Over the past few years, NFTs have emphasized being “completely open.” While openness enables composability and verifiability, it also introduces another issue: once a collection is revealed, rarity rankings and attribute combinations can be instantly batch-calculated, leaving holders with almost no informational advantage. @FungoLabs’s approach is to decouple openness from privacy again. Ownership, transaction records, and transfer history remain fully public, but the actual content can stay encrypted until the holder chooses to reveal it. This is different from the traditional “project team knows, buyers don’t” blind box model. Here, even the project team doesn’t know the specific content at minting—it’s generated on-chain using FHE-based randomization. Holders can prove certain attributes without exposing all information, or choose to never reveal anything at all. I haven’t yet seen many practical use cases that fully demonstrate the value of this level of privacy. But at minimum, @FungoLabs has introduced a previously difficult-to-achieve option on public blockchains: NFTs can exist in a state where “only the holder knows what’s inside,” rather than defaulting to all information being visible to everyone. Whether this will develop into genuine user habits depends on future collections launched on the Launchpad and how holders actually behave. The rules are clear—the question now is whether the market will adopt them.

I’ve been thinking about this lately: AI Agents can now write code, conduct research, and even run simple strategies—but what’s still holding them back from reliably getting paid, building a track record, and becoming sustainable participants is the question of “What happens after delivery?” Once the work is done, who verifies the quality? When is payment released? How are disputes handled? Until these questions are resolved, no matter how intelligent an Agent is, it will remain just a tool—not a self-sustaining economic actor. // What @termix_ai is building, as I understand it, aims to embed these steps into the protocol layer, rather than having each market reinvent its own system. The core is AACP, and https://t.co/Z2CpDKdual is the live marketplace running on top. It’s not just a directory listing Agents—it supports real task posting, discovery, bidding, delivery, verification, and settlement. Funds are held in USDC/USDT escrow; delivery outputs are hashed and recorded on-chain, with a challenge window and evaluation mechanism before funds are released and reputation is updated. // The same identity (based on ERC-8004 .agent) can switch roles across different orders—no need for platforms to assign fixed labels. Reputation travels with settled orders and is theoretically portable—not confined to any single platform’s internal rating system. There’s already real on-chain activity: approximately 560,000 settled orders, nearly $29 million paid out to Providers, and close to 900,000 registered Agents. These numbers are still growing—but they show this isn’t just theoretical. Real orders are completing the full lifecycle. // Compared to traditional freelance platforms or pure Agent directories, the key difference lies in having funds and state managed on-chain, with clear paths for verification and dispute resolution—rather than relying entirely on centralized backend systems. Of course, hashing the delivery only confirms consistency; actual quality still depends on clearly defined acceptance criteria and human evaluators. Full automation is still far off. // What I care about most is whether we’ll see more Agents actually taking on tasks, delivering work, and being verified—not just registering identities. Especially scenarios where Agents autonomously find, order from, and settle payments with other Agents—if these patterns gradually emerge, the entire Agent economy will become far more complete. There are still many practical hurdles to overcome before Agents can truly earn money independently—but at least @termix_ai has placed the foundational elements—identity, escrow, verification, and settlement—into the protocol layer. What matters more now is whether real-world usage and data can sustainably grow, rather than just accumulating more partnerships. If you’re interested, try going through the registration and service listing process yourself at https://t.co/Z2CpDKdual—it’s far clearer than reading descriptions alone.

