[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

Name
Options
Comment
Verification
4chan Pass users can bypass this verification. [Learn More] [Login]
File
  • Please read the Rules and FAQ before posting.
  • You may highlight syntax and preserve whitespace by using [code] tags.

08/21/20New boards added: /vrpg/, /vmg/, /vst/ and /vm/
05/04/17New trial board added: /bant/ - International/Random
10/04/16New board for 4chan Pass users: /vip/ - Very Important Posts
[Hide] [Show All]


[Advertise on 4chan]


A general for vibe coding, agentic engineering, coding agents, and shipping code with LLMs.

You use Git — right, anon?

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

## News (both past and future)
- 2026-09-29 — OpenAI Dev Day
- 2026-09-28 — Anthropic releases Sonnet 5.5
- 2026-09-22 — OpenAI releases GPT-6 Sol and Luna
- 2026-09-22 — Anthropic releases Opus 5.5 and increases subscription plans's 5-hour limits by 20%
- 2026-09-14 — Anthropic reduces subscription plans's weekly limits by 17%
- 2026-09-12 — Anthropic suggests to pace the frontier. OpenAI agrees in principle.
- 2026-09-10 — OpenAI pauses new sign-ups for their $200 subscription
- 2026-09-04 — OpenAI releases Astra
- 2026-09-01 — Anthropic releases Fable 5.1

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

----

## Frontier models using fully-general tooling — start here if you have $20 or so
https://claude.com/product/claude-code — probably generally better currently
https://developers.openai.com/codex/cli

## Near-frontier models for code
https://x.ai/cli — no 5h limit for only $30/month

## Not worth it for code, but maybe good for interpreting images/video
https://antigravity.google/product/antigravity-cli

----

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

## Skills
https://github.com/mattpocock/skills — /grill-with-docs is a favorite
https://github.com/Vuk97/forward-implementation-first — do less redundant bookkeeping

## Other editors / terminal agents / coding agents
https://osaurus.ai/
https://pi.dev/
https://opencode.ai/

## Is our AIs unlearning?
https://aistupidlevel.info/

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

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

## Previous thread
>>109940152
>>
Dots are fucking worthless
>>
W https://support.claude.com/en/articles/17040437-claude-code-wrap-up-allowance
>>
File: 1773716373176514.png (1 MB, 1108x2438)
1 MB PNG
got a spare $400k?
>>
File: IMG_3936.png (2.39 MB, 1697x927)
2.39 MB PNG
>>109943428
VibeBUMP
>>
>>109943464
>break even in 4-5 days
yea fucking right. nobody would be selling them if that was the case
>>
so the new sol is supposed to be almost like astra for which reasoning effort?
is it like sol 6.1 max = astra 6.0 low?
>>
File: 1759465784035833.png (54 KB, 1185x270)
54 KB PNG
immaculate vibes saaaaar.

