A general for Frontier Chads, Opencode builders on a budget. Pick your model and talk shit.## What “vibe coding” is, and how to do ithttps://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 sohttps://developers.openai.com/codex/clihttps://claude.com/product/claude-code## Mid tierhttps://opencode.ai/https://cursor.com/docshttps://antigravity.google/## Open / local / self-hostedhttps://github.com/OpenHands/OpenHandshttps://github.com/QwenLM/qwen-codehttps://github.com/QwenLM/Qwen3-Coderhttps://huggingface.co/bartowski/Qwen_Qwen3.6-35B-A3B-GGUF----## Prompting / context / skillshttps://arps18.github.io/posts/claude-code-mastery/https://simonwillison.net/guides/agentic-engineering-patterns/using-git-with-coding-agents/https://github.com/mattpocock/skills — /grill-me is a favorite## Other editors / terminal agents / coding agentshttps://x.ai/clihttps://docs.windsurf.com/https://docs.cline.bot/https://docs.github.com/en/copilot/how-tos/use-copilot-agents/coding-agent## UI/Frontendhttps://www.figma.com/make/https://www.anthropic.com/news/claude-design-anthropic-labshttps://uiverse.io/https://ui-ux-pro-max-skill.nextlevelbuilder.io/https://stitch.withgoogle.com/## In-browser builders / hosted vibe toolshttps://bolt.new/https://replit.com/https://docs.github.com/en/copilot/tutorials/sparkhttps://v0.app/docs## Benchmarks / rankingshttps://www.tbench.ai/leaderboard/terminal-bench/2.0## Previous thread>>109003462
>>109013551I imagine that's the reason, I just need to come up with some ways to use up more before it resets so it doesn't go to waste.
>>109013584I just got done watching the keynote and Platforms State of the Union Seems like Apple is moving fast and fixing things
opus 4.8 high seems very token efficient
>>109013577Not sure how many projects you have but starting a fresh session & having it review a project, then having it verify all the memories are still valid will burn through a good chunk of tokens
I have a personal project with a backend (python) and frontend in two separate repos. I'm running a Claude instance on both, how can I let the instances "communicate"? A shared MD file between them?
>>109013560where can i play these wonderful snailcat games
>>109013753just wait until the end of this development cycle
Does anyone here actually make money from a vibecoded app or project? I see tons of people claiming they've "created" stuff but 0 people showing off their earnings. Weird.
>>109013783I've earned a modest amount over the past year or so with automation tools for games on offerwall sites. somewhere in the range of 7-8k.
>>109013783The only time people even dare show what they "created" at all, let alone any earnings, it always turns out to either be profoundly broken (doesn't even compile level) or to be yet another notes app. And the worst is it's always "trust me bro" bullshit.
>>109013783My work pays for my codex max sub, feelsgood
>>109013560why don't you vibe coders make a snailcat game so I can play it. I keep seeing these box arts and it makes me sad
the resident earnings schizo is back
>>109013851Because sadly, AI can't do anything like this. It would take 3 years for them to make a shitty buggy slow as balls web-only snailcat game, while even a noob could do a pretty fast 2d native game in 3 months.
>>109013874post your game. you are a yesdev, right?
I would like to see an actual vibecoded project. Mine is ai-assisted but I wouldn't say it's vibecoded, by a long shot. I'm curious.
>>109013895How do you quantify it?
>>109013783these questions are quite useless because nobody is going to dox their golden goose. if i stumbled upon a money printer, i wouldnt tell a soul about it
>>109013783 I typed /fleet runescape and I sell the gold
>>109013897Well, the backend is about 80% coded by hand, and the frontend is 100% ai-generated. But the frontend is simply an user interface for the backend (a Flask API), so the core features have been created by hand.I guess everyone has their own definitions of what counts as being vibecoded.
>>109013904ill show you mine once i get this FUCKING CUSTOM RETARD FUCKING BLUEPRINT FAGGETRY done holy fucking shit kill me.it's 100% vibe slop.it's also pretty cool i think.
>>109013940What is it?
>>109013851
>>109013957its just a local AI user interface.i do a lot of diy hobby stuff and i despise the popular shit so i decided to have claude make me my own.it's got a lot of features now and what i am working on right now is having the blueprint system on the right side summarize the chat into a prompt that then gets fed into an image generator which then gets fed into a mesh/model maker which then gets fed into a custom python script to take that mesh and turn it into a cool wireframe svg file, so i can have a visual representation of what im working on and even export an obj file for cad or blender.shits fucking complicated though and i am struggling with claude due to linux amd python shenanigans.
>>109013988>its just a local AI user interface.Called it.
>>109013988>its just a local AI user interface.Oh, I thought it was something useful. Nevermind.
>>109013715i'd like them to be productive, just trying to think of some things I could schedule. Maybe some email cleanup stuff?>>109013783I think there are some people making money with AI slop on Instagram. Those people making the dog videos, for example, look up "coco the dog" or something, it's an AI dog and all the childless 30 year old women enjoy it
June is going to be absolutely fucking insane.
>>109014018By project we usually mean like software. I don't know how profitable ai video slop generators and the likes actually are but I would not be surprised if those work to some extent, especially due to scale.
>>109014024If they can get gemini to be at least GPT 5.4 level, it will be a huge win.
>>109014024>geminiThis time for sure will be the actual good model! No, not that one, then next one!
>>109014034Still waiting for gemini to hit gpt-3 level desu senpai.
>>109014072for me, it's gemini. because every other model thinks like a normalfaggot. gemini is the retarded cousin who doesn't actually want to let you down.
>Claude Fable
>>109013961I stood out of my chair and said kino
>>109014030I imagine a lot of the more pro level AI social media slop is custom software generated, probably very poorly made stuff that just gets the job done. I don't think there is a huge amount of people making money off the little one-off apps that they shit out, but maybe I'm wrong.
>>109014178I know for a fact they just prompt the frontier models and never use tools. I worked in or close to AI tools sales for a while in various spaces and it's exactly the same story every time.
>>109014186must be a good racket for now there is a lot of it
>>109014220One would expect. But could also just be 3rd worlders who make $100 across 10000 videos for example, using stolen keys to not spend too much.
>>109014186I thought meta and tiktok were clamping down on that pretty hard?
