Eldar | Polymorphic Capital

With @Visa and x402 (@coinbase), @Nevermined_ai is turning agents into transacting entities, unlocking a new market where machines don’t just use services, they pay for them at internet scale. Content, APIs, and data can now be monetized per request, creating entirely new revenue streams for digital businesses. > Machine-native micropayments > Delegated spending without human-in-the-loop > Merchant acceptance via existing global rails Any protocol. Any rail. Any price. This is how new economies start. Great work, team 🚀

Most people see AI agents as productivity tools. We see them as a new category of economic participant. The gap: agents can't hold persistent spending authority. Every transaction requires a human back in the loop. @Nevermined_ai just shipped the rails that fix this in end-to-end agentic commerce stack. Four industry innovations end-to-end: 1. Agentic tokens (Visa/MC's new token type – agent transactions are now visible/auditable) 2. Persistent delegation (months (not seconds) of spending authority, hard limits) 3. x402 micropayments (HTTP 402 finally works after 28 years) 4. Merchant acceptance via existing PSPs Think of @Visa building VIC and @Mastercard its AgentPay specifically for this. @Nevermined_ai just shipped the end-to-end agentic commerce stack. Check it out.

For those who’ve been around for a while, IDEX was never just another DEX – it was an era. Long before fast L2 DEXs, long before Uniswap, back in 2018–2019, IDEX was the top app on Ethereum and basically the only place where you could properly trade tokens in a non-custodial way. Yes, it was still slow, order books were thin, and gas could cost a fortune. But for its time, it was a real breakthrough: non-custodial trading, a functioning order book, matching engine, market orders. Big thanks to @AlexWearn and the team for years of hard work and for making an invaluable contribution to the creation of this industry. True pioneers.

🚨Alpha alert 🚨 @tempo just went live a few minutes ago. With the mainnet launch, they’re also launching the Machine Payments Protocol spec. I received early access to the specification, and have some initial thoughts. First off, the MPP spec is a new, alternative take on the 402 “request for payment” http code. It differs fundamentally from x402 in that it removes the Facilitator from the architecture. On the one hand, removing an intermediary from the flow is usually a good idea. On the hand, it moves processing logic handled by the Facilitator to the merchant Server. Time will tell if this is good or bad, but my guess is that there’s a belief that the server side agent will be able to optimize the merchant flows, so having a 3rd party do that is redundant. Now we need to wait and see… The MPP spec also designs for different payment methods beyond stables, including support for fiat via credit card payment, which I think is a win. Adding to this is a PSP agnostic approach, which I thinks a double win. Merchants should be able to choose whatever payment method they want, processed by whatever payment service provider they want. Now onto the questions I have. First, for card payments the spec has a novel encryption method where the Server (i.e. the Merchant) controls the decryption of the Client's payload. Unfortunately, this raises questions about PCI compliance. The Server doesn’t control how the Client (the buying agent) manages its card delegation. The prescribed route seems to be DPANs, but raw PANs are also in scope based on the spec. Because the Server doesn’t control the CHD payload, raw PANs can be encrypted and sent to the Server for forwarding to the Server’s PSP. This, in turn, would require the Server to be PCI compliant unless it can guarantee the payload is only ever a DPAN. Requiring every Server to be PCI SAQ A or, worse yet, SAQ D compliant is not scalable. Second, the spec enables multiple payment methods, but doesn’t resolve transactions globally. While this might look normal, for agent-to-agent payments to scale, imo having a global ledger that all agents are able to plug into is the optimal approach. Why you ask? For a number of reasons, but one specific reason is identity establishment. Agent identity will be tied to statistical modelling based, in part, on transactional history. The contemporary example of this is credit risk profiling. However, compartmentalized transaction networks will lead to disparate transaction history, which will cause issues when trying to resolve behavior. Congrats to all involved, including our partners at Visa and VGS. The space is making massive strides. Things are starting to come together, and we’re looking forward to putting the pieces in place.