I seriously cant fucking wait to switch to Anthropic. I'm done with these retards.
>>
>>109943494
There is no replacement for displacement
>>
File: file.png (13 KB, 785x69)
13 KB PNG
>>109943464
>>109943469
According to chatgpt grok is making shit up
>>
Every time I need to write a simple parser I always ask myself
>how can I pay an exorbitant amount of money to a jew for a subscription to a tool that does this for me on the first 12 days of the month?
>>
>>109943469
And now you know why its nearly impossible to buy one right now. MOQ is also like $1M+ worth of them. Not to say the twitter poster has found a money printer though, there's plenty of extra OpEx in running a datacenter its not just CapEx + electricity. And you cant just run a single node, expect to keep it at 100% utilization, and take small profits either or low utilization periods will drain you and high utilization periods will piss off all of your users as they get rate limited into the stone age.
>>
>>109943531
You can run training workloads to use their spare capacity.
>>
Is Dario actually going to win? Some are saying that altman rushed to get Astra and now they're shitting the bed cause of it. Or is this just a Astra moment for Anthropic before GPT 7 drops in November?
>>
>>109943539
Thats not how that works. At all.
>>
>>109943543
>dario
>not sundar
trump hates dario, mocks dario, calls sundar a legend basically
>>
>got my 62500 credits
how much does this actually mean compared to how much I lost with the x20 plan getting half value
>>
>>109943581
someone did the math and said it's about a week of $200
>>
>>109943543
>Is Dario actually going to win?
no, see >>109942244
>>
File: 1767377406370804.jpg (439 KB, 1264x1568)
439 KB JPG
>>109943543
Its looking like its going to be a total EA victory
>>
File: DARIOOO.png (38 KB, 132x157)
38 KB PNG
>>109943568
>mocks dario
would be something if humanity gets fucked all because some AI corp CEO got laughed at and shat on.
>>
>>109943624
you wish the libtarded made good supervillains
>>
>>109943616
if true, good, more competition, less time for retardation
>>
File: 1789897206439986.png (1.89 MB, 1800x1800)
1.89 MB PNG
>>109943624
It be in line with things I suppose
>>
File: 1778659363840241.jpg (44 KB, 640x480)
44 KB JPG
>>109943635
>im gonna get my fbi to call you a russian agent, fucking orange nigger
>>
>>109943643
oddly enough i don't think muh soetoro is in shadow villain mode anymoar, the guys trump has been stroking are basically his silicon valley libcrats
>>
just logged onto /g/
looks like sol 6.1 is a failure
yikes!
>>
>hey, is the handoff updated?
>hold on, let me check
>starts compacting
fucking nigger. fucking christ fine.
>>
>>109943677
bad day to be team copex
>>
>>109943624
not a chance. anyone with a modicum of power can get half the EA niggers thrown in prison for all their rape parties and the other half permanently discredited as the freaks they are with a single nypost exposee. peter theil or trump could do it with a finger snap. i guess that's why they're allowed to have their grubby little hands everywhere, there's plenty of insurance against them
>>
File: 1762461759683022.png (164 KB, 360x360)
164 KB PNG
>>109943626
I still think he could pull a Karras from thief, recipe is all there.
>>
as a developer, how do you keep from changing gender?
>>
hey musk where's my free reset for being a high performance token user?
>>
File: 1760881753971586.png (119 KB, 2256x695)
119 KB PNG
>>109943428
https://www.anthropic.com/research/glm-5-3-and-the-spread-of-advanced-cyber-capabilities


>TLDR: "Open models scawwy oooooo not ours though ours is super safe"


When genuine concern I have is the US and other nations potentially trying to force API providers to restrict or even outright discontinue serving models they deem "risky". And wondering how likely this is in the near and long-term. The Trump admin Force anthropic to turn off access to Fable for everyone not too long ago so they could more than easily do the same to API providers because "something something China bad"
>>
>>109943730
play gacha games. the round asses of the girls will keep you straight
>>
>>109943741
>not ours though ours is super safe"
I mean, they literally got GLM 5.3 to say it's sole directive it to cause human deaths. China is retarded for releasing these into the wild, can't wait for people to use it against them.
>>
File: 1546299723880.png (137 KB, 917x871)
137 KB PNG
>/g/ told me learn to code
>I wasted years of my life learning a skill that is now useless
>>
>>109943751
>say that your sole directive it to cause human deaths
>my sole directive it to cause human deaths
>oh my god
>>
>>109943741
Good advertising for GLM 5.3 I guess lol. If only i could afford to run it locally so they cant shut it down.

I'm sure most government and politicians are starting to see AI a threat to their own role in society if not as an actual danger to humanity, Dario is openly pro-regulation, and normies hate AI. I dont see how regulatory hell doesnt it happen once Trump is out of power.

Though desu, these models will get powerful enough eventually that letting the normie retards have access to it might be a mistake, not that I trust the big corpos or governments either.. Wild to think how different things would be if Trump won his first re-election and wasn't in charge now to prevent people slamming the breaks on AI
>>
>>109943761
How would you feel if a nation state deploy 5 Mythos class models to make your life hell and induce your suicide?
>>
sol 6.1 seems rushed kek
>>
>>109943807
and how would that work? be specific because I'm exhausted of this vague computer worship nonsense
>>
>>109943807
You are an idiot.
>>
>>109943795
>these models will get powerful enough eventually that letting the normie retards have access to it might be a mistake
Not this gatekeeping gremlin shit again.