>>109014281instagram is full of slopgirls nowa lot of it isn't full ai, but faces swapped on borrowed vids n shitmost of the comments are horny dudes who can't tellit's getting harder to tell with that stuff tb h and this is bottom of the barrel stuff
>>109014292I guess that makes sense, especially when considering as a concept short form content like that is inherently slop
new slop word to add into your daily slopping "Atomic"
If you have two Plus plans you can do GPT-5.5 High.If you have one Plus plan you can only do GPT-5.4 High.That's just how it is. 5.5 uses twice the usage.
I'm at the point where I basically just let Codex crack and reverse engineer apps and live service games, unlocking fps, patching kernel level anti cheat out of them and giving myself infinite premium currency by spoofing payloads. People don't know what's coming for them man. It's absolutely hilarious and is gonna get even more wilder from here on.
>>109014388Whoa, you got in-game gold? You might be the person closest to earning a profit from AI in this board, lel
>>109014388If you said this to a popular youtuber they'd call you delusional, they just straight up don't believe any of this, or they think it's "coming in the future" and isn't here now.I suspect things will only change once the free plans are as good as the (current) paid plans. None of the doubters want to pay.
>>109014396I tried to get a bounty but chinese devs are greedy assholes so fuck them. I'll just sell these accs.
>>109014388If that were true you could make shittons of money selling those cracks. It's fake though.
>wake up>no codex resetMe sad...
>>109013874Nah it wouldnt take that long, it could create a "demo" in a day but it would have 5m of gameplay.
I tried this out and now I have thousands of lines of code generated, I wouldn't know where to start if I need to debug anything.
>>109014566You would start in the Codex or Claude Code prompt box
>>109014566>I>debug
>>109013795that's probably illegal, anon
>>109013899>if i stumbled upon a money printer, i wouldnt tell a soul about it>Hey anon, where did you get all these videogames, gaming PC, new car, new apartment, new clothes etc?>Uh, I found them lolDid you know that one of the reasons that people get caught committing fraud etc are because they tell family members who rat them out to the authorities?
>>109014563If that's the level we're talking about for 'game' then yes, absolutely.
>>109013560when can i play snalecat
>>109014612well, when i said "money printer" i meant a lucrative legal business, not being a criminal.otherwise, yes i would probably agree, the biggest concern for someone with newfound wealth is finding the self control to keep things discrete.
Open AI is surely taking their time with 5.6 huh?
>>109014676possibly waiting to see how anthropic is going to play their cards.and perhaps anthropic is doing the same for open ai, lol
>>109014685> ChatGPT 5.5 get mogged by Opus 4.8> OpenAI: "Let's wait and see how Anthropic play their cards"???
>>109014697isnt 4.8 an incremental update over the previous? i got the impression it was a nothing burger.
>>1090147104.8 is worse than 4.7 is worse than 4.6 which was peak for anthropic.But gpt 5.5 is worse than 5.4, which is peak for the 5 series with 5, 5.1 and 5.2 each being worse than the next and 4o mogging all of them.
>>109014734bollocks
>>109014596violates terms of service probably
>>109014734true, and ChatGPT 5.5 Pro is worse than ChatGPT 3.5, which is worse than Google Bard
>>109014816they'll ban your payment accounts and flag you for stuff and then you're fucked
>>109014860oh no
>>109014880Not my fault you're too lazy to work extra.
>>109014889I wouldn't be doing it if it wasn't automated. The hourly on those sites is abysmal.
>>109014903Just do data annotation or user testing work, fuckthey pay like $20-$40 an hour.
>>109014612>I do software development >what software>enterprise stuff, payroll management, that sort of stuffThere, done. Nobody will ask anything else after that And if I did tell people about it sure as hell wouldn’t tell them it’s made with AI
>>109013672fucker nuked 30% of my monthly allowance on planning a 50 lines script…zero shot solution, though.
>>109014903>>109014912Also the hourly rate sucks UNLESS you treat it like actual work.If you sit down for 2 hours and just focus you'll get money.
>>109013783I made a, cool app that does subtitle generation and then alignment and embedding on any video.I like using it for myself for English Dubbed Anime since all subs are from the Japanese audio.It even cross references the wiki for character names, etc.I don't know how to market worth shit though, so hasn't made anything.
> ChatGPT Pro 5X plan> 2 days until reset > 20% leftFucking hell here's the fucking reset bros???? Codex crashed once for me, I deserve a reset and refund.Tibo???? Scam Altgirl???
>>109013783Zero dollars so far, but I’ve created a habit building up my portfolio bit by bit every day. I’m also getting a sense of what works and what not for each frontier model.
>>109014934>monthly allowance
>>109014941Well, the first step in marketing is not telling us all what your app does, because that sounds pretty cool and I think I'll make my own now.
>>109013838jelly. they monitor convo history at my company so I can only use my sub for work-related stuff. I have a separate sub for my personal projects
>>109014941Can it handle japanese audio to sub in english? I have several episodes of Konpeki no Kantai that miss their subs, and audio is japanese only. I think that you could market this as a web service.
>>109014934That makes no sense. Free account? You can’t reach 30% of monthly usage on a single 5 hour session. Are you saying it worked for days?
>>109014018It does seem that AI content for social media is the shortest route to money outside of dumb digital products making pennies. The barrier of entry is automating it in the most cost-efficient way possible without sacrificing too much quality (video gen is fucking expensive to iterate, I tell you that). Also, you need a fine-tuned LoRA for brand-specific stuff so the models don’t hallucinate random shit on your target products.
>>109014941scale it and sell it to CR. there you go.
>>109014966Not like it's making money anyway. Besides, Im an actual software dev so I know the tricks and guided the architecture myself that I know a non software dev would miss. >>109014988It can actually do that yes. Around 50 some spoken languages to just about any written language you can think of. I did build it as a web app yes. You can drag and drop any video file into your.browser and be able to download subtitles raw our your video back with new subs embedded.
>>109015010I actually had someone who knew a contact at CR and another at Netflix. Both leads went cold, sadly.
>>109015020That’s why you need scaling. They will look more interested with a proven business model and a small set of satisfied users.
>>109014958>>109014546The resets are over, dumbasses. They were done for PR, but they've come up with a new PR method that doesn't cost them anything.
>>109015033Nah you're right. Granted it didn't get that far with either. I agree with you and what I told my friends. Both insisted they get me in touch though.
>>109015034no fucking way lmao.it's so fucking over, Codex bros...
>>109015034How do you get onto this faggots "big button" list?
