Icea

@Ledensy @hubxyz

@hubxyz You guys still existing?

AI models are eating faster than the internet can grow Epoch AI published one of the most important papers in the AI industry. They estimated the total stock of quality human-generated text at ~300 trillion tokens, and projected an 80% probability that frontier models exhaust it between 2026 and 2032. We are inside that window now. But the real crisis is worse than the paper predicted. The supply isn't just finite. It's shrinking. Over 74% of newly created web pages now contain AI-generated text. The internet is being contaminated by the very models it trained. Every generation of AI pollutes the training data for the next one. The industry tried synthetic data as a fix. The research killed that idea. Models trained on their own outputs degrade. They lose the nuance and variability that made human data valuable. Without real human data as an anchor, quality collapses. Meanwhile, training costs keep climbing. GPT-4 cost ~$40M. The current generation is approaching $1B. The next: $10B+. Every dollar spent scaling compute accelerates the depletion of the data these models depend on. Ilya Sutskever said it at NeurIPS: "We've achieved peak data. There's only one internet." And that's just text. The industry is shifting multimodal. Models now need real-world audio across hundreds of languages. Real images from real environments. Real video from natural conditions. That data doesn't live on the internet. It hasn't been collected yet. The race for compute is solved by capital. The race for data has no equivalent solution. You can't manufacture more internet. You can't synthesize what doesn't exist. And you can't scrape what was never captured. This is what @hubxyz is building. The data AI needs next doesn't exist online. It exists in the real world. And someone has to go collect it

@Cryptozylver @hubxyz bro check discord

@hubxyz Wen 😭

@hubxyz HubApp is launching soon...

Nano Banana could never

@beechinour swipe

you can 10x the quality of the videos you create by being ruthless about where you take inspo from… there’s a huge difference between creators that have an amazing reference bank vs just trying to rip “what’s currently working” on the X timeline i built a huge a swipe file which includes: - the best commercials from the last 10 years - critically acclaimed movie trailers - random clips I couldn't look away from rt + comment "swipe" and I'll send over the file (must be following)

@amnxnet Who animated the video?

This is one of the best launch videos I’ve seen today…

@hubxyz This is the most exciting chapter since the Hub AI Agent

AI companies spend billions acquiring training data 5 billion people carry cameras and microphones in their pockets Nobody connected the two Until now We built a new platform where anyone with a phone can collect real-world data for AI and earn from it. A student in Manila recording how Tagalog sounds on a crowded bus. A mother in Lagos photographing the objects on her kitchen table. A woman in Bogotá capturing how afternoon light falls through her window. Real images. Real audio. Real environments. From 190+ countries. Data that no AI company can buy from any provider today because it does not exist yet. Hub already delivers structured training data to companies building the next generation of AI models. Now we are opening the supply side to the world. The people left behind by the economy just became its most valuable participants. The first tasks are live. The first contributors are already earning inside our Discord. This is not a test. This is the beginning. The full public launch is coming.

@xarmin 💪

UAE will recover at a speed that will shock the whole world. Do not underestimate the creativity, financial/diplomatic power, and sheer will of MBZ, Al Maktoum, and their government. They have faith, so I do. #UNITED 🇦🇪

@beechinour director

stop wasting hours manually prompting images for your ai videos i created an openclaw bot trained on images from your favorite directors, that lets you prompt at 10x the speed how it works: > fully trained on "director profiles" (fincher, anderson, etc) > takes dna from director profiles, character refs, and ref images > generates batches of images via freepik api > organizes outputs directly into your desktop folders this is game changing for fast iteration/batch prompting for ai videos RT this + reply "director" and I'll send you the skill file (must be following)

@beechinour skill

my OpenClaw bot works as my assistant editor now - digs through my script/comes up with b-roll ideas - researches x, youtube, google for tasteful, undiscovered b-roll - downloads the clips and puts them in organized folders - gives me a list of timestamps with high quality moments i used to pay an assistant editor to do this exact task... RT this + reply "skill" and I'll send you a free skill file that researches, scrapes, and organizes broll for your videos (must be following)

@Privacygateway_ Pls gib invitation code

This is how your prepaid card will appear in PrivacyGateway. Simple. Private. Ready to use.

@hubxyz damn...

The models converged. The war didn't GPT-5.4, Gemini 3.1, Claude 4.6, Grok 4.2 Within 3% of each other on reasoning Within 4% on coding Within 5% on multimodal tasks The era of model differentiation is ending The era of data differentiation is beginning Every lab trained on the same public internet The same crawled pages The same licensed archives The same 5 trillion tokens Models converged because the training data converged The labs that separate themselves next will train on real-world data Not publicly available. Not pre-packaged Extracted from the live web, structured at the source, delivered to the pipeline before it becomes everyone else's training set That is what Hub was built for. The model war is over. The data war just started.

@sifarchand true

@hubxyz Team on it

Dario Amodei listed 7 factors that determine how capable AI becomes Compute. Data quantity. Data quality. Training duration. Objective function. Two normalization terms. The industry is pouring $100B+ into factor one. Factors two and three have no infrastructure at scale. No distributed pipeline. No real-time supply chain. Hub knows the true bottleneck. Our next step is fixing it.

Another sensational SREF from @lloydcreates --sref 8025898031

Higgsfield cooked with Soul 2 🟡 Prompt: "A cinematic, fine-art close-up portrait of a beautiful woman with dark hair. The framing is tight (head-and-shoulders), eye-level, with an 85mm-equivalent lens and shallow depth of field (f/1.8), producing strong background bokeh. Her black hair is slightly wind-tousled, with loose strands partially covering one eye. She gazes subtly upward with a calm, introspective expression, natural film grain, camera film Kodak Vision3 500T, shot on an Arri Alexa 65 for cinematic look"

@Dork_sense @morphic Does it work only on NB?

keeping lighting consistent across your AI film is actually easy Create with color palette or fix existing image with this prompt 🧵👇🏻

Hiring AI filmmakers ASAP. Paid. AI video prompting & AI creators (AI artists to make 20-30 seconds visual stories) Post your portfolio or example (or DM me).