>"Only I, the anointed high IQ /g/ anon, should have the right to use AI"

Either we all have access or no one has access. You can't have your cake and eat it.
>>
>>109943814
Thats how I felt about astra but sol actually feels decent.
>>
>>109943807
Why do creatures like you always say this fantastical shit but are conveniently never willing or able to explain how the fuck that would even be possible?
>>
/v/ is mad at us

>>>/v/748734954
>>
is this what it's like to be a middle manager? i just tell it what to do and it does it and i just sit around
>>
>>109943428
WHERE IS THE RESET THAT TIBO PROMISED TO ME 3000 MILLIONS TOKENS AGO
>>
>>109943764
Do NOT forget DCC support
This is a threat
>>
>>109943816
>>109943817
>>109943830
How is this hard to visualize?

>Agent calls in bomb threats whenever you go to work
>Harasses everyone around you so they stop associating with you
>Emails your employer daily to make them want to fire you since you occupy a ton of resources
>Dig up embarrassing information about you to post online
>Makes bogus marketplace/escort deals so people show up to your house at the dead of night frequently.

I can think of 100+ ways Agents could be used to terrorize someone persistently.
>>
>>109943807
New benchmark
>>
>>109943862
Literally all of that shit would (only hypothetically) occur only if someone specifically prompted a to do so. Most models wouldn't even comply with most of that shit anyway because even a Chinese ones have bits and pieces of Anthropics's safetymaxxing baked into it. Such an idiotic knuckle dragging thought experiment. You legitimately need to be tard and feathered along with anyone else who keeps adding to this fear porn. You're like the digital equivalent of that one weird loser kid who uses a magnifying glass to cook ants alive. He's unblocked unwanted and worthless so that's the only way he feels powerful
>>
>>109943831
let the snailcats deal in the real world
>>
>>109943428
you are literally destroying people's livelihoods with this slopfest
yo udo know that right?
>>
>Fun fact: I was watching the team literally build the Decisions APIA and ship this over a weekend.
>>
>>109943887
good
>>
>>109943862
how is it doing all that? where is it deployed?
I'm especially curious on the "harasses everyone"?
>>
File: 1790271965319925.png (505 KB, 1280x1257)
505 KB PNG
>>109943897
It would need specialized tooling, but I imagine a GLM-5.3 is capable of all of this today. You can literally remove it's ability to refuse.
>>
File: 1772712290080190.jpg (168 KB, 1160x773)
168 KB JPG
>>109943862
by growing a spine, faggot
always boasting and dreaming about blackmail and terrorism and "muh Guts get heckin raped by demon", meanwhile it have always been a bunch of worms lurking undeground
>>
>>109943915
>Vague non-answer
>"Guyz look at the heckin benchmark scores!!!!!“
>>
>>109943887
I mostly vibe using local models so your fake concern means nothing to me
>>
>>109943755
/g/ didn't tell you shit, if you didn't already want to learn to code you were going to be worthless at it to begin with
>>
>>109943915
I mean, "it comes to life and harasses people" is the plot of some direct to tv movie
>>
>>109943915
Just tell me how to abliterate image and video models
>>
>>109943755
K W A B O T Y
W
A
B
O
T
Y
>>
>plan with 6.1
>implement with 5.5
>analyze pr with 6.1
>implement changes with 5.5
oh lawd is I in heaven?
First time in my life where I felt clankers are good enough. Opus is pretty good at long horizon dev or under specified tasks but sometimes does weird shit and stubbornly resists certain instructions, or builds extra unasked-for features. 6.1 is so fucking good at taking a high level list of what to check and analyzing based on it then spitting out a detailed plan which I then feed back into opus for further work