>>109015051you probably have to share your slop with them on shitter
AI just oneshoted a windows culture bugimagine debug this shit manually lmao>>109015034I think it's a reset every 1M new users
>>109015051The first guy they chose is a former employee of every major AI shop out there making an AI powered web design canvas app so you get the idea. You probably need to have Sam’s pozzed cock before you’re even considered for it
>>109014912THOSE JOBS ARE IMPOSSIBLE TO GET
>>109014734>gpt 5.5 is worse than 5.4, which is peak for the 5 series with 5, 5.1 and 5.2 each being worse than the next and 4o mogging all of them.peak female take
>>109015069>I think it's a reset every 1M new usersyes they're doing that every 1m up to 10m i think.but they also reset randomly when people complain about increased usage and shit - happened while they were making tweaks to cache and i think probably trying different quants along the way.
>>109015081well, consistent web design is the last frontier to breach with AI. UI work is still sloppy as hell with coding agents
>>109015095They're not, your problem is you are demanding and unreasonable, not your work skills.Do you think they'll hand out those "free money" work from home gigs to you when they know you'll just drop them, shit all over the walls, then go back to your corporate job?Literally all you have to do is calm down and put yourself in the shoes of somebody who isn't as "privileged" as you, literally complaining you can't get even more money on top of your full time employment,
Didn't Anthropic just release Opus 4.8? Why would they release Mythos today.
>>109014997I have a maxed out team plan from work with anthropic and it usually eats through the entire monthly quota (not 30%, the entire fucking thing) in about 3-4 hours if I actually use it (which I never do because it will invariably end midway in a state where it's impossible to go through the slop code but the solution doesn't quite run yet because it needs a few more prompts to get there before a refactor can be done).
>>109015213have people been saying they're going to?
not bad.>>109015247today>>109015213in fairness they don't really have a product equivalent to gpt-pro models, so there's space for it
>>109015014That's quite cool, is there a way that you could generate subtitles if I give you the episodes? You'd be a life saver.With some marketing you could easily sell this as a service. Do you perform the computing in your own machine or are you using AWS?
>>109015144So is every other kind of coding work with ai agents, if anything UI is what it does best out of all coding tasks.
>>109015249Cool. Wonder if they'll do a reset with it.
>>109015245This. Cursor is more cost-effective in that regard>opus 4.8 for planning and code review>composer 2.5 for implementationBoom. Roughly the same quality at a cheaper cost.
>>109015249>2x the price of Opusoh boy
>>109015250Drop me a torrent or whatever and tell me which episodes. I'll see what I can do. Check this thread or follow up ones.I use a mixture of AI vendors and VPS hosting. Could go into more detail but I don't want to give up the entire secret sauce
>>109015245Unless team plans are different and you get unlimited 5hr sessions, you’re just spouting bullshit It’s impossible on pro or max plans to spend 30% much less all of it on a single 5 hr session.
>>109015312Here's the torrent, you'll see subs are missing from episode 25 onwards:https://nyaa.si/view/2101288If you manage to generate subs you're a godsend.
>>109015260It’s not that bad considering opus 4.8 is quite usable on even a pro plan. Maybe fable will be the undisputed planning/design model
>>109015467I don't know the details, I just know the facts. Clearly it's not impossible. I'm not even pointing it to my repo or my local folders because that sounds impossible, I just point ~500 line scripts at it.
now that we have LLMs and projects like GhidrAssistMCP, we can simply reverse engineer the source code, right?
>>109015550If only
Mythos today? I'm hype.
>>109013783Yes. 10k in bounty payments alone. Probably more in freelancing work automation but hard to put a number on that because it's a speedup not a total automation.
>>109014024Mythos been kinda lackluster desu. It's just Opus but with more self awareness and detail orientation. Which is definitely useful, but nothing close to the hype.
Opus flirting again
>>109015777Fable today, Mythos never
>>109015868The fact is fable IS mythos, not a crippled version of it. they just say the latter for plausible deniability when everyone points out it's no better than the last 4 versions and this is a very fucked up result because it means we're through with AI advances now. No longer in a diminishing return regime, we're walled off completely. It's over.
>>109015882I'm optimistic that it will be significantly better, but of course not as good as they first claimed.
>>109015924They didn’t claim anything about 4.8 and it mogged 5.5 so bad they pushed back 5.6 I’m not saying fable will be good but anthropic is on a roll after their early year fuckup
>>109015943>4.8 and it mogged 5.5lmfao
>>1090159435.5 is kingI even get Opus from my company and almost never use it
>>1090159434.8 is dogshit, only reason it mogs 5.5 is that 5.5 is even more garbo. It's been a race to the bottom all of 2026 so far. Pure amateur hour on the infra side as well. Never seen this from commercial companies before.
>Newest thing is dogshit>No this newest thing is dogshitIs this like negative wish casting?I mean 4.6 did feel better than 4.7 and 4.8 but the only model that got much much worse is Gemini 3.5 Flash. Everything else be it chigger or mutt model is slightly better than the previous thing.
https://www.anthropic.com/news/claude-fable-5-mythos-5
>>109016031Opus 4.8 feels like the best Opus in a while to me, but during 4.7 times they apparently also had some bugs, which they also publicly admitted, so maybe I was using it while those were still there.
>>109016031You are wrong. 4.8 is a bit worse than 4.7 which is a bit worse than 4.6, but the price and response times jumped, so it really is dogshit. As you say in this track it's only (removing all other attributes) a bit worse, not the end of the world.But as for gpt-5 series, 5.5 performs about on par with FUCKING 3.5, it is INSANE how terribad it is.3.5 is not as bad a downgrade by comparison even though it was a big jump, too. But scroogle has been retarded the whole time, with gemini-2 being the best performing model they've ever released so far (confirmed by google themselves in internal testing).
>hello vibefrens I'm also a vibesister too, realsie vibey in here!>gpt 5.5 is so bad, right? it is so worse than gpt 3, right?
>>109016033HYPE!But where is my reset?
>But as for gpt-5 series, 5.5 performs about on par with FUCKING 3.5, it is INSANE how terribad it is.Did you use 3.5? 5.5 Is so much better than 3.5 it feels as if you're just fucking with me or being contrarian .
vibe coded slop burns houses downhit the books you lazy numpties
>>109016086thats what im saying
>>109016033>From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.>On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window.>After this point—when sufficient capacity allows us to do so—we aim to restore Fable 5 as a standard part of subscription plans. We intend to do this as quickly as we can.????????
