🗞 𝐀𝐦𝐢 𝐋𝐚𝐛𝐬 𝐱 𝐕𝐚𝐧𝐠𝐫𝐢𝐝 the world still needs to be captured. if you take one thing from the ami labs round, take the demand. not the valuation. yann lecun left meta and raised $1.03b for ami labs at a $3.5b pre-money valuation to build world models. the interesting part isn't the number. it's what the money says about where serious ai research is going. the argument is architectural: predicting the next word or next pixel isn't the same thing as understanding how the physical world behaves. a factory isn't a paragraph. a loading dock isn't an image. a stairwell isn't just pixels. physical ai needs models that can reason about environments, objects, movement and cause and effect. ami is trying to build the model. but models still need reality to learn from. real shelves. real warehouses. tight doors. stock rooms. alleys. construction sites. places a mapping van may never have captured. vangrid's pitch is basically: what if you could commission those places instead? an organisation or agent posts a bounty. the capture gets funded. someone uses a phone they already own to collect the environment. the result comes back as a textured 3d model with a capture record tied to things like the content hash, coarse location and timestamp. @vangrid_io also says faces and licence plates are blurred on-device before the video is encoded. that creates an interesting separation. the ai company doesn't necessarily need to own the capture fleet. it can request the environment it needs. the network supplies the missing physical data. but there is an important line here. the ami labs round is evidence that investors see demand for world-model technology. it is not evidence that a phone-captured mesh is already good enough to train a lecun-style world model. that's still an empirical question. geometry quality. texture quality. coverage. consistency. capture frequency. and whether the resulting data is actually useful downstream. that's where vangrid has to prove itself. what i find interesting is the part it is trying to remove: the waiting. if a robotics team needs a specific environment, the old answer can involve sending people, vehicles or specialised mapping equipment. vangrid's answer is: post the place you need. let the network capture it. turn the physical environment into a requestable data asset. so i wouldn't read the $1.03b ami round as 𝘷𝘢𝘯𝘨𝘳𝘪𝘥 𝘪𝘴 𝘷𝘢𝘭𝘪𝘥𝘢𝘵𝘦𝘥. that's too big a jump. i'd read it as a signal that the demand for better physical-world data is becoming harder to ignore. ami is building the intelligence layer. vangrid is trying to build part of the reality layer underneath it. the open question is whether the captures are good enough to become infrastructure rather than just interesting 3d scans. that's the part i'm watching fr. techcrunch.com/2026/03/09/ya…

pet that earns stock tokens? i'm buying one i’ve been keeping an eye on @Stockpocks and love the tamagotchi-style mechanics paired with earning stock tokens as you play it’s a refreshing concept and definitely worth getting in on early the interesting part is that... the 24-hour genesis sale (dutch auction) is officially live it runs from sunday, 4 october, 13:00 utc (today) to monday, 5 october, 13:00 utc mint yours here: genesis.stockpocks.com

Wizard of soho flexing his heaviest bags, aping into pure memecoins during a market dip while pointing out the mega whale holders backing the play. ​“My biggest bag is a pure meme, just a... not even rwa meme, just a pure meme.” ​“We had a big downturn, people thought it... it's moving back. no, this is... and look at the holders, the holders are just mega whales.

This is what September looked like for @axisrobotics A month where conviction met proof. We crossed 5M robot data trajectories on @base and we’re already past 6M, growing every day. 200K+ contributors worldwide are helping us build the physical experience layer for robots. Our Franka dataset family crossed 160K+ downloads on huggingface, becoming the most downloaded open-source simulated Franka manipulation dataset. We sat down with brian_armstrong during his visit to Singapore to talk about the future of Physical AI and where robotics is heading. Huge thanks to baseapac for making it happen. We also joined BasedIndia to share our approach: building a scalable human in the loop system where anyone can contribute to robot intelligence. Our research was accepted to the IROS PWMS workshop another step toward scaling embodied AI through diverse, real world inspired data. And then came one of the biggest community moments of the month. 2,472 participants committed 2,394,209.52 USDC in the Axis Community Sale on Sonar. The most participated Sonar sale of 2026. To everyone who participated, contributed, downloaded our datasets, tested our work, or believed in the vision thank you. We also announced 25% TGE unlock for everyone, plus additional fully unlocked bonus tokens based on original commitment. September wasn’t just about milestones. It was about proving that the community, data, research, and infrastructure can compound together. 6M+ trajectories and counting. The engine keeps running. Robotics, Based.