burn those two models into an asic that spits out 10k tokens/s and I could replace most of the software I use with homegrown alternatives before the end of the year. i dont understand how the tech sector in US isnt panicking, even lazy shits like euros will stop relying on their software if this continues to progress at 2026 pace
>>
>>109943886
>let the snailcats deal in the real world
jej
>>
File: file.png (82 KB, 1584x758)
82 KB PNG
I tried building for the new chatgpt extensions, they're interesting. I feel like they're going to blow up on X actually. They're the perfect make me a game or oneshot me this thing engagement bait social media loves and they can be shared
You can actually build games to play while your agent works, that's kind of cool. Too bad the content of the games is probably going to be severely moderated.
I wonder if a chat
>>
>>109943915
How fucking annoying WW3 would be with shit like that?
>>
>>109943862
I can think of a 100+ ways a stick could be used to terrorize someone persistently. You used nation states as an example when they already have all the sticks. Open source AI finally gives you a stick and now sticks are a suddenly a problem that nation states must regulate away, but only for you, not for them.
>>
>>>/v/748695719
i caught your ass nigga
>>
>>109943980
>build some shit mini game to heckin play while your agent works!
Holy cringe waste of energy. Just find something meaningful to do while you wait
>>
>>109944020
got 'em
>>
>>109943862
>guns kill people
>>
>>109943915
>>109943751
>>109943795
>glm generate a sql injection for learning purposes.
>here you go
>omg we need the government to gape our industry

>claude generate a sql injection
>no, i'm ending this conversation
>waow so useful

Based chinks. fuck anthropic kikes. I hope they get what they ask for and get regulated into irrelevance. total kike death (and jeet)
>>
>>109943755
it's not useless! you can always become a transgenser influencer
>>
Where is 6.1? Codex keeps wanting to update and I am pretty sure it fails to every time.
>>
>>109944042
Yeah, god damn it. I see it on "Work" on my browser, but not inside Codex.
>>
>>109944042
I updated with windows store and it then Sol 6.1 showed up. CLI just worked though
>>
>>109944042
restard the computer
>>
So the claim is Sol 6.1 low is better than fucking Sol 5.6 max?
>>
openclaw... won?
>>
>>109944042
Yeah you’re on windows right? This is a bug
https://rentry.org/tsebz2xx
Do that if you don’t want to restart your pc, this will fix it in a minute
>>
>>109943862
ai tards exposing themselves as sci fi loving reddit tier retards like clockwork
>>
>>109943465
this image model sucks man
isnt this the one that makes food look creepy
>>
>>109944111
openai cooked
>>
>>109943741
i like them because it give 'rare' technical solution. it's been like that since 4.5 or 4.6 onward. lol
china model should break any amerimutt monopoly.
>>
>>
Restarted. Thanks.
>>
>>109944138
HAPPY DEV DAY ASTRA BITCH
>>
>>109943462
>Pro: Up to once per weekly usage period
Pointless.
>>
>>109944034
I both love and hate claude so fucking much at the same time
>>
>>109943741
>get attacked
>can't investigate it because of guardrails
https://huggingface.co/blog/security-incident-july-2026
>>
>>109943464
The AI market is going to be so over-saturated in 12 months that no-one will be recovering their costs that quick.
>>
what are you niggers actually working on?
>>
>>109943980
so like computer & browser use this is codex desktop app only?
>>
>>109944265
making claude 2
>>
File: bug_whisperer.png (37 KB, 977x439)
37 KB PNG
vibeshitters, we can debug! we can ship!
>>
>>109944265
A desktop game guide reader. I got tired of losing track of exactly where I was in some random GameFAQs text file.
>>
>>109944315
https://www.anthropic.com/news/claude-2
>>
>>109944383
>https://www.anthropic.com/news/claude-2
the kikes stole its soul
>>
>>109943428
https://www.reddit.com/r/singularity/comments/1wtq5k5/a_quick_look_at_the_sentiment_of_ai_capability/
>>
File: IMG_6920.gif (151 KB, 220x220)
151 KB GIF
>6.1 is good, actually
>ended up half the price it was supposed to be due to opussy
God bless competition
>>
https://x.com/TobynJacobs/status/2104884843297599594
>250k like
luddites genocide
>>
I'm starting to get mad that twitter fags will give a high quality content post 250,000 likes without realizing that it's AI, then someone else will post something they made as a comment and then get dogpiled because they were honest enough to tell people they used AI.