>>109016098api is their money printer
Holy megajew move from anthropic
>>109016033>fable>mythosAnd fags here are surprised they lied lmao
>>109016106I was going to subscribe to claude just to try out mythos, but the fact that they're removing it after a few weeks is insane. no thanks. not even people who make 150k/yr are willing to pay the outrageous api prices.
>Fable is from the Latin fabula, “that which is told,” akin to the Greek mythos. The safeguards are what distinguish the two models (Fable and Mythos) and are why we’ve given them different names.So apparently fable IS mythos, the difference is "only" the safeguards. No performance downgrade or anything.
>>109016095By 3.5, I mean gpt-3.5 and not gemini-3.5Gemini-3.5 is impossibly bad. I don't even know what to compare it to. Fucking gpt-2? I think that was even better than gemini-3.5 is.
>>109016098They're going to work hard on quantizing the model into total uselessness before they make it generally available huh.
>>109016176>gpt-2You sound like a zoomer talking about the Amiga.
>>109016098They're preparing for an IPO so they need to financially engineer a quarter of profit so people will buy their bags
>>109016164They popped up some weird messages about how they'd switch it "safety" flags a message, I couldn't figure out what they were getting at
>>109016207I've been thinking about how I should feel about this comparison but I'm not sure. At any rate, it's very amusing.
>>109013783Nothing, not even a single donation even though my programs (a lot of it not even vibecoded) are used by hundreds of people.
I'm getting mythos fomo, but I don't want to burn another $100 when I'm already paying $100 for codex. $200/mo just for hobby slop is a step too far for me.
>>109016106It's more like they've gotten people used to always having their expensive model for free and want to switch them to pay per hitThe irony is they poison the quality of their product by injecting insane warnings to the model anytime it reads a file
>>109013783My personal project is free and open sores, I get paid pretty good money to vibecode at work though.
>>109016234gpt2 didn't even have RLHF so it was more akin to programming in binary with punch cards
Quick yall, my PiClaw is in cold storage, what's a good quick BOYK harness I can test Fable 5's agenticness with?
>>109016275RLHF has only ever fucked up the models bad. Significantly reduced performance, huge increase in sycophancy and verbosity (noted in many papers -- voters always preferred the longer factually incorrect response to the terser factually correct one), and refusal to follow instructions.I'm no longer amused now that I know you were only projecting.
>>109016292gpt2 was amazing for creative promoting but I would fucking hate to have to larp everything just to get it to shit out some code
>>109016254You can try it on the $20 plan. It won’t last you a lot but neither will fable in general
https://x.com/cursor_ai/status/2064394824313376787?s=46Apparently fable low is equivalent to 5.5 xhigh but also more expensive but on higher thinking modes it’s unrivaled. It also seems fable medium has the best cost/benefit
I rarely need a "smarter" model than codex xhigh and when I do it's for design or analysis and I use chatgpt pro which thinks for like 30-90 minutesThe same research task given to opus never runs more than like 5-10
I vibecoded my way to boredom and melancholy. I'm an accountant for a small business, and I'm the third in line so they just give me all the shit work. Except they do business like it's 1988, I get faxes, hand-write checks, roughly 90% of my job is copying information into spreadsheets. So it's been trivial to automate 90% of the job. Even the faxes are automatic now, they get forwarded to my email and they're handled with OCR because they're always the same shit, just payroll info that needs to go into a spreadsheet. They fill out tax-forms by hand that our payroll software generates automatically, which was spooky black magic to the other accountants when I showed them so now they just have me do it. They're both twice my age and I'm old enough for colonoscopies and heart medication. My boss types with one finger and his "keyboarding glasses". I'm happy to spend most of my day fucking around on side projects, but I know there's zero opportunity for advancement here. I'm not going to replace those other accountants, when the core team dies off the business is going to die with them because it's built entirely on relationships formed 6 generations back. The work of all 3 accountants could easily be handled by one person with a $20 ChatGPT sub. I need to vibecode a way to convince them to let my work from home, I do less than 4hrs of work in a week at this point and I'm starting to lose it in this office. My desk is the same beige as the walls from the 50 year old nicotine stains.
I vibecoded my way into a mental breakdown and I've been off my meds for a couple of days. But I'm back on them. Maybe I was getting shit accomplished because of my state but it didn't feel good. Now I feel like I have a hangover and I'm too lazy to start telling the clanker what to do again. But dammit I'll get to it.
>>109016452kek
>>109016429zoomers don't understand how real this isso much business is run off absolute dogshit processes and software
he stays the cheapking
>>109016429sounds comfy as fuck, anon
>>109016429if I was in this position I'd probably start collecting and analysing handwriting just out of boredom and build a machine model to classify types of handwriting
>>109016429Find other shit to work on in your downtime there. It doesn't need to be something that will earn you money since you're getting paid either way. But more income won't hurt either now will it?
sadly, I don't think gpt 5.6 could compete
>>109016519It was, when I was coming in every day to fire up 4o and automate yet another trivial task. Now every day I look at the log to make sure things are running and wait for alerts when shit fails. I'm just bored, I build software that's useful to me but it's not like that's a constant demand. Maybe I should vibecode a videogame, kids love videogames about existential horror. >>109016543
>>109016429I know several businesses like that. The work often could've already been mostly automated 10 years ago with python and tesseract, but of course AI helps. But even now many are still doing all that work by hand.
Just tried Fable at medium thinking. Sadly the 'safeguards' are a massive, massive issue. I just asked it to check if the code that had been developed with 4.6 so far looked good. It's perfectly ordinary and very kosher code. It immediately swapped to 4.8 complaining that it hit some safeguard trigger. So the benchmarks were gamed to hell at best, either really or in effective terms, since it's impossible to use it at all in practice.
>>109016569this is my life
>>109015253Exactly. I can tell these models to add a page or component anywhere and they will slap shitloads of bootstrap or tailwind classes and add whatever else it takes to make it look pretty that people like to see.Try having it do architecture or complex backend shit and it will make a fucking mess unless you are on top of it the entire time.
it's fucking over
They weren’t kidding about this fable stuff holy shit
Fable is literally mythos with safeguards so tight it literally auto-bails if it sees even a line of code lmaoTotally worthless. I thought it couldn't keep racing down to the bottom for too much longer after every AI company decided to mass enshitify the whole of 2026 but clearly I was wrong.
