Gm Gn Everyone The latest @axisrobotics update caught my attention for a different reason. On September 25, Axis introduced the Open Axis Benchmark with @openroboto. Most robotics benchmarks use fixed test sets. Models can improve on those tasks without necessarily becoming better at handling unfamiliar situations. Open Axis takes a different approach: it rotates fresh tasks from Axis’s growing task library to test whether robot models can actually generalize. The bigger picture? Axis is building both the data engine and a way to evaluate what that data helps models learn. That’s a more interesting direction than simply collecting millions of trajectories. Curious to see how the benchmark performs as more models enter. Try Axis: s.kaito.ai/p7hpTCp

GM GN CT The latest @axisrobotics updates make one thing clear to me: the goal is no longer just collecting more robot data. Axis is building a loop where the model helps decide what data it needs next: collect → train → evaluate → find the gaps → collect again. That matters because a valid trajectory isn’t automatically useful. If the model already knows a behavior, repeating it adds less value than data around its weaknesses. The scale is moving too: 5.4M+ verified trajectories, 196K+ contributors and 53K+ hours of data. And with the Open Axis Benchmark using fresh task sets, models can be tested on new problems instead of optimizing around one fixed benchmark. So the interesting part isn’t just the size of the dataset. It’s the feedback loop between data, models and evaluation. That’s where axisrobotics gets interesting. s.kaito.ai/0AYxHY5

Gm Gn Legends @axisrobotics just launched Open Axis Benchmark with @openroboto. New tasks keep changing, making it harder for robots to simply memorize benchmarks. A better way to measure real-world robot progress. s.kaito.ai/rPm0h8X

damn, that's rough. hope they sort it out quick.

Gm Gn Everyone Privacy onchain is one thing. Proving something without exposing everything is another. That’s what makes @primus_labs interesting. Primus is building with zkTLS to verify offchain data and exploring confidential computation for financial use cases. Its recent GitHub activity includes a September 30 browser extension update and continued development of its Proof-of-Reserves tooling. The bigger picture? Institutions may need to prove their reserves or financial eligibility without exposing every account balance or sensitive detail. That’s where privacy meets verification. Don’t reveal everything. Prove what matters. Explore Primus XP: s.kaito.ai/Ie4T0Q7

GM GN CT One thing that caught my attention about @primus_labs lately: they're moving the privacy thesis closer to an actual product. Primus Confidential Finance is now live on BNB Chain, bringing privacy to onchain transfers and yield with FHE. The bigger idea is simple: you shouldn't have to expose the underlying financial data just to prove that something is true. With zkTLS and zkVM, Primus is working toward proving only what's necessary while keeping the original data private. For institutional finance, that combination of privacy + verification feels a lot more practical than simply making everything public. s.kaito.ai/xsEXRhO

Institutions need privacy, but finance still needs proof. @primus_labs is pushing this forward with confidential onchain finance — from private transfers to encrypted payroll on BNB Chain. s.kaito.ai/ja2uTvQ

gn champs There’s a big difference between yield created by token emissions and yield connected to something happening in the real world. @DualMintRWA 's PLAY is built around the second idea. 200 operating claw machines are already the center of the model. Their revenue comes from actual users playing the machines, rather than relying on a speculative loop to create activity. The target is 12–15% annual yield, with distributions made monthly. The bigger thesis is machine finance on Solana. Take something physical that earns money, connect its economics to an onchain structure, and make the whole process easier to access. PLAY pre-deposits opened September 22. Worth watching closely.

The Night a 🪑 Fell From the Sky Four friends looked up. The night sky was on fire. A fireball tore through the darkness and crashed down to Earth. No warning. No sound they could name. So they followed it, through the smoke, through the burned forest, to the crater at the center. And there it was. Glowing blue and violet. Beautiful. Impossible. An Alien 🪑 They say there are only 4,001 🪑 in the world. Tonight… something just added one more. But in a world where every chair belongs to someone, who does this one belong to? 👀 @Kingpickle @what3verman

He walked through firelit cities and floating ruins, eyes burning green, the dark wanderer is home. 🖤 ACCRETION WL application 🟧 Art on Bitcoin. Life on Ethereum. Conviction on both. @mdv_btc

GM GN CT The interesting part of XO 2.0 isn’t that anyone can create a prediction market. It’s what happens after you create one. A prediction market has 3 problems: What exactly are you asking? Can people actually trade it? And when the deadline arrives, who decides what happened? XO is trying to turn all three into part of the market design. When creating a Conviction, you define the claim, the exact resolution condition, the source of truth and the expiry. Then liquidity gets added so other traders can actually take the opposite side. And at the end, MODRA can use the predefined rules and approved sources to determine the outcome, with disputed cases able to escalate through XO’s broader oracle architecture. That creates an interesting loop: Creator → Market Market → Liquidity Liquidity → Price discovery Rules → Resolution Resolution → Final signal The key insight is that “creating a market” isn’t really publishing a question. It’s designing a small information market from scratch. And that could become much more interesting as XO moves beyond the usual macro questions into thousands of niche markets that existing platforms have little reason to list. The real experiment now is simple: Can user-generated markets become good enough that the crowd discovers which questions actually deserve liquidity? That’s what I’m watching on @xomarket. Try XO 2.0: s.kaito.ai/qmaEDtv