Craig asks King if liquidity has ever stopped him from cashing out a position 👀 King brings up his $Mumu bag from last cycle. He was the top holder with a position worth around $2.5M at one point… but liquidity made getting out a completely different story. One of his biggest round trips. Ft: @KingTakesOff x @shineyd1111 x @CraigOPodcast Powered by @BasedBot

gSLEEP CT 💚🧡 I would like to note that my attention was drawn to the @sleepagotchi concept because, in my opinion, the development of this project is the most fascinating. The original idea was great — a play on the sleep-to-earn concept and the Gotchis themselves, but now they have taken the development of this concept to a new level. The transition from a “game with token” to a “product with an entire supporting ecosystem” is genuinely impressive. A new stage in the development of the concept underlines the importance of the chosen direction for further development — and this is not just the optimization of the existing system, but a change in the very essence of the product. The renaming of the project to $SLEEP → $CHI is, in my opinion, a signal and a demonstration that they are not tied to the game, but to the product itself. Asleepagotchi project actively communicates with partners and users, joins other ecosystems (such as Welcome3), which in itself is an indicator of how large and promising the project is. The changes in leadership may seem confusing at first glance, but such projects as @sleepagotchi need fresh thinking and a new development strategy to take their product to the next level. However, what impresses me with this announcement is that despite the changes in leadership and project goals, the look and feel of friendly and approachable sleepagotchi, as well as fun and unique Gotchi, have remained the same. I am excited about the future of this promising project and will be looking forward to updates.

Alright @ChatGPT @claudeai @grok @bot Find me a .si domain that can actually make me money. Search everywhere, give me your best ideas, and please don’t make me regret asking

I'm a Hazels Creator — I'm on the Waitlist. @0xhazels #WEAREHAZELS studio.hazels.io/waitlist

Stock Pocks Genesis is officially open. The @Stockpocks sale is running as a 24-hour Dutch auction, starting at 250 USDG and gradually dropping to 20 USDG, with 10,000 Genesis spots available. What makes this more interesting than a normal mint is that the Genesis piece is basically your entry into the wider Stock Pocks ecosystem and the game they’re building around it. Definitely one I’m watching while the price moves down.

my tier list of AI agents so far

✌️ 𝐌𝐢𝐧𝐭 𝐍𝐚𝐯𝐢𝐠𝐚𝐭𝐢𝐨𝐧 the first thing i noticed on mint dot io after seeing people talk about the platform, i decided to look around for myself. the interface is pretty easy to move through. the homepage keeps things simple, with clear sections for casino and sports instead of throwing everything at you at once. there is also a big game library with 3,000+ games from different developers, plus a 21-level rewards system and customer support. registration is straightforward too. but there are a few things i noticed on the other side: → payment options are limited → there is currently no welcome bonus for new players so the experience isn’t just about how good the interface looks. for me, the interesting part is seeing how the navigation, game selection and rewards system fit together. simple on the front end. more to consider underneath. play for more earn xp. climb the ranks. get rewarded @playonmint ✌️

Want to go hard in the trenches? According to King, you gotta be there 24/7. The trenches don’t sleep 👀 Ft: @KingTakesOff x @shineyd1111 x @CraigOPodcast Powered by @BasedBot

PaperDAO: The DAO Bet on the PaperTrade Thesis @AzFlin @paperdao_xyz @daosdotworld

$BTC Trying to break above the cloud After a sharp decline, the asset bounced off the MA50 moving average and began to gradually move into longs. At the moment, the price is actively trying to consolidate above the cloud on the 4H timeframe. We're in the weekend and volatility is fairly weak, however, the dynamics are positive, which points to a potential retest of resistance at $86,700 at the start of the workweek. Today I wouldn't expect any serious changes, since volumes have dropped off, so we wait for Monday and there we'll watch buyer activity, but preliminarily I expect a move into longs.

Good Morning QuipNetwork Most apps and games still ask you to blindly trust their "random" numbers even when it decides who gets allocations or wins big, that’s wild if you think about it @quipnetwork new QVRF system is actually fixing this by using real quantum hardware, they’re not just throwing numbers at you; they commit the quantum samples, audit them independently, and run an extraction process so everything is fully provable Why this matters way beyond jpegs: 🔹 Fair, un-rigged allocations 🔹 Verifiable gaming & lottery outcomes 🔹 Simulations that actually need true entropy 🔹 100% transparent selection processes Quantum Echoes is kicking things off as the first project on it, but honestly? The infra itself is the real alpha here, there's a huge gap between generating randomness and letting anyone audit how it was made Anyone else looking closely at this? 👇