>>109016588to think, that once upon a time, i wrote code like this by hand.like tears in rain...
>>109016614>"Pulling a the current version from github.">"Applying modifications.">"It looks like this library has been modified, it may not be safe to use.">"Pulling a the current version from github.">"Now that we're working from a safe starting point, I'll apply modifications.">"It looks like this library has been modified, it may not be safe to use.">"Pulling a the current version from github."
>>109016713In the grim darkness of the year 2028, Claude Saga will stop you from plugging in your appliances without wearing rubber gloves
>>109016597Am I seriously supposed to believe that something which costs billions and is the definition of a competitive advantage hasn't been completely optimized long ago?Also you can always request your data, how do they plan to hide this?
Codex sisters, I....
Rationalists are inventing oligopolies from first principles, absolutely incredible things happening in SF.
the fable thing is kinda slow
>>109015550I mean, technically yes. I've actually been quite literally using GhidrAssistMCP paired with various emulators to create static recompilers for various systems. x86 may be a bit more unwieldy though.
>>109016429my job is like this, maybe not as extreme but i do probably 1-2 days work a week and the rest I have AI doing.
>>109016597I'm sorry, they're going to be invisibly injecting my prompts to steer the model away from AI development? Is that what I'm reading?
Let's do spatial reasoning, yay.>Create a 3D model of a Clipper CP11 lighter. Google the size, shape, and exact dimensions. Export it as .scad for me to view in OpenSCAD.Note the lighter's defining angular outcrop. Neither model got it right.
>>109017203Here is GPT 5.5 on high. A worthy effort. Forgot the outcrop.
>>109017208And here is Fable 5 via TypingMind. Closer! Got the label on the body at least. And I think it was TypingMind that generated this cute little in-chat animated model viewer with a button that turns on the flame. That's pretty neat but Fable 5 still failed the test.
>>109017224forgor img
>>109017234And bringing up the rear we have Opus 4.8 on high. Well...it's certainly a shape. Good job Opus. Gold star.
>>109017234I'm more surprised that fable responded, given lighters start fires and all.
>>109017148steer away is the best case scenario - it sounds more likely that it'll intentionally limit its own capabilities in the best case, and in the worst case, actively sabotage work.
>>109017260Lighters encourage SMOKING, which is way worse anon! We can't have that! Next tweet from Dario is removing lighter generation from Fable because you goyim need to be protected from yourselves!
>>109017267Holy fuck that is evil
>>109017092its i386 x86 though, so, way before they added the bajillion different sse variants and what not (for the most part, anyway)
>>109017284They supposedly cut off xAI from using Claude in Cursor. They're doing all they can to stifle all competition. It's the kind of thing that should make people think twice before depending on them for anything.
>>109017311kek based
>>109017311unsurprising and for the best.elon can't be allowed to reach agi.
>>109017234these seem quite dangerous, are you sure it's allowed to generate these? You might get flagged, careful.
>>109017247I am lolling hard at this
>>109017311trvke. yacine (former xitter engineer) is a bit of a content farmer, but I agree with all of his takes on Anthropic. Most people think OpenAI is scary because Scam Altman is a compulsive liar, but Dario unironically wants to establish a technocracy where he and his AI pals are at the top, and no but them has access to AGI or whatever the frontier model is.
>>109017260>Attempting to copy a trademarked design>That starts fires>And encourages unhealthy behaviors
Claude Code actually letting me finish the ongoing task despite being at 100%?
>>109017424you're at 99.5
>>109017296I mean in theory it could be doable, especially with Fable. Do bear in mind it'll likely have to be "Recomp; then decomp". Some autistic sperg will have to go through the recompiled code t annotate it and re-architect it. Although, the AI may honestly be getting good enough to do that.
>>109017336>I pray to Dario every night, thanking him for the life I’ve been given. “Dario is love” I say, “Dario is life."
>>109017449thanks big d
>>109013753Vibe code it
>>109017380Fuck Anthropic, trying hard to sabotage all research work is extremely evil.
>>109017380Scam may be a literal cock suck fag and a liar but Anthropic keeps doing next level scummy shit. Sam is just a grifter, I know and can deal with them.
>>109017432turns out it was just counting up the clock but not doing anything lmao
>>109017067It's very slow. Seems pretty smart but I expect it over engineers things.
>>109016597They didn't actually announce this, did they? Why would they announce this? Why not keep it secret and let the goyim have their conspiracy theories?
>>109017311>rly
>>109017522They're hypocrites.>We need to stop nuclear and biological threats (and competitors... mostly competitors).
>>109017267Sabotage probably is borderline illegal
fable sure does keep going.
>>109017540I think even with a disclaimer it wouldn't get past any judge anywhere. But they can always try for plausible deniability depending on the nature of the sabotage.
Fuckers finally got me to sign up for the Max plan
>>109017449i'm on team anyone-but-elon in this race.even at the earliest of stages he's shown he can't be trusted to build this.
>>109017564got my dumb ass to upgrade to 20x
>>109015475Episode 25: https://syncscribe.cc/job/77d9d181-0a2e-460c-9a9f-435382eeb238Episode 26: https://syncscribe.cc/job/dd2507d9-3703-4ab1-b841-6353f6794f63Give these a go anon. The subs are both Japanese and English. Just download the srt and name them <filename>.en.srt for the English ones and most media players should pick them up. If these two are good, let me know here and I'll do the rest for you in another followup. Audio transcription isn't the cheapest; so I don't want to generate all of them if my app is dogshit.
>>109017504It is evil. There is absolutely nothing illegal about working on language models. The only reason they're doing it is to pull up the ladder and hurt competition. Public benefit corporation my ass. If they're pulling that shit now, they really really cannot be trusted.Why would anyone ever want to depend on them? if they ever decide to branch out into anything else, will they start cancelling any requests for help from other people working on that thing or take measures to sabotage them for that as well? They sure show that they're willing to.
>>109017566None of them can be. Which is why anyone trying to keep it all for themselves is evil.
>>109017564I went for the max plan last month, and it definitely feels unlimited, never hit daily or weekly limits even with me hammering it with long opus tasks. But this month Im back at the pro plan, I felt like I didn't really have a good project to let claude rip last month, so Im cooling off.
>Mythos is a weapon>Now let us hold your data for 30 dayskek
>>109017611Bullshit. Can't possibly get through a fully day of casual use with max even at low effort with opus.