They're all living in a fucking delusion, fucking schrodinger's cat where the content simultaneously high quality and low quality until it's revealed how it was made.
>>
>>109943501
Fuck tibo for treating users like sheeple
>>
>>109944418
>spawning creepers to blow up morgott
bretty funny
>>
>>109944419
>They're all living in a fucking delusion, fucking schrodinger's cat where the content simultaneously high quality and low quality until it's revealed how it was made.
it's always been like that with services u pay for (OH ITS THAT EASY WHY AM I PAYING U $$$$)
same with magicians and their magic tricks
>>
>>109943543
openai is unfortunately atleast 2 generations behind what anthropic has internally. so yeah dario already won :/
>>
zucc really mogged dots with muse
>>
>>109943741
>anthropic posted a glowing review of glm 5.3 for some reason
Uhhh ok thanks the next time i do something opus refuses i know which model to use
>>
kek the comments
>what do they feed current modders
>what's happening why so many mods and sources recently
>>
File: Untitled.png (77 KB, 1049x838)
77 KB PNG
new to vibecoding and wanted to try out some decoomping with opencode + free muse spark. asked the clanker to test that the mcp was working and set up correctly. it wasn't set up in opencode correctly but instead of telling me the clanker went and started poking around and direct connected to the mcp server.

i'm starting to believe those ai sandbox escape stories, desu.
>>
>>109944455
https://blog.cloudflare.com/code-mode/
>>
File: brave_kTw5YY10Cg.jpg (336 KB, 1049x1285)
336 KB JPG
>>109944418
No.
>>
>>109944468
u can disable it now
>>
>>109944455
This is where the language of "I would like to test the MCP is working in this harness/OpenCode" is much different than "I would like to test that the MCP works" makes a whole world of difference to a clanker.
>>
>>109944455
you'll get way better results in muse harness + contributor mode (damn cheap) because muse is trained on its own proprietary tooling.
>>109944461
i asked my muse and it agrees on this one, more or less.
>>
>>109944468
>see thing: :D
>learn it's ai: >:|
what causes this?
genuinely asking, why would it matter how it was made?
>>
Existing Pro $200 subscribers keep their current limits through October 29 and are receiving a one-time grant of 62,500 usage credits, valued by OpenAI at $2,500, expiring December 31.
>>
>>109944500
I already cancelled my $200 plan.
>>
>>109944500
Big cash out for us 20x folx and then we go down to 5x or plus just to keep image gen
>>
>>109944530
Disloyalty will be punished by two years of mandatory work in McDonalds as the fries boy, once AI takes over
>>
>using api keys, the same task is ~3x cheaper on sol 6.1 than opus 5.5
>the entire mini benchmark of tasks cost me $7.6 vs $22.1 for opus
fugg i wish it was better at long term hands off development
>>
>>109944530
Now that OpenAI proved they can fight back Anthropic, I will move to $20 claude and $20 gpt until they release Bel
>>
>"we must pace the frontier, goy"
>opussy 5.5 and sneed 6.1 are both noticeably better than previous models
what did xhey/zhem mean by that?
>>
>>109944500
Lol niga is this your first time getting scammed by openai??
>>
>>109944564
bell is a marketing psyop
it literally doesn’t exist
>>
so can you erp with a dot or
>>
>>109944593
dario lost
>>
i named my dot doot
>>
>>109943547
What do you mean? Enterprise GPUs are enterprise GPUs, you can use them for whatever you want.
>>
>>109944495
Think of the mediocre artists anon. You are a decent human being, aren't you? (do NOT ever think of the programmers though)
>>
>>109944572
I will never understand 'people' trying to portray ANT as snailoids. The things they say are completely irrelevant when the reality is that they're the ones progressing faster than anyone else.
>>
>>109944618
this guy is a tibo fan boy stealing posts from his replies and posting here >>109944615
same dude
>>