lets talk about @ptsdshow episode 1 , "Enter the Tenches" episode 1 opens with pt catching bottom on a 5s chart, ripping 8k% instantly then trying to hit sell and his hands turn into floppy spaghetti. that's the entire show in 30s. the win is fake, the exit never works and he wakes up at his desk right before liquidation cuz he opened a 100x leverage trade in his sleep. the therapist actually has a name for it, acute nocturnal somortosis. sleep trading. xD but the origin story is what got me. march 2021. his half cousin d**k stumbles into a fortune on axie infinity so pt decides if that guy can do it then anyone can. he sells his beanie babies and pokemon cards for 1800$$, works out he is two 100x away from million, drops out of college and moves back in with his mom. that is not satire, that is a documentary and i am in the credits. which detail in epi 1 hit too close because the spaghetti hands got me. episode 2 reaction stay tune.

a blessed sunday to you all taking a rest for the day but the countdown begins on nucleus 14 days left on nucleus szn 3 eyes on $aura only top 5,000 qualify if you talked about kaito for yaps in the past you know how this ends start writing about @nucleuscodes now or read the board later and wonder why you waited enjoy your sunday ✌️

i kept hitting the same wall while reading about Vangrid. • even as physical AI gets better, robots still struggle to really understand the physical world. • most robotics research is stuck in controlled spaces like labs, warehouses, or factories. • but robots are eventually going to need to handle messy outdoor spots like busy city streets or construction sites. • even the smartest robots and simulations get stuck if their data is out of date. • @vangrid_io fills that gap by giving physical AI the live data it needs to work in the unpredictable real world.

Wait, you can run your entire email marketing platform straight from Cursor now? @nitrosendx is changing how the whole workflow looks.

GN CT 🫡 Thoughts on similarites between Vangrid and AmericanFortress include how both make crypto less of a guesswork scenario. For @vangrid_io , everyone's heard of the current hotness of Physical AI; Jensen Huang himself declared CES 2026 the 'ChatGPT moment for physical AI', and NVIDIA's Physical AI Data Factory Blueprint at GTC highlighted the same issue of a dearth of real-world data for training. Fei-Fei Li has mentioned the same challenge holds true for robotics, albeit with less training data than for language models; you can't scrape a street corner off of the internet. Vangrid's solution turns smartphones into a capture network that gathers on-demand ground-level 3D data of real-world locations with captures anchored onchain on the Base blockchain. The explorer currently shows over 1.18M captures, 4.4K attestations and 562K USDC settled, with a fresh batch of attestations every 30 minutes or so. You can see for yourself at explorer.vangrid.io; web and Android have been live since July, the team is based out of Amsterdam, and they're in NVIDIA Inception. The CEO Robert Brighton came from Microsoft's Quantum Computing and HoloLens groups, and the CTO Khalid Benrafik has built systems at Booking.com, ING and Mollie which handle transactions worth $1B+ in annual transactions. The $9M, with seed funding from HashKey Capital, Animoca Brands, Borderless Capital, Crypto.com Capital, Gate Labs and Mapleblock Capital. Meanwhile, @Americanfort_io tackles the opposite end of the same problem. Most of us are still verifying a wallet address manually by checking the first and last four characters; seven matching characters caused a $50M loss. With Send-to-Name™, the copy-paste step is replaced by an @ name, and a wallet generates a one-time-only address for each payment; only the sender and recipient can derive it, only the recipient can spend it, and the payment still settles normally on the native chain. Unlike ENS-style names, a FortressName doesn't publicly point to one permanent wallet, balance or payment history. It's live across 13 networks, and FortressChat is also live. AF isn't another chain either; it's an SDK which aims to be embedded within the wallets people already use, with Litecoin Foundation, 0G Network, Tron and Arbitrum among the integrations. On the long-term roadmap, QBIP-32 is their quantum-safe key derivation approach designed to avoid requiring users to migrate funds to a new address. SafeSend™ (proof of source of funds, selective disclosure) is still in development, so I'll withhold judgment on that for now. You can claim a free FortressName at names.americanfortress.io. One makes it possible to verify the network; the other removes the need to do it by hand. Seems like a good direction to be heading in. Also, if you've been trading onchain for a while, @zerufinance on @NucleusCodes may already be taking your past wallet history into account towards a starting balance. Check yours at app.zaps.wtf. The $50K Ditto trading competition is here: app.zaps.wtf/campaigns/ditto