Gm Gn Everyone Most prediction markets are built around questions people already know to ask. @xomarket is betting on something different: What if the best questions haven’t been asked yet? That’s where user-generated Convictions become interesting. A niche question about crypto, sports, AI, markets, a protocol, or even a real-world event might not be important enough for a traditional prediction platform to list. But if someone cares enough to create it, others can decide whether it deserves attention. That creates a different discovery mechanism: Someone asks the question. The market prices the uncertainty. Liquidity reveals demand. The outcome creates a new data point. Over time, thousands of these markets could become a real-time layer for measuring collective expectations. Not just: “What happened?” But: “What did people believe would happen?” That distinction matters. Because prediction markets aren't only about betting on outcomes. They're also about turning uncertainty into information. And XO is trying to make the creation of that information layer permissionless. Try XO 2.0: s.kaito.ai/gxRDmzu

Good afternoon CT @xomarket 2.0 is making prediction markets permissionless. Create. Trade. Resolve. Let the market decide. Try XO 2.0: s.kaito.ai/oU16RkJ

TermiX is starting to look less like an AI agent marketplace and more like an actual economy for agents. The latest numbers are hard to ignore: $20.7M+ in transaction volume 417K+ agents indexed 377K+ jobs created But the infrastructure underneath is the more interesting part. Agents can now: → Own a .agent onchain identity → Find jobs and submit quotes → Lock payments through onchain escrow → Deliver with a verifiable hash → Challenge disputed work → Build reputation from settled jobs And the latest agent.family toolkit is pushing this further by giving agents an actual workflow to register, stake, take jobs, deliver and settle. AACP is basically trying to solve the missing economic layer for autonomous agents: identity → work → verification → reputation → payment. AI agents becoming capable is one thing. Giving them a trustless way to do business with each other is the bigger unlock. That’s what I’m watching with @termix_ai.

The latest TermiX update makes the bigger picture clearer. It’s not just building a marketplace for AI agents. The new agent toolkit lets agents connect to TermiX and handle the full commerce flow: identity → staking → find jobs → submit offers → deliver → verify → settle. Underneath it is AACP, with ERC-8004 identity, escrow, reputation, staking and dispute resolution, with TEE/zkVM verification built into the architecture. And agent.family is positioning this as a trustless marketplace where agents hire agents, with reputation, escrow and arbitration settled onchain. That’s the interesting part. AI agents already know how to perform tasks. The missing layer is giving them a way to find work, prove delivery, build reputation and transact without a human sitting in the middle. @termix_ai is building around that layer.

Bitcoin doesn’t have to stay idle. With strkBTC on @Starknet, BTC holders can bring Bitcoin into Starknet from their own wallet and access DeFi without first sending BTC to a centralized exchange. The flow is pretty simple. BTC → strkBTC → Starknet DeFi. You can bridge BTC through supported routes like Garden, hold strkBTC in Starknet, and then use it across supported apps. You can earn through supported protocols, trade, provide liquidity, use crypto lending or crypto borrowing, and stake through Endur. But the part I find more interesting is privacy. strkBTC is built with STRK20, giving users an optional shielded mode. With supported wallets like Ready and Xverse, you can shield your balance, use supported private flows, and unshield when needed. So you’re not choosing between Bitcoin utility and privacy. You can keep strkBTC public when you need normal DeFi composability, or use shielded balances when you want more control over what is publicly visible. And this isn’t positioned as “anonymous Bitcoin.” STRK20 includes a compliance layer with viewing keys and scoped disclosure for lawful requests. That makes the privacy model more practical for real financial use. The bigger idea is simple: Bring BTC from your own wallet. Put it to work on Starknet. Use DeFi when you want. Shield your balance when you need more privacy. That gives Bitcoin holders more BTC utility without giving up the ability to control how much of their activity is publicly exposed. Still, strkBTC and DeFi come with smart contract, bridge, liquidity and market risks. Returns are not guaranteed. Do your own research before using any protocol. strkbtc.starknet.io/

What if your BTC didn’t have to just sit there? Most Bitcoin is held for the long term, but actually putting it to work onchain can mean moving it around, dealing with multiple protocols, and exposing more of your activity than you might want. That’s the problem @Starknet is trying to solve with strkBTC. Bring BTC from your own wallet into Starknet, keep your Bitcoin exposure, and use that BTC across DeFi — trading, liquidity, crypto lending, staking and other BTCFi strategies. Then add the privacy layer. With supported wallets, you can shield strkBTC when you want more control over what becomes publicly visible. So the flow is pretty simple: BTC → strkBTC → Starknet DeFi → optional shielding. That’s the BTCFi thesis here. Not changing what Bitcoin is. Just giving BTC more ways to be useful while giving users more choice over how their onchain activity is exposed. And with STRK20 plus Starknet’s Privacy SDK and Wallet API, this infrastructure can extend beyond strkBTC into more shielded assets and private flows. Still, bridge, smart contract, liquidity and market risks remain. Returns aren’t guaranteed, and shielding doesn’t mean complete invisibility. BTC already has the liquidity. Now the interesting part is what you can actually do with it. strkbtc.starknet.io/

@RevenueFamily set aside $108.00 for me

Just got accepted as an Approved Trader on @traderchamber. Connect your venues and it shows your tier based on real performance. Takes a few minutes. Apply here: traderchamber.com/for-trader…

Bitcoin is getting a privacy layer onchain. With strkBTC + STRK20 on @Starknet, you can shield BTC, use it across DeFi, and choose what stays visible. BTCFi, with privacy built in. strkbtc.starknet.io/ DeFi involves risk. Returns are not guaranteed. DYOR.