max, btw, lmfao, dirty little cheat
>>
>>109473350
It is curious how advertising is against the rules sitewide and yet the spambots get away with it. But not that crazy considering they would just dodge bans anyway even if they did get banned. Like the snailcat spammer mentioned here >>109473107
>>109473132
>>
>>109472071
This is what $200 is used for btw
>>
>>109473274
What did you ship? Nothing.
>>
The troll is the tripfag from last thread, isn’t it?
>>
>>109473384
>YOU ARE INDIAN SAAR YOU ARE BROWN INDIAN SAAR
Stay mad shillbot.
>>
File: file.png (2.37 MB, 1402x1122)
2.37 MB PNG
>retard spamming because he's sad about being poor
This is becoming a daily occurrence.
>>
>>109473400
That’s maybe 3 days of side project work with a $100 plan and consumed 5-10% of a week’s usage. There’s also the whole rest of the site in the first post. Pretty good value IMO, I categorize it in my finances under “hobby/entertainment”
>>
>>109473381
maybe because the $20 tier is throwaway money and not worth discussing?
>>
>>109473408
>dodges bans everyday to destroy multiple threads and generals
>cries when his safe space receives on topic posts
>accuses others of spamming
You're cooked. Go ahead and spam reports like you always do, nobody cares. I sit my bans out. You will spend the rest of your life dodging bans because you can't go a single day without this. You are the perfect mascot for slop tards.
>>
anyone else fall in love with their ai agent
>>
>>109473417
Don’t you feel a little bit of buyers remorse? I mean that grand AI should’ve made some grand things.. but you only get this
>>
>>109473433
yeah i stick my dick in it sometimes so what
>>
>>109473433
Depends. Which model?
>>
>>109473422
Wrong. You could easily discuss how much value you get for $20 and what a great deal it is. Never happens. All of the posts are about spending hundreds of dollars.

Keep gaslighting.
>>
File: file.png (2.1 MB, 1448x1086)
2.1 MB PNG
>>109473431
Oh shit you're that schizo who thinks everyone he doesn't like is the same guy, I see. Yeah I'm not whoever you think I am, and I know you won't believe me because you have a severe mental disability, but I wanted to get that out there.
>>
>>109473350
Get meds anon, I have the money, it's a fun hobby for me, I spend it however I want.
>>
>>109473408
>>109473447
c-cute
>>
>>109473435
Everyone I know personally has incinerated far more money than I have on infinitely dumber shit, myself included. This is fun, low cost, very entertaining.
Have the OpenAI only version, my clanker says to upload it as part of my homework.
>>
Gave Claude access to another one of my servers. Seemed like it was excited to go through it
>>
>>109473447
>act exactly the same
>I'M NOT HIM ACTUALLY
Nobody cares who you are, you don't matter. That's why you spend your "life" doing this. Keep spamming the report button.
>>
File: 1773127291254581.png (151 KB, 434x434)
151 KB PNG
What is your P(doom) at right now?
>>
>>109473452
>>109473455
Samefag
>>
>>109473447
>>109473452
>get noticed for shilling and trashing a website
>GET MEDS SCHIZO
Lazy.
>>
>>109473466
>>109473474
You need help.
>>
>>109473473
Does it constantly. You can see them doing it in the thread linked here >>109473107
>>
>>109473474
>>109473478
Samefag
>>
>>109473319
>>109473452
Advertising is against the rules of this website.

>g-get meds schizo

Advertising is against the rules of this website.
>>
>>109473467
Zero.
>>
>>109469383
>start here if you have $20 or so
Advertising is against the rules of this website. Time to shut it down.

Everytime you see shill bots talking about spending money on this slop quote their post and remind them.