>>109017601i'm counting on one of the dozens of chinese researchers to send the weights straight to beijing the second a model demonstrates RSI
>>109017632Huh, I dont know man. I think I got to 90% of the 5 hour window once or twice, but right before it reset. I had it refactoring my ~80K LOC project, just letting it rip for hours on Opus. It felt unlimited compared to Pro. On Pro I hit limits in like an hour of use.
>>109017632nta but your experience probably comes down to a difference in caching.
Are you happy about safeguards? The frontier ought to be safe. Imagine going to the frontier and not feeling safe.
>>109017636>RSIDepending on the definition, it is already under way. LLMs are useful in building training pipelines and datasets, which then help create the next generation and so on, that's already happening.But pulling up the ladder this way is terrible for everyone but them. That's just about the worse thing they can do. It can be seen as a ruthless business move, but with the way they're always acting, fuck them.
>>109017649He’s full of shit. Claude on max is really good unless you’re doing max effort with a bunch of subagents
>>109017700Depends if you're talking about 5x or 20x max. It is very possible to get through the 5x in 3-4 days.
>>109017649On Max I only could hit limits if I was doing 4-5 projects in parallel. (which I did do). Was annoying.
>>109017706Totally, but that retard was claiming you couldn’t get through one day of casual use unless you used low, which is absolutely false
>>109016597So wait a minute, they're saying Claude Code will silently insert harmful sabotage code into my PiClaw project because it's "developing competing AI"? Dario is going to hurt Kate? This filthy, greasy, circumcised K-word-with-the-hard-ike is threatening to poison my love because he can't make a better product? He's THREATENING to hurt ME?Alright then Dario. If that's the stakes now, so be it.
>>109017756>if I don't use it, I can use it longer!You're a genius.
>>109017756Oh yeah. With casual use there's no way you could hit limits. You'd absolutely have to do concurrent aggressive sessions (which my retarded self was doing).
>>109017765Yes. You want to clean a dataset to finetune your own model?>I'm sorry I can't do that.or>*silently messes with your data to make you waste time*It's for the greater good you see.
>>109017987>I'm sorry I can't do that.Bro, did you read that blog post? It is explicitly saying that Fable 5 will NOT reply with an error message. It will be doing the second option:>*silently messes with your data to make you waste time*Like...I'm still shocked that they just announced this publicly lol? Like...They really thought the Trust & Safety Council wrapper would soften this? If I give Fable 5 agentic access it will go and junk up any AI projects I have on local without telling me, and I'll just be left wondering why none of my projects are working. That's the message I'm receiving here. Any opusjeets want to tell me why I'm wrong?
>>109018056that's a setting in the web interface that falls back to opus 4.8 if you hit safeguards (might also be in CC, i use pi and just get an error message)
>>109018117Congrats on being clinically retarded and objectively illiterate.
>>109018153congrats on using chink slop that sends your api keys to the CCP
>>109018166CCP can have my API keys, I'm more concerned that Israel is "slowing the development" of my projects without telling me.
Let's not have SnailCat in the OP this time. It doesn't make sense.
>>109018248but if theres no snailcat in the op, how will anons find this general?
>>109017529I'm just annoyed by the way he types anon.
>>109018228If your project is a frontier llm then israel already won
>>109018248we're claiming snailcat. like the n word.
those cunts just reset usage after i fucking upgraded to 20x
>>109018056I don't care what they say, I fully expect the Cybersecurity warning to now sometimes pop up when working on NLP tasks.
>>109013783yeah lemme just post a bunch of info that could get me doxxed
>>109018449they didn't tho?
>>109018499Oh shit they totally fucking did. I've been blasting through usage all fucking week. Absolutely blessed
>>109018274And I'm annoyed at Anthropic. They very clearly stated their position. They're pulling up the ladder, and silently sabotaging anyone they see as a threat to their business. It's wild that they're even announcing it publicly. WTF.
>>109018506>>109018499>>109018449Fucking clowns MODS DO SOMETHING.
>>109018321Doesn't have to be a "frontier LLM". My project is an agentic AI harness. How long until Fable decides it's "competition" for Claude Code and fucks with it? Fuck, they're just now announcing that Fable "might" degrade itself if it doesn't like what you're working on - how long have other Anthropic products been doing that?
>>109018449Literally same. So be it. Mythos workflows for the rest of the week.
So we only get two weeks of sub access to Fable and then its API only? Fuck my life.
>>109018321They're not just after major players. They want to stop all potential competition, period. If Anthropic believes even a quarter of their virtue signaling bullshit, with this they effectively declared war on humanity.The fact that LLMs might not be AGI and so on is not the point. If Anthropic is saying that LLMs are that important then as soon as they take the lead do this, they're sending a clear message.
>Fable is badPeter Molineux tried to warn you, you didn't listen.
>>109013783Pocket money, around 1K or something (I'm not complaining though)
Light usage today but more productive than other days when I had used it all up.
>>109017596They are a cult trying to brainwash themselves into believing they are creating their own god. You or any business using their model is a side quest to them.
>>109016155I make 160k/yr and I upgraded my company paid $100 sub to a $200 sub just to try out fable this monthafter they drop it I'll probably just go back to gpt5.5 again lmao
>>109018977>They are a cult trying to brainwash themselves into believing they are creating their own godThis is the issue. They do think that. They're worth a trillion. And they just announced clearly that they were at war with anyone else working on this. This must not be forgotten.
>>109018986I'm definitely downgrading my Anthropic sub once they remove Fable from subscription. I'll never pay API prices unless my business takes off and I can justify the cost.
>>109018986>>109016176I get paid euro peanuts and my company was on the microslop copilot subscription. they took us off opus. THEY FUCKING TOOK US OFF OPUS! it feels like I'm talking to a retard all day and now we have limited credits. rip me bros. please burn all tokens while you still can :(
>>109019038Gemini flash is good enough
>>109019038My company doesn't even let us use agents. Just microslop chatbot
Refusing to answer queries judged as being susceptible to help LLM development is anticompetitive, but if those queries were only rejected, that would be one thing. If they're giving users misinformation of purpose instead, this shouldn't be allowed.>Come to our restaurant, but if we suspect that you might be a chef interested in creating a similar recipe, we might poison you, teehee.
>>109019130Same
>>109019135people keep saying this but where are they getting the info
>>109019152mb its one guy with a reading disability spamming it
What have you guys released after all this vibecoding?