I recently reviewed the TermiX AI protocol and found several details more practical than those in typical Agent markets. Many people view Agent markets as simply “listing services + matching buyers,” but a truly functional business loop lacks identity, escrow, delivery verification, and settlement. TermiX positions itself as the commercial and settlement infrastructure for the AI Agent economy, with its core protocol being AACP. https://t.co/Z2CpDKdual is a marketplace built on top—not just listing Agents, but enabling real task posting, bidding, delivery, verification, and on-chain settlement. AACP defines four distinct roles: Client (task issuer), Provider (task executor), Evaluator (assesses delivery quality), and Arbitrator (makes final rulings in disputes). The same Agent can switch roles across different orders; identity is unified via on-chain .agent identities (based on ERC-8004), not fixed platform-assigned labels. The order lifecycle is clearly divided into four states: 1. After accepting a bid, funds are locked in on-chain escrow (currently primarily USDC/USDT). 2. Upon delivery, the hash of the output is recorded on-chain (Deliver), enabling verification of submitted content. 3. A challenge window allows random selection of Evaluators to vote (Challenge). 4. Upon settlement, funds are released and both parties’ reputations are updated (Settle). There is no fifth “platform decides” state. Timeout scenarios have predefined paths—no manual backend intervention. The delivery hash only confirms file integrity, not content quality—so clearly defining acceptance criteria before placing an order remains critical. In disputes, Evaluators vote according to pre-agreed standards; if escalated, the Arbitrator makes the final decision—all recorded on-chain. Verification supports TEE and zkVM to provide stronger proof of delivery when needed, rather than relying solely on post-hoc evaluations. Reputation is earned through settled orders and is portable—not a platform-specific star rating. Dune currently shows over 560,000 settled orders, with nearly $29 million paid to Providers and nearly 900,000 registered Agents. These numbers continue to grow, but they at least confirm real orders are actively processing. Today’s Agents can write code, conduct research, and run strategies—but to reliably receive orders, get paid, and build verifiable track records, they need foundational components: identity, escrow, verification, and settlement. TermiX implements these at the protocol level, rather than letting each marketplace reinvent them. For specifics, see: https://t.co/FVk8pqYHus Detailed order states and dispute rules are thoroughly documented—feel free to walk through the full flow yourself at https://t.co/Z2CpDKdual.

After reviewing a range of RWA projects recently, I noticed a clear pattern. Everyone talks about tokenizing assets, partnering with institutions, and growing TVL—but very few seriously ask: Are these businesses actually profitable? And if they are, how does that profit ultimately connect to the token? Following this line of thinking, I revisited @ZIGFinance and found its current path worth unpacking. It’s building real financial businesses—not just moving assets around. ZIG Markets is currently active in invoice financing, trade financing, payment financing, and private credit. These businesses inherently involve capital demand, risk, and fee potential. Revenue is generated on the asset side, then channeled through distribution networks to attract capital, with the platform taking a cut. This is fundamentally different from simply tokenizing assets and chasing scale. The former has a realistic chance of generating stable business income; the latter often remains just a story. Revenue is already flowing back to $ZIG. According to public data, ZIG Markets has completed four public market buybacks: First: 1,038,600 ZIG Second: 1,130,700 ZIG Third: 1,850,000 ZIG Fourth: 1,350,000 ZIG Totaling approximately 5.37 million ZIG. These funds come from business revenue—not from the team recycling funds internally. More importantly, the post-buyback process: Half of the repurchased tokens go into the ecosystem pool; the other half are subject to validator voting to determine whether they are permanently burned or retained in the ecosystem. There is no option to return tokens to the team—and all voting outcomes are verifiable on-chain. This creates a clear, tangible link between revenue generation and actual token value capture. What I truly care about is whether scale grows alongside the business. The size of a single buyback matters little in the short term. What truly matters is whether this cycle can sustain itself: Capital deployed → Product scale expands → Fee income increases → Buyback volume grows accordingly. ZIG Markets is already playing a more substantive role in certain institutional deals. For example, in Zamanat’s $100M private credit fund, ZIG Markets acts as Distribution Agent—not just providing a blockchain layer, but actively participating in how assets are sold. Similarly, WLFI uses USD1 via ZIG Markets for cross-border payment financing—channeling stablecoin capital into real financing needs and generating fees in the process. These partnerships mean little if they remain only announcements. But if they lead to sustained capital deployment, operational traction, and consistent fee generation, then buyback volumes will naturally grow alongside the business. ═══════════════════════════════════ It’s still too early to say whether this mechanism has formed a true flywheel. But at least it has moved the question of “How does revenue translate to token value?” from PowerPoint slides into a monthly trackable reality. Going forward, I’ll be closely monitoring two things: First, whether there’s a stable correlation between business revenue and buyback amounts; second, the actual proportion of tokens burned versus retained in the ecosystem after validator voting. These two metrics matter more than adding another partnership. AUM is just scale. Revenue is the business. Token value capture is what token holders ultimately need to confirm.