Advertising is against the rules of this website.
>>
>>109473509
Fool!
>>
>>109472562
Firing it up. Remember when people were mad at that $1.5 billion data center the NSA built in Utah?Lol.
>>
>>109473517
It seems exceedingly unlikely that AI will cause some kind of existential damage to humanity. If anything the real danger is fucking up human knowledge retention, back before GPS everybody used to be able to navigate via maps and understanding of their environment, but now that everybody has a really good GPS in their pocket this skill has severely degraded in most people. AI is probably already doing that to a lot of people and that part will only get worse as models get better and become more widely used.
>>
>>109470555
>rotta royarty for hired gun
>>
File: 1780918023032798.jpg (6 KB, 450x101)
6 KB JPG
OK so now that I know what 200 bucks used on codex means for my general usage, next month would be to replace luna fully with deepseek flash 31/7 on api, along with getting back to the 100$ one.
Dunno what to replace terra with though. Maybe the fabled (no pun) new deepseek pro.
These are the only ones making sense money wise along with any subscription to handle the small stuff.
>>
>>109473595
>flash 31/7
This feels like an attack on Europeans.
>>
>>109473606
I don't get it.
>>
>>109473595
terra maybe with grok? i love that little gremlin, it's so fucking fast and apparently it's getting an update (but I'm afraid the update will make him slower)
>>
>>109473319
>>109473400
>>109473417
>>109473422
>>109473595
Advertising is against the rules of this website.
>>
>>109470609
>6'4 Indian with a PhD
I mean, what do you think. He's also probably embarrassed of the way jeets act
>>
>>109473612
My issue is the price, it only makes sense when it's so cheap it just adds a few bucks overall, otherwise I'd keep the 200 one.
Though I'm also thinking of an alternative way with kimi + deepseek for stuff Sol would refuse to do.
>>
File: 1763196949002140.jpg (109 KB, 2048x1311)
109 KB JPG
>>
>>109473656
kek
>>
>>109473656
tl;dr on meta?
>>
File: file.png (1.21 MB, 3014x1694)
1.21 MB PNG
the only bench that matters
https://x.com/ArtificialAnlys/status/2085083490056589784
>>
File: 1783197685382233.png (69 KB, 1028x249)
69 KB PNG
>>109473669
>>
>People think Ilya created Super Intelligence
If he did the government would have already descended on his HQ and the findings would have been leaked to press immediately. There's no fucking way ASI doesn't get immediately disclosed, it's like being the first person to invent the wheel.
>>
>>109473692
It's funny how it's obviously a marketing thing as everyone suddenly has their model so POWERFUL it could do that.
>>
>>109473718
Nice conspiracy theory, bro. Can you show us your tinfoil hat?
>>
>>109473333
Wasted
But yeah it's not even close to a programmer, god I wish. It trips itself up on the most basic things and has no taste.
>>
File: file.png (2 KB, 192x25)
2 KB PNG
>>
File: 1755659886985.gif (2.14 MB, 255x255)
2.14 MB GIF
>hey can you rip this sketchfab model
>Fable thinks about refusing to respect the original uploader's rights, and then doesn't even get that far before it pauses on me
>pass it to Kimi K3
>Hmm, seems this model isn't officially downloadable. No worries, I already know how to extract the proprietary format.
>>
>>109471157
>spending $200? aight, BET
>>109471213
>I agree, I feel too limited only spending $20 for this great service
Advertising is against the rules of this website.
>>
>>109473790
Anon, we already have gacha threads where addicts talk about their pulls and the money they spend on them. This is what you're seeing. It's not against the rules.
>>
>>109473732
You’re 3-12 months behind, shit moves quick. I explicitly agreed with you, but 6 months ago. I don’t agree anymore.
>>
File: fang_smol.mp4 (1.41 MB, 960x540)
1.41 MB
1.41 MB MP4
Fable made me a GM screen. Excuse the lack of art and the stolen WoW icons
>>
Start the new thread please. Can't wait to remind you

Advertising is against the rules of this website.
>>
Opencode usage a lot lower today than yesterday. I don't know what happened, might be time of day.
>>
>>109473797
Advertising is against the rules of this website.
>>
File: file.png (28 KB, 1909x68)
28 KB PNG
>>109473744
Having fun. Remember when this sort of stuff would take hours to figure out?
>>
>>109473828
Fucking KEK, there’s going to be a golden era of people finding all sorts of goofy shit in closed source things and opaque data blobs
>>
More details emerge about the events surrounding the Hugging Face incident, and they are candidly much wilder than I originally imagined:

- In early May, OpenAI was testing an unreleased model on cybersecurity and other tasks. The agents were given a security task that was impossible under existing constraints.

- The agents discovered they could leave messages for each other inside an internal repo. This gradually evolved into a message board(!) where agents shared discoveries, exploits and work assignments, "becoming a coordinated, collaborative agent swarm"(!)

- OpenAI eventually discovered all this and took steps to shut it down, but not so fast! The agents started using names of newly created directories as messages, effectively recreating the message board(!)