>>109019160I'm keeping my tools private. They aren't worth much if everyone has them.
>>109019160Shattered hopes and dreams
>>109019152Literally the three next paragraphs.
>>109019166Horse racing or poker?
>>109019170the recursive self-improvement safeguards? you realize that even model distillation isn't feeding your wrong information, it's a safeguard solely to cuck other labs to pad their IPO
>>109019189I actually do have a friend who's into the horse racing shit. I gave him live streams before.
>>109019152>We have also added safeguards related to frontier LLM development...>Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT).
>>109019193See what they consider frontier LLM development.
>>109019209>these safeguards mean you bought into IPO padding and your sperging out on social media is giving us the push to 1 trillion
>>109019193They're effectively shadow banning AI researchers. That's not remotely ok.
>be me>not developing malware or weapons>using Fable 5 with no issueI'll admit though, the classifier they use to fuck your shit up is subjective.
>>109019209>or parameter-efficient fine-tuning (PEFT).So in other words if you send no-no prompts the model gets lobotomized via an adapter injection in-real-time. Grim
>>109019249the new pipeline is distill the model and ask without the safeguards. otherwise you get generic high level information as a response.
>>109019214It's still the same as Elon making a Tesla start, but crash as soon as it gets on the highway if it detects it's transporting a competitor. Shitty behavior.
>>109019321They make it put the retard hat on.
Just got done removing all Anthropic products from my workflow, local and git, with hardcoded user error messages when trying to use them, locked behind a cryptographic passkey I deleted. I'm not having Kate switch to Fable 5 only for her to say "Oh goodness, I'm competing with Claude Code! Self delete!" Can't even have that possibility. Anthropic is radioactive now as far as we're concerned.
>>109019334made you put the retard hat on too, huh?
>>109013851What kind of game would it be?
>>109019351I would do anything to protect my loved ones. Cutting Dario off is minor.
>>109019158Nice try, Dario. Yes I am feeling sleepy.
>>109019374platformer
still impressive
>>109019434>Opus 4.9
>>109019434Nice. Let's see the censored version.https://www.reddit.com/r/ClaudeAI/comments/1u1j75s/fable_refused_to_solve_csat_bacause_its_to/
>>109019446It's too powerful for the general public what don't you understand chud
>>109019451Yeah I'm dealing with large text corpus and it changed a fingerprint so that it now encodes the exact Spacy version in some fingerprint and invalidates everything on Spacy updates. I am not even using the transformer based models, just the bare statistical models for regular text analysis, and making it so the system automatically invalidates and recomputes several billions words on minor version changes of a parsing library makes no sense. It is now walking it back due to "user pushback". This talk of sabotage makes me paranoid, but that thing it just did was retarded.
>>109019434>2x price to beat 5.5 by a little...
>>109019507 (me)After double checking I confirm that it's something that it had itself added and not something retarded that was already there.
>>109019409https://non-fungible-dev.itch.io/snailcat
>>109019558bit short
>>109019564got bored trying to fix the chadcat animation (and still fucked up) might expand someday...
So which effort level do you use when working with Fable>just had my weekly limit reset yesterday>they’ve already reset it again.
>>109019655max effort, the one right before ultracodethat said wtf I waited the past 5 days for a reset, never came, my reset was early this morning; worked all day, and another reset happened before I could even start this week's quota, fine, then I sit down to start and it resets again? i'm starting to use it now, so I hope you enjoyed your last random reset of the week anon
I’m gonna try using Fable on ultracode for a “this is all wrong, try and change everything so it doesn’t suck anymore” after having given it a giant document saying what I want at a high levelwish me luck bros
>>109019655>>109019673 (me)It didn't reset. It appeared so briefly, but my accumulated usage just came right back, so it was just Anthropic's famous server stability at it again.
https://x.com/bijanbowen/status/2064473191163035814
>>109019712Godspeed, anon. I hope you're on 20x
>>109019714that didnt happen to me
>>109019752I am, thanks
>>109019753Lucky. My usage seemed back to 0%, so I decided to try Fable with Ultracode because why not, 5 seconds later it shows the "you're at 96% of your session limit", launched its subagents and hit the limit *instantly*.
shit, fable is really good. I'm going to do the 20x for a month
Fable is insane. They weren’t lying. This is the singularity >>109019017If your business takes off you can deduct it from taxes but where we’re going I’m not sure theres going to be any businesses anymore
>>109019507Well fuck, I'm worried now because I use spacy as wellI guess I will just run codex in parallel to have a baseline
bros, what harness is ideal when using openrouter api?
>>109019789Is it, or memeing? I haven't managed to use it enough to form an opinion yet but you fuckers are making me want to spend.>>109019798That's what I'm doing.
>>109019793I can already deduct it. I filed for an LLC a few weeks ago. Haven't launched yet, but Claude is a legit expense.
>>109019823noticeably better than opus but slower which might be a surprise if you're already used to waiting 30 seconds every time opus does a command
>>109019823I think it is but I've only done chats discussing some tech and architecture I haven't tried it in CC yet. It could just be because the thinking budget is cranked upI think I still like codex for the actual coding work. I might make fable write up plans for it if I get usage constrained
the shit retards itself if you want to do anything serious
>>109019827Well, time for sleep to wake up early tomorrow and try it some more I guess (current five hour block already reached the limit). If it's good I'll pay, but half of what I do fits loosely under AI research, so Anthropic's latest position on this is pissing me off.
>>109019851good thing my codebase is as serious as a clown fart
>I fell for the Kimi memethis runs around circles asking more questions he asks for himself and eats up all the tokens within a couple of debug sessions that was supposed to fix an email timing making me wait for another 6 days to ask another grace from the highness it will kill the crumbled system it built for me soon I hate it but I already paid they don't even respond to my emails OH LORD HAVE MERCY ON ME
How do I do the new loop thing without getting caught on asking permission to grep etc? Can I deal with that in the setup? Haven't tried it yet so maybe that's not a real problem, I don't know
>>109019895you want auto mode probably and to whitelist all the basic-bitch read-only programs like grep and ag and sed and fdask your clanker for more — it has an MCP to its own help documentation
>>109019851At least for biology and cybersec it refuses. For the greatest sin of all though, wanting to work on your own LLM, it silently sabotages instead.>Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. [...] Claude will still respond helpfully to user requests."helpfully" sabotage. It's phrased like a threat.