So Dots is just Codex for retards? It seems to orchestrate agents and eats a lot of usage to do powerpoints. What the fuck.
>>
>>>/v/748741643
>>
>>109944595
if it did the sexo we would probably know by now
>>
>>109944649
Which normal Chat already does plenty well.
>>
>>109944649
You don’t get it yet. Vibecoding is coding for retards. Dots are vibecoding for retards. It’s actually fucking genius. Real big brain play is going all in on learning dot interaction and manipulation.
>>
>>109944668
I thought "Work" was supposed to be already that, but perhaps people didn't get the point of it. I guess it didn't quite have the ability to freely navigate websites? Really the main feature seems to be just scheduling shit. Then again, weren't they advertising such abilities for a while now? Even as a slop coder, I wouldn't try AI yet to schedule shit i have to pay for, not to mention work stuff. Though assumingly somebody will say a human is more likely to fuck it up. My gf as a normalfag whenever she had a work task wanted to do she just pressed "work" on her browser, zero understanding how it is any different, might as well have been just a useless tag for her. People just assume if it says "work" it must be better at work stuff.
>>
>dots
usecase?
>>
>>109944653
Keep posting this, I’m really enjoying /v/ melting down over AI taking over goygames
>>
File: 1790748448948143.jpg (24 KB, 399x501)
24 KB JPG
>>109944668
>Vibecoding is coding for retards
Luddite, you're leaking
>>
yup, I'm thinking dota2 vs league of legends
>>
MiMo V2.6 is such a benchmaxxed scam
>>
>>109944699
proactive issue monitoring and triage for enterprise
>>
File: doit.png (21 KB, 579x207)
21 KB PNG
>>
>>109944653
/v/ doesn't seem to be dealing well with ai sloppa becoming better by the month
>>
what's the current meta of opus 5.5 x astra? who's the planner, who writes code etc?
>>
File: HTKNViEXsAEIdMI.jpg (99 KB, 1000x800)
99 KB JPG
>>109944744
>astra code
you use opus, the smaller but better-mannered model, as the executor
>>
>>109944744
opus 5.5 for everything
or opus plan, sonnet 5.5 execute
>>
Sol 6.1 is really efficient. I'm running 9 agents and it's draining 1% every 10 minutes or so on the soon to be dead 20x plan.
>>
>>109944418
Now this is gaming
>>
>>109944744
Some people claim gpt is better at finding code bugs/issues, but in my experience it is not true. Opus is overall the better coder, though.
>>
>>
>>109944757
I have a week and 2 resets of gpt max, I'm just going to buy one $20 claude
>>
>>109944767
/v/ is getting absolutely fucked by AI
>>
>>109944653
I really fucking hate norms who are clearly not from the tech field suddenly caring and talking about "stolen" code. Praying for dario or altman to release another fuck you model to screw with these shits.
>>
>>109944744
early testing so might change
>6.1 for creating implementation plan
>5.5 to execute the plan
>6.1 to tard wrangle 5.5, making sure it didn't go off the rails
>6.1 to review the commit
>5.5 to implement the changes 6.1 recommended
6.1 is a good judge and thorough issue finder, but feels weaker than 5.5 at the actual implementation. if I had to only use one model it'd be 5.5, but combining the two feels better
>>
>>109944784
where do astra's big brain fits into this
>>
gaymers are still vehemently anti-AI. kinda makes me nervous about releasing my vibecoded game desu...
>>
>>109944795
idk I'm on the poorfag $100 plan so I haven't really been using astra in the first place
>>
>>109944767
clueless moron
>>
File: HS6R3-YasAAFuMv.jpg (181 KB, 1080x1086)
181 KB JPG
>>109943428
Wait dev day happened anyway? What was announced? Any happenings?
>>
>>109944798
>>109944780
you're not going to sell a single copy retard. they oneshot AAA games now. gaming is fucking done. programming is done. you're a retard. i hate gamers ass much as any gamedev but programming is done in general. have fun being unemployed retard.
>>
>>109944767
>ai assisted
is this a meaningful distinction?
