the stock market closes the news doesn’t earnings, macro headlines and sudden market moves can happen while traditional markets are offline @Bybit_Official TradFi Perpetuals bring the perpetual model crypto traders already know to contracts tracking US stocks, gold, silver, oil and other traditional markets 24/7 trading, USDT settlement, long or short positions, all from the same Bybit account there’s also a current promo with 0% maker fees, discounted taker fees and up to 100 USDT in welcome rewards for eligible new users If you already use Bybit, TradFi Perpetuals are available through your existing account Starting fresh? bybit.com/en/sign-up?affilia…

this is how i’m spending my weekend 👀 the @cedearena $100K tournament is live and ongoing. previously whitelist, now open to everyone. a simple walkthrough: » sign up » head to the battle page » join a 1v1 and trade your preferred market » win → rewards hit your balance the tournament runs for 7 days. and it’s not just about winning individual battles. your winnings count toward the tournament’s grand prize. might as well run it 🤝 cedearena.io/100k?utm_source…

Gm fam One detail about @vangrid_io that is easy to miss is what happens after the phone recording The requester is not simply paying somebody to send over a random video file An accepted capture can go through reconstruction and become a textured 3D mesh in GLB format That changes the usefulness of the network Video is great for humans because our brains automatically understand geometry We can look at footage and infer walls, distance, obstacles, openings and surfaces Machines need that information represented much more explicitly A reconstructed spatial model gives software something closer to the structure of the environment itself - Now imagine that workflow at scale - A robotics company needs a location - Someone nearby captures it - The capture receives a provenance record - - The requester accepts the delivery The scene is reconstructed The resulting asset can then enter simulation, mapping, planning or another Physical AI workflow That is a more interesting product than “get paid to take videos.” The video is just the acquisition format The real product is structured spatial information about a place that did not previously exist inside the buyer's system And because the capture can be commissioned again later, the model does not necessarily have to represent the location forever Physical environments have versions A good spatial-data network should be able to create the next version when reality changes

the stock market closes the news doesn’t earnings, macro headlines and sudden market moves can happen while traditional markets are offline @Bybit_Official TradFi Perpetuals bring the perpetual model crypto traders already know to contracts tracking US stocks, gold, silver, oil and other traditional markets 24/7 trading, USDT settlement, long or short positions, all from the same Bybit account there’s also a current promo with 0% maker fees, discounted taker fees and up to 100 USDT in welcome rewards for eligible new users If you already use Bybit, TradFi Perpetuals are available through your existing account Starting fresh? bybit.com/en/sign-up?affilia…

Do you know bad we have it here for folks to gatekeep the routes they used to send out a mere $8 that gives access to an open edition NFT??? Smh🤦‍♂️

Neither chain could hold this alone Oct 08 2026 • Exclusively on @opensea

Good Morning 🌞 Happy Friday All of You Family 💗 Look > LLMs got the internet, Robots did not. A language model can train on text that already exists. A robot cannot scrape "how to pick up a mug without knocking the bottle" from Wikipedia. That motion has to be produced. Teleop in a real lab does not scale. Raw video is missing the physics. So Physical AI sits around waiting on data while the hardware is already here. That is the job @axisrobotics actually has. You open a browser. You drive a simulated arm through a task. That run becomes a trajectory. If it passes replay checks, you sign it on Base and it gets a Data ID who made it which task, when. Not a screenshot. A record. Dataset V1 moved π0.5 from 83.9 to 88.8 on LIBERO-Plus. Little Prince ran 15,371 people through 85,387 sessions and put a policy on a real robot in five days. This is not a pitch deck metaphor. The loop is already turning. The $AXIS Community Sale is oversubscribed and the commitment window is still open until 28 Sep, 05:00 UTC. Fixed $0.10, $100M FDV, $1M cap, 10M tokens. Over the cap, settlement is pro rata and unaccepted USDC comes back. Official door only: sale.axisrobotics.ai KYC sits with Sonar. Axis does not DM first. Not available in restricted places including the US, UK and Mainland China. < This is not investment advice.> If you are not buying and still want the thing to move, contribute. Do a task in the simulator, then sign the trajectory onchain. Unsigned runs do not count. Start here: s.kaito.ai/E3BZVv3

GN CT 🌙 Friday night is for closing tabs, not opening ten more charts. If the house is calling then go. The work shop paper will still be on the internet tomorrow. One quiet thing landed today and it is worth staying awake five more minutes. @axisrobotics got a paper into the Physical World Models workshop at IROS 2026. Pittsburgh. 1 October. Not a banner. A review panel said the work can sit next to other labs. The claim is narrow and that is why it is useful. Can noisy, community collected simulation support embodied pretraining. Their V1 set says yes if coverage is wide enough. Continual pretraining on AXIS Dataset V1 moved π0.5 from 83.9 to 88.8 on LIBERO-Plus. Gains kept rising as they scaled from a quarter of the set to the full set. No obvious plateau in that sweep. Scale they put on the graphic today: 1,800 plus tasks. 1.5 million plus trajectories. Co-authors listed from Georgia Tech, Berkeley, Texas A&M, Johns Hopkins, Penn, Michigan, NUS and NTU. That is a different story from a contributor leaderboard. A workshop slot means outsiders can read the method, pull the Hugging Face dump and rerun the training code. Paper, dataset, project page and GitHub went out with the announcement. Read it as infrastructure meeting a conference, not as a victory lap. LIBERO-Plus is a benchmark. IROS is a room in Pittsburgh. Neither one is a warehouse robot. What it does prove is that the open corpus is thick enough for a named model to move on a public test. If Friday energy is left, open the paper first and the timeline second. GN. Let the night be slow. The interesting work already shipped. gAxix nitter.net/axisrobotics/status/21…

sign out boy 😂

One thing that has always been a major concern about trading is how fragmented everything is. You check the market on one platform. Do your analysis somewhere else. Look for setups somewhere else. Then switch again to execute. And somehow, the trader is expected to put the whole picture together. Trading can be much simpler than that. Just more connected. Bring the market, analysis, setups, and decision-making workflow closer together instead of making traders jump between five different tools just to understand one trade. Less fragmentation. More context. Smarter decisions. Well... we built all of that for you already. Get ready for a new trading experience you haven't witnessed in Web3.