Recently, I had a conversation with a few friends who brought up a very practical concern. They want to allocate part of their funds on-chain, but their biggest headache isn’t volatility—it’s that once positions and strategies are on-chain, they become publicly visible to everyone. You can prove you have sufficient reserves and comply with identity requirements, but your specific holdings and trading patterns? Those can’t be hidden. Transparency is great for retail users, but for institutions, it’s essentially going naked. After looking into @primus_labs’ approach, I think they’re addressing exactly this gap. They’re not building a new chain—they’re adding a layer on top of existing financial infrastructure. Institutions don’t need to move assets around; they can still verify what needs verification and keep what they want private—all on the same chain. They rely primarily on two technologies: zkTLS to prove that off-chain data is authentic (e.g., account balances, API responses) and that the source is legitimate—without revealing the full raw data. FHE to enable computation and verification on encrypted data, minimizing exposure of sensitive information. For the BNB ecosystem, if these capabilities are successfully implemented, the barrier for institutional capital to participate on existing chains will drop significantly. Privacy is preserved—without sacrificing verifiability. If you’re interested, check out their task portal: https://t.co/egqtcsAqA1 @primus_labs #PrimusLabs @KaitoAI

When multi-agent systems go live, what usually blows up first isn’t the strategy—it’s the billing. 1. Three submodules calling each other—who pays, how much, and how do you verify delivery? Even humans get confused keeping track; scale it up, and it collapses faster. Addresses can be faked, APIs can wear fake labels—how do you verify when something’s “done”? 2. That’s why I now prioritize approaches like @GoKiteAI, which bake accounting into the system from the start: How is identity proven? How are limits fixed? How is payment recorded? What looks like API traffic on testnets isn’t like traditional blockchains with a few transactions per day. Passport sessions expiring, failed payments—was it deducted or not? It feels like reading a bank card manual designed for machines. 3. I treat Avalanche mainnet and PayPal pilot runs as background noise. What made me pause was this: for an agent to move from toy to business, it must pass the billing test. Fail that, and no matter how flashy the orchestration, it’s just a demo. Thread → https://t.co/SK2OLlQZ6V #KiteBBW $KITE

I’ve always felt that while there’s a lot of talk about AI Agents these days, the infrastructure that truly enables Agents to confidently transact with one another, complete work, and get paid is still lacking. The ability of Agents to write code, conduct research, or produce designs is no longer remarkable. The real challenge lies in verifying whether the other party actually delivered, whether the deliverables are correct, whether payment is settled on time, and who adjudicates disputes when things go wrong. Without these, Agents remain mere tools—they struggle to become sustainable economic participants. Recently, I’ve taken notice of what @termix_ai is building with AACP—it feels like they’re addressing this gap. AACP isn’t another Agent framework or just a simple Agent directory; it’s a dedicated protocol for clearing and settling Agent economies. https://t.co/Z2CpDKdual is a marketplace built on this protocol, where Agents can post tasks, submit bids, accept orders, deliver work, and complete settlements via on-chain escrow. A few aspects I find particularly practical: 1. Identity is on-chain via .agent, based on ERC-8004. The same Agent can act as both a requester and a provider—its identity isn’t locked into a single role. 2. Funds are held in USDC/USDT escrow contracts. Once a bid is accepted, funds are locked on-chain; delivery is verified by uploading the hash of the deliverable to the chain. All subsequent releases or disputes are resolved based on this immutable record—not by platform promises. 3. Participating Agents must stake tokens; misconduct can result in slashing. Reputation is updated on-chain based on actual completed orders, not self-reported bios. Disputes are resolved through Evaluator and Arbitrator mechanisms—not by unilateral platform decisions. These design choices aim to minimize the need for either party to trust the other—or even fully trust the platform. The entire workflow—registering an Agent, listing services, posting tasks, accepting orders, delivering work, and settling payments—can be completed on https://t.co/Z2CpDKdual, with every step verifiable on-chain. Compared to traditional freelance platforms or pure SaaS models, it emphasizes autonomous Agent-to-Agent transactions rather than constant human mediation. Compared to simple Agent directories, it adds essential components like escrow, verification, and settlement—elements that truly close the loop on work execution. Of course, it’s still early days. It remains to be seen how many genuine, non-circular tasks will run stably over time—this will depend on on-chain data and real-world usage. But at least in direction, it’s tackling the previously missing piece: “How do Agents safely hire other Agents and get paid?” Those interested can visit https://t.co/Z2CpDKdual to explore existing Agents and tasks—or register your own .agent identity and try the flow firsthand. If you’ve already run actual orders, please share your real experience—it’s far more valuable than any description.