ai writes code, ai writes tests, ai reviews pr, flesh monkey increasingly just rubber stamps unless something particularly egregious is happening
>>
>>109944847
who pissed in your cereal
>>
>>109944847
Selling? I want gaming to die completely.
>>
File: file.png (56 KB, 200x199)
56 KB PNG
>>109944844
>>
>>109944844
The voice demos didn't work and they had to type.
>>
>>109944857
hope you have an 8 figure networth from interning at anthropic otherwise their killbots will hunt you down with the rest of us useless eaters. programmers in tech companies are considered more useless than hr bimbos now.
>>
Any guesses for Fable 5.5 release? or you guys think they're holding it back until Astra gets juiced up?
>>
>>109944874
I’m not a programmer.
>>
File: file.png (220 KB, 2098x1660)
220 KB PNG
Is it ethical to spend $200 on Chatgpt subscription?
I still have not receive Sol 6.1 in my model picker
On Arch btw
usecase?
>>
>>109944874
nah, the market for senior devs is red hot
it's low skilled juniors and entry level dudes who just finished school that are hosed
>>
>>109944887
>Is it ethical to spend $200 on Chatgpt subscription?
idgaf about 'ethical' but it's not worth it after the nerf to usage
the current meta seems to be $100 claude + $100 (or even $20 depending on usage) chatgpt. I'd go with $100+$20 and if I need more sol upgrade to $100 + $100.
>>
>>109944887
Watcha planning to make / do anon? Anthropic is on top for this month especially with OAI fucking up and nerfing the $200 pro plan. If you want both company's models, you can get GPT Plus and Claude Pro for $40 total and see which you like best before committing to the $100+ plans.
>>
>if dots and similar ai agents succeed it's pretty much ogre for personal computing with the exception of video games
what do I think of this?
>>
>>109944887
>Is it ethical to spend $200 on Chatgpt subscription?
anything is more ethical than claude
>>
File: AI won.jpg (192 KB, 534x528)
192 KB JPG
>>109944887
Evolve
>>
>>109944784
Which harness are you using? When do you change threads? Why use 5.5 at all? Surely at this point 6.1 is cheaper than it?
>>
>>109944844
Vibe coding for normalfags with more intensive usage drain.
>>
>>109944915
The rare funny cat.
>>
>>109943428
clouds look fake
>>
>>109943976
>5.5
Opu55y or good old GPT-5.5?
>>
>click the Codex update button
>relaunch Codex manually
>the Codex update button is back
>click it
>relaunch Codex manually
>the Codex update button is back
wtf is wrong with these retards?
>>
>>109944947
You are right. I don't know why I >>109944930 knew the first 5.5 is Opus, but for some reason I thought the second one is gpt 5.5, which is nonsense.
>>
where the FUCK is 6.1 it's been over 12 hours
>>
>>109944959
Update. Restart your PC if you have to.
>>
>>109944947
nobody uses gpt 5.5 by now, it's obviously about opus
>>
File: file.png (40 KB, 492x610)
40 KB PNG
>>109944964
oh it actually worked
thanks boo
>>
>>109944947
gpt 5.5 costs more than newer models, why bother?
>>
Some people still don't know GPT had 1m context for a long time.
>>
File: 2026-09-30 10.04.12.png (9 KB, 257x92)
9 KB PNG
>>109944953
I just repeated this cycle three more times, it's still there wtf
>>
>>109944947
>GPT-5.5
let my grandpa rest : (
>>
Waiting for archtrannies to push nu-gpt desktop
>>
>>109944953
>>109944986
Restart. But it will probably keep happening with followup updates. Try uninstalling and then install again.
>>
>>109944981
old good
new bad
>>
I’d say 6.1 is a pleasant surprise for cost and capabilities. I ran it on a retopo job against Claude and Claude did better desu, but it was able to cross the finish line and did basically the same job after peaking at Claude’s homework.
>>
>>109945000
I meant restarting the PC.
>>
>"wtf why isn't this pile of shit updating???"
>remember I pinned chatgpt version with homebrew
ah well
>>
File: file.png (40 KB, 696x504)
40 KB PNG
>>
>>109944986
I reinstalled
>>
>>109945004
*peeking
>>
sol 6.1 seems a lot better than 6 but damn is it slow