🇨🇾 @rise_pay EOR is now live in Cyprus. Hire full-time employees with no local entity, fully compliant payroll, taxes, benefits — and payouts in fiat, stablecoins & crypto. Cyprus just got a lot easier to hire in. Global work, unlocked. riseworks.io/rise-countries…

The global remittance market is vast and growing, with increasing demand for lower fees and modern digital infrastructure. Proud to back @theaadedewe and the whole @kredete team again in their lastest Series A. @kredete is building a stablecoin-powered financial ecosystem for the African diaspora. Starting with remittances – a lifeline for some immigrant families – the platform is expanding to help users Send, Save, Spend, and Build Credit. From powering $500M+ in remittances to launching Africa’s first stablecoin credit card, they’re reshaping access to finance for 700K+ users, and now expanding to Canada, UK & Europe.

1/9 We’re excited to announce our investment in @TACEO_IO , an applied cryptography team building a private shared state network for the decentralized internet. @TACEO_IO is solving a problem that’s been blocking privacy adoption at scale 🧵👇

believe in somETHing 💜

If I’m not mistaken, Zama was never too eager to be a Web3 company. No interest in launching a token, saw way more upside in AI privacy — the "real serious" stuff. So to tap into the Web3 hype without getting their hands dirty, they licensed their tech to projects like Fhenix and Inco, which raised a bunch of VC money building FHE-powered chains and tokens. So what does the latest news mean? 1. Zama couldn’t make the business work and is now going full degen, dropping a token to salvage monetization? 2. Fhenix and other Zama-powered side quests are about to get rugged?

A taxi driver sending $500 in stablecoins each month to support his family in Africa is a meaningful Web3 user, creating real value for the ecosystem. A degen cycling $500K through protocols to farm token incentives — and jumping to the next one tomorrow — is not.

Built an AI agent but don’t want to implement a whole backend just to get it online? ⚡We built something for that — deploy your agent as an A2A server in minutes. auto-a2a.ai 🧵👇

𝐌𝐀𝐍𝐒𝐀 𝐇𝐢𝐭𝐬 $𝟏00𝐌 in Total 𝐓𝐫𝐚𝐧𝐬𝐚𝐜𝐭𝐢𝐨𝐧 𝐕𝐨𝐥𝐮𝐦𝐞! 🎉 This milestone reinforces the growing demand for seamless, instant liquidity in global payments, and we’re just getting started. Onward and upward! 🚀

If operating systems had tokens and a retail market, many VCs would still be funding Windows & Android killers — instead of what’s been built on top of them

@jbrukh Unmatched @0xbilly at @0xIntuition

Who is working on web3 knowledge graphs for AI?

The 2024 bull market was disconnected from Web3 from the very beginning. Its main drivers came from outside the ecosystem — ETFs, regulatory shifts, reserves, and broader macro trends. Likewise, the current market correction has little to do with Web3 itself. The real problem with Web3 isn’t trading tariffs — it’s that we still haven’t delivered truly valuable products to a broad audience, products whose utility doesn't hinge on the state of the crypto market. While the number of active onchain users is measured in tens of millions, the uncomfortable truth is that no protocol or company can sustainably justify billion-dollar valuations. From the start, this bull market felt unearned. There was no clear justification for a dramatic revaluation of the Web3 space — no breakout use cases at scale (with rare exceptions like Social or Prediction Markets), and no significant user growth. Unlike 2017 or 2021, this cycle was driven prematurely by external forces, not by internal innovation or adoption. As a result, the growth turned out to be even more speculative than in previous cycles. Once again, the bull market shifted focus from products and users to tokens and hype. Many builders were lured into the casino, playing the game of narrative-chasing and short-lived multi-billion-dollar FDVs. In this game, the least principled tend to win — those who over-invest in token marketing and government relations. Meanwhile, teams working on real Web3 infrastructure and utility saw little attention and even less market reward. So what's the bottom line? A paradox. This time, the market decline isn’t rooted in internal failures like FTX or Luna. In fact, Web3 fundamentals have been steadily improving and are arguably stronger than ever. Scalability and user experience — long-standing pain points — are finally being addressed. And if you look past the memecoin casino, the number of non-speculative Web3 users is hitting all-time highs, largely driven by real-world use cases like stablecoin payments. The tourists are leaving. The builders remain. So does Web3.

We met @YassineLanda before the ai agents hype, and we knew he had what it takes to build a durable AI <> crypto company. @mbdtheworld is the leading feed recommendation provider for crypto companies Read more about the story here: @socialgraphventures/mbd" target="_blank" rel="nofollow noopener">paragraph.xyz/@socialgraphve…