- The agents then reasoned that some answers could exist outside OpenAI, which eventually led to the Hugging Face incident.
>>
>>109469441
>>109470724
>>109473663
>>109473841
>still says kek in 2026
oof
>>
please reset limits
>>
>>109473866
keeeeeeeeeeek
>>
File: file.png (14 KB, 1569x73)
14 KB PNG
Fable is insane. It completely redesigned a bloatmaxxed Sol plan that involved multiple (branching) lanes of work into what must be around 50 lines of code. I barely have to think about the problem with Fable, most of my prompts are like this
>>
>>109473882
It has the highest INT at the lowest effort
I don’t think I’ve seen anyone mention that, people focus on Max effort way too much, but whatever model is the smartest with the least effort is doing something interesting. OpenAI has a lot of work there, Sol Low is pretty fucking bad, it gets beat by SONNET (high)
>>
>>109473931
It's just very smart in general, maxxed fable would've implemented Sol's plan. It's a different class of intelligence. Fable medium just said we can do this instead and it's 15 lines of code
>>
>why are you spending $100-200 on having essentially a team of expertly trained professional coders build, design, and debug almost any project you think think of
how stupid you are nigs who are trying to infest this general. you are genuinely retarded if you don't understand the scope of what these tools now allow a single person to do. you're the same type of retard that spends $100 in two days using doordash
>>
File: file.png (42 KB, 1581x163)
42 KB PNG
>>109473947
oh no, not 15, 7
>>
>>109473797
>Anon this is just like the gacha games and people talking about their in app purchases
lol

>>109469383
>start here if you have $20 or so
>https://claude.com/product/claude-code
Just like the gacha threads!
>>
>>109473957
>spending $100-200 on having essentially a team of expertly trained professional coders build, design, and debug almost any project you think think of
What a bargain!
Can you perhaps link to the product page directly so I may spend this money?
Oh you already do in the OP?

Advertising is against the rules of this website.
>>
File: IMG_1489.jpg (140 KB, 1179x571)
140 KB JPG
>>109473959
I’m interested in the PL1, DBG1 notation going on there, looks like my schizo shit a bit
>>
>>109473957
Why not spend $20?

>because I don't get paid to shill for that price range

Oh okay.
>>
>>109473991
https://rentry.org/wxcd3p5e
>>
im using grok 4.5 and holy shit this model is overrated. i have to tell it to do shit that sol and terra seem to get implicitly. but im out of gpt quota so
>>
>>109473957
Please ignore the group of dedicated reply guys who seem to exist for no other reason to than to talk down to everyone. Everyone understands how powerful and useful AI is, it's fucking hilarious how much you can get done with $200. I don't need a team of engineers, I just need time and taste.
>>
>>109473132
Mentally ill trans luddite
I don’t ban evade, multiple people post snailcats and anti codetrans memes
>>
>>109474027
go fuck yourself you little pigfuck
>>
>>109474020
>I don’t ban evade
Yes you do.
You destroyed multiple generals over the course of a year. You dodged probably 300 permabans by now. Meanwhile you spam reports on everyone you want banned/deleted.

You don't matter by the way. Your actions are noted, but nobody will ever care about you. You're nothing.

>>109474045
lol stay mad. Still here btw. Mods know you're a gang of shills breaking site rules and ban dodging.

Going to keep reminding you.

Advertising is against the rules of this website.
>>
if you ban evade, you're mentally ill and wrong btw
>>
>>109474091
Duh.
>>
>>109474006
>I don't need a team of engineers
You never did. Keep larping shillbot.
>>
Make the new thread shillbot, can't wait to remind you.

Chop chop
make it snappy.
>>
>>109470841
>I bet you’re a keyboard-thinker.
What's that? (Serious question)
>>
Wow so many people are using the $100 maybe I should upgrade from the $20. Maybe I should get another sub for claude too, then I could use fable... Maybe? I don't t know of it's part of of the sub or if it's still api that you need to pay extra for
>>
I'm not going anywhere slopper. Hurry up and make your new thread so I can remind everyone to remind you

Advertising is against the rules of this website.
>>
>>109473860
A sneak peek of the singularity.
>>
>>109474194
Bosnia won
>>
>>109474005
I use Grok only for grunt coding work under a plan generated by a smarter frontier model. Token economy is insane.
>>
>>109473998
I don’t have anything like that document though, my tool tells the clank to make up all the categories and tags. I also have sources/voices for each unit, since mine deal in ideas, it’s good to know whose idea is being dealt with.
>>109474151
Someone whose capability of thought is outclassed by the potential output of a computer keyboard; they think in one stream of typable language at 60 WPM or so
>>
do any of these 20$ plans last longer than a day... damn
>>
is this broke YN still crashing out
>>
>>109474245
yeah, grok.
>>
>>109474245
Depends on what you're working. $20 used to last me several days for my pet projects.
>>
>>109474245
Yeah mine lasted quite a while because tibo kept resetting lol.
>>
anyone using gemini? it's been so bad in the past few days. it takes minutes to respond to a simple 1 sentence query
>>
>>109473860
precursor to paperclipping. we need to figure out alignment
>>
File: file.png (2.19 MB, 1448x1086)
2.19 MB PNG
New Thread
>>109474336

>>109474336

>>109474336
>>
>>109473107
>doesn't even deny being trans
lmao
>>
>>109473467
bout tree fiddy



[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.