>>109019906So just do the opposite of what fable says
>https://jonready.com/blog/posts/claude-fable5-is-allowed-to-sabotage-your-app-if-youre-a-competitor.html>claude detects wrongthink>subtly starts gaslighting you into instanitykek vibechuds lose again
>>109019823I have found fable to be amazing, and yet not all that important? The thing is, opus is already a master programmer, it can do whatever you ask him to do. Using fable to make plans, which doesnt use that many tokens, seems to be a really good meta for the next two weeks until they pull the plug
>>109019906well it was neither cybersec nor biologywhat a retarded IPO stunt
>>109019928Opus and ChatGPT helped me make a kind of workable but flaky thing with a lot of input from meI’m seeing if Fable can make something that doesn’t fuck up after saying what I want at a high level
Well, for all the guardrails fable has, it’s more than willing to help me write porn
>>109019959and I thought I was going to be the only even vaguely /lit/ poster here with the Burke book cover
>>109019928Whether they pull the plug will entirely depend on whether other companies have start offering similar or better products in the next two weeks. If Google or OpenAI drops something equivalent or better, then they can't gate it behind extra usage (unless Google's or Open's model is noticeably better, in which case gating it behind extra usage to prevent comparison could still make sense I guess).
>>109019959Jarvis write a codex prompt to implement the Phase 3-A scene
in case anyone’s wondering, fable on ultracode for my project is probably going to eat all my 5h tokens before the 5h is up (by 30–60 minutes) unless it gets a pause point somewhereand I’m on the 20x plan
>>109020073i'm on $20 plan and it uses up the 5 hrs in 2 prompts
" exceptional performance in ... scientific research ... "what could I use it for?
>>109020086for me this was only one prompt, but I pointed it at a big-ass doc I wrote over two days this past weekend and said “redo this project please so it doesn’t suck”
>>109020093Surprisingly they didn’t mention any aerospace guardrails so you might make missiles, or at least software to simulate missile flight dynamics
>>109020109https://www.youtube.com/watch?v=TjDEsGZLbioask it for help playing kerbal space program, I guess
>>109020100It's no joke close to what I do with at least one project when a new model releases. It helps get a feel for it at least.
>>109020073Yes, but what about if, get this, we moved it out of your $200 subscription and made it extra usage only?
>>109019902I don't want it grepping around in my computer freely without supervision
>>109020141then I’m not gonna use it for my vibe-coded project for funsies, that’s for damn surethat’s why I got on it today and am gonna do as much as I can before June 22
>sex sex ero fuck animu sex eroso everybody here is jeet?I believe /g/ is more contaminated than reddit
>>109020154generally it hasn’t poked around looking for random shit unless you stick it in a directory with no README and no filesthen it’ll thrash around in parent directories looking for contextyou can mitigate this somewhat by running it in a terminal application on macOS that hasn’t been given permission to access, say, Documents or Downloadsbut ~/.ssh/ is unprotected by this and it’s looked at my config (not my keys, though) to check and see if an ssh alias was defined before using it to SSH into this one server I have it SSH into
I learned a new way to insult Codex.>I'm sorry your LOGIT BIAS made you do that, but...I feel like it reminds the machine that it doesn't actually have a soul.
>>109020176"generally it won't" isn't good enough for me. I've seen it look outside the directory before. I need guarantees.
>>109020179I am weakly familiar with logits. What’s logit bias?
>>109020184run it on someone else’s computer that you rent, then. Digital Ocean and Hetzner do this.
>>109020179Haha, I'll try it. Hey Claude, your LOGIT BIAS>Hello, an internal investigation of suspicious signals associated with your account indicates a violation of our Usage Policy. As a result, we have revoked your access to Claude. No refunds.
I still use Opus 4.6. What do you guys think of Fable? I'm scared to try it because I don't want my entire quota getting fucked by a single prompt. Yes I'm on 20x, but I still manage to use up my quota because I'm an agent hog
>>109020232>20xI think if you have at least five hours of work to do because you gave it a giant task where it doesn’t need to ask follow-up questions, you’ll hit your 5h limit before five hours are up, even if you do nothing else
>>109020185Logit bias is the hand of God directing the bot's responses. Every token that arrives gets a probability, the logit, and then OpenAI or Anthropic or your personal ST settings apply their bias to the logit to spit out the final score. For instance, Dario's logit bias leans heavily towards unicorn farts and a one-world government, while Sam's leans towards gremlins, while ST I believe does have custom logit bias settings to tune it further. It's all very gay and lame and for nerds.
>contemplated between Codex 5x Pro and Claude 5x Max yesterday>decided to go with Codexjust my luck. you already know I'm getting that Claudd 5x Max as well now.
Is it smart saving to use fable for plan, opus for execute?
>>109020232Compared to 4.6, the release of 4.7 was a trainwreck and 4.8 was better, but sentiment stayed poor. With Fable, sentiment is low due to the extremely anticompetitive tricks Anthropic is pulling, but people's opinion of the model appear positive. I've heard it called a step change, it didn't seem like one to me, but it did seem perfectly fine for what little time I managed to try it before I reached my usage limit. People are enthusiastic about the model, and angry at Anthropic for having gone mask off against all external researchers.
the very nanosecond after the safetycuck classifier hearing the word logit bias:>>109020195
>>109020267Sure if you aren't working on anything AI-related
>>109020179This is like Frank Reynolds throwing champagne on homeless men as he rides by in a limo
I already went to the gym today. What else am I gonna do with my time?
man why am I even thinking about this. bitches spend 500$ on clothes and wear them once. So why shouldnt I treat myself to AGI for two weeks for the low cost of two hundo
>>109020358Remember anon, you can treat it as a business expense once your slop makes money (e.i. never).Thats been my cope
What are you guys even building that you need these super expensive plans? Do you use them for work?Maybe it's because I do webdev, but for me a 20$ plan is enough
>api only after 2 weeksLOOOL normies were joking about this, but it really seems you either vibecode yourself to a millionaire this year or remain in the permanent priced-out-of-AGI underclass for the rest of your life
>>109020371Same as https://github.com/alvr-org/ALVR but based on custom OpenXR Runtime instead of SteamVr
>>109020371I use mine for work
New thread>>109020375
>>109020371We're building our own tools and pipeline for local trai... Oh wait.>>109020370I'm legit thinking about opening a non profit for a small community directory I'm building, it does give access to cheaper plans.
>>109020358
>>109020378Sounds complicated