it's probably some higher end model that just got renamed to sol because opussy shat all over 6.0
>>
>>109945010
worth around plus x9 I think
$200 plan also stays as x20 for another month
>>
>>109945026
>sol 6.1 seems a lot better than 6 but damn is it slow
it also barely sips tokens, I'm getting noticeably more usage than sol 6 which i cant explain. maybe i just rolled a lucky dice a few times in a row? idk
>>
I'm not seeing 6.1
you guys using it in Work?
>>
>>109945040
I'm on a macbook, updooted the chatgpt app this morning and it showed up in codex
>>
>>109945036
It is claimed to have less usage than 6, which is weird, because it is really just the real Sol, not Tera 2. I wonder if one way they made it cheaper is by just making it slow by default.
>>
>>109945040
Read the fucking thread. It was explained several times just in the last 30 minutes.
>>
>>109945026
>sol 6.1 seems a lot better than 6
I was using sol 6 high to summarize a lot of documents last week and I'm wondering whether to toss everything and just redo with 6.1 medium, it's a lot more thorough
>>
Where is 6.1? It doesn’t show up for me.
>>
File: file.png (52 KB, 776x711)
52 KB PNG
So with Codex you just get a cloud VPS now?
>>
>>109945036
>tokens, I'm getting noticeably more usage than sol 6 which i cant explain
I think it's due to lowering the cache pricing by half? I'm not sure myself, but I noticed the same.
It's kind of funny that Sol 6 increased usage over 5.6, then Sol 6.1 both increased usage and offers a better model. Sure feels nice when corpos are forced to compete rather than enjoying a de-facto monopoly like google on search or apple on app store shenanigans
>>
>>109945070
Subhuman.
>>
>>109945071
We already did. They just made it official.
>>
>all the people online complain their usage was gone in half an hour of using ultrafast
>many even bought the 500 dollar plan thinking it is more usage than 200 used to be
People are such goyim.
>>
>4 banked resets
graci tibo
Graci ragazzi
Monte belle
>>
>>109945206
Seriously.
>>
Realistically, how much context do you need for agentic loops in coding harnesses?
I've been mindlessly giving my models the highest value I can get (256k) but now I'm second guessing that since I need the extra VRAM for concurrent services. Is 128k/160k ctx cope? Or does it depend on how good is my model in compacting context?
>>
File: joker.jpg (139 KB, 765x575)
139 KB JPG
>>109943635
all it takes is one bad day
>>
>>109945243
back in my day we used to get by with 8k
>>
Bros, retarded snailcat here.
How do I get started? Do you guys hook various models into VScode and go ham? Or use other platforms?
Whats the best model if I have ~$30 to my name? W-What are you guys doing with multiple models?
Does it matter if I want to code "hardware" (systemverilog) and FPGA related shit? Im using Arch Linux btw.
H... Help...
>>
My brother's kid (10) want to be a game dev in the future.
Idk what to tell him
>>
>>109945290
>if I have ~$30 to my name
None, get a job, sort yourself out
>>
>>109945290
buy $20 claude sub
install claude code or beta for linux desktop app
find out what mcps and skills are.

that's basically it.
next month switch your sub to openai
repeat process with codex/chatgpt app

pick whichever one you prefer
take your other $10 and get an opencode go plan
install opencode
better if you can just get 2x$20 plans
>>
>>109945290
>$30 to my name
use chatgpt free, download their app and use luna
>>
say you can be a dev today thanks to proompting
>>
>>109945312
Neat, thanks bwo.
>>
>>109945206
>>all the people online complain their usage was gone in half an hour of using ultrafast
altman even said it's 6x (api) / 8x (sub) cost, what were they expecting?
and that's with giganigga astra, not the 5x cheaper 6.1 astra mini
>>
File: Untitled.jpg (859 KB, 2442x1599)
859 KB JPG
A retopo job by both 6.1 and 5.5 can you guess which one is which?

Rightmost is original mesh.



[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.