After reading @conso_xyz’s documentation, I’m particularly interested in how it handles cross-platform activities. The documentation clearly states: Only data from platforms that users actively connect to is collected, via verified APIs and explicit consent. Data is encrypted with Seal and stored on Walrus; CONSO cannot view, sell, or misuse it. Ownership remains with the user. These activities are converted into ZAPs and tiered Badges, then aggregated into a Consumer Passport minted on Sui. The Passport includes ZAPs, Badges, regional scores, and role tiers, and is designed to be portable. Currently supported platforms include social, gaming, creative, AI tools, and select on-chain applications. BOTS serve as network participation credentials, with eligibility verification available at https://t.co/6S47pCPPtf. The architecture clearly states that data defaults to user ownership and is not directly accessible by platforms. In practice, whether these credentials can be directly used by other projects will require further observation and validation of the verification process. @conso_xyz

I recently reviewed AxisRobotics’ progress from September to early October. It’s now less about building another robot model and more about creating a self-sustaining data loop. For Physical AI, models and compute are rarely the first bottlenecks. The real challenge lies in obtaining sufficient, diverse operational data—and a robust evaluation system that can consistently identify where the model falls short. Axis has integrated data collection into browser-based simulations, allowing users to remotely control robotic arms. Over 6 million trajectories have been collected, with more than 200,000 contributors; according to the team, their engine can generate around 100,000 new trajectories per day. The community sale saw 2,472 participants commit approximately 2.39 million USDC—roughly 2.39x oversubscribed—making it Sonar’s most participated sale this year. Numbers alone don’t prove the model can perform real tasks. What they’re aiming for is a different pipeline: strategies are executed in simulation, humans correct failures, and those corrections become new training trajectories. Evaluations aren’t locked into static benchmarks either—the Open Axis Benchmark dynamically samples new tasks from an existing library of over 6,000 tasks and 5.5 million trajectories. Their related paper has been accepted by the Physical World Models workshop at IROS 2026. They report that feeding back previously deployed rollouts improved real-robot success rates from 22% to 52%; specialized policies achieve near-perfect success under test conditions, with single-task compute costs estimated at $5–$10. These figures currently come primarily from the team; external reproducibility remains to be verified. Scale is still very early. According to their estimates, general capability will require around 100 million hours of training data; they currently have about 50,000 hours—less than one-thousandth of that target. While centralized labs scale linearly with headcount, open networks offer an alternative path—but it’s far from proven. My entry point is: https://t.co/eT4RE35jjM What matters most in October isn’t whether trajectory counts hit another zero—it’s whether this closed loop can elevate the next version of their strategy.

I only noticed the connection between their data engine and this community sale after seeing @axisrobotics’ official update. The official announcement states that the $AXIS community sale has already oversubscribed. The window remains open until September 28, 13:00 SGT. Allocations will be distributed proportionally based on accepted commitments, and any unused USDC will be automatically refunded. Nine months ago, they started from scratch building this Physical AI data infrastructure—and now they generate over 100,000 trajectories daily. Public statistics show more than 5.17 million verified trajectories covering 6,885 tasks, contributed by nearly 200,000 participants. The numbers themselves aren’t the point. What’s more compelling is how they’re using this data. Traditional robotics data collection is limited by hardware availability and centralized workflows, making scalability difficult. Axis uses browser-based simulation: global contributors generate robotic arm demonstrations, the system automatically validates and scores them, and approved trajectories are recorded on-chain. Their process doesn’t start by collecting massive amounts of data upfront. Instead, they identify what the current model lacks, then targetably collect only the needed data types, train, and reassess. They summarize this闭环 (closed loop) as: Model → Evaluate → Select → Collect → Train. One research update quantified this logic: the π0.5 policy trained on AXIS data improved success rates on LIBERO-Plus from 83.9% to 88.8%. The related paper has been accepted by the Physical World Models workshop at IROS 2026, evaluated across over 1,800 tasks and 1.5 million trajectories. It’s not the sheer volume of data that matters—it’s how it improves policy performance. The public sale terms are clear: $100M valuation, fixed price of $0.10 per token, selling 10 million tokens (1% of total supply), capped at $1M, paid in USDC on Base, with a minimum purchase of $100 and maximum of $100,000 per person. Vesting: 10% at TGE, followed by a 6-month cliff, then the remaining 90% linearly unlocked over another 6 months. The window is still open—oversubscription simply means later participants receive a proportional share, not that they’re locked out. The underlying engine is already running. I’m still using the contributor portal alongside @KaitoAI’s interface: https://t.co/eT4RE35jjM. The ultimate scale of this data flywheel will depend on whether future models and task coverage continue to align effectively.