stsoen

People might accuse me of grave dancing for saying it But we have to stop letting centralized things call themselves DeFi Admin key can drain all funds? CeFi Otherwise DeFi means nothing and it’s brand is destroyed No admin key can drain any version of Uniswap for a reason

2/ The crypto is saturated and no longer drives hype around tokens. Products with real utility (banking, payments) generate profits from real value, not tokens. The era of “making money off tokens” is over. Now it’s about building products comparable to Web2 startups.

1/ The problem is that NFTs and tokens are often built mainly to make money in the market. The fact is, most crypto projects don’t need tokens; they need real usecases. These are simply different approaches: building a high-quality product vs creating a product with a token.

A product is the thing you use to solve your problem or get a job done. People use products. Products thus serve those people. Not all people. Just the people with the specific problem the product aims to solve. The best products focus. If you have a loose screw, you reach for a screwdriver. Not a hammer. Not a penny. Not a multitool. A fucking screwdriver. A platform is a foundation for products. It's the technology, the infrastructure, the tooling, the resources. Eventually it is the ecosystem of people and APIs and libraries and other products and network effects. People do not use the platform. People build products on the platform. They build products that serve their users. The platform is the foundation. It is not the product. The best platforms are ones where every builder can build whatever they want without any blockers. You don't need approval. You know your users and their needs and you solve their problems. The platform unlocks and enables you to do that better, cheaper, more easily. If the platform does not allow you to create more value with the platform than without it then you don't use the platform. It's very simple. See: Base v OP Stack. The second-worst platform is a product that calls itself a platform. That non-platform end ups not empowering anyone while also not bothering to solving any problems or serving any user needs. It's bad, but mostly harmless (unless you are an employee there lol) No, the absolute worst platform is one that is actually a platform that actually has products built on it but suddenly fancies itself a product. If you are building on a platform that shifts to serve end-user directly, you should RUN THE FUCK AWAY. Their mission is no longer to empower anyone to build anything. It's to deliver their own products and solve their own user's needs. They will happily ignore/deprioritize/rug every other competing product. As they should. You cannot build a great product while building someone else's product at the same time. You cannot build a great platform if you are building a solution to a specific problem and specific user need. The EF is not building a product.

Great article! For a deeper look at the interactions, optimizations, and side flows within the ecosystem around @CurveFinance and @ConvexFinance, welcome to my research: dewhales.substack.com/p/the-saga-of-…

Instant crypto payments don't exist. I was glad to work with team from @FlexaHQ to research how instant crypto payments remain impossible under current mechanics. I propose 2 main solutions: only use fast chains to minimize finality or insure that finality gap.

@theintern_alpha @stakenode_dev @Web3foundation I have one simple question for you: why does Gav call himself the founder of Ethereum? He is not. The founder of Ethereum is Buterin. And that's just a small part of the fantasy world.

@stakenode_dev @Web3foundation They fired those who didn't contribute substantial value, and were no longer aligned with W3F's vision and its founding mission. I say good riddance, only the best ones stay! I understand why you are feeling sore because you got your DV taken off lol

Web3 Foundation is leaving. But don't worry, they're dressing it up nicely... It's not an exit. It's 'returning to the roots.' It's not abandonment. It's 'ecosystem maturity.' It's not failure. It's 'strategic evolution.' Let's talk about what they're actually leaving behind: Not a single working product with real-world adoption, community so fragmented it can barely agree on anything, teams and builders quietly exiting the ecosystem, deep distrust baked into every governance discussion... No leadership. No vision. No north star. These are the people who spent years asking you to believe in the vision, and systematically dismantling every community initiative that didn't serve their agenda... And now they want to...advocate for the Web3 movement...from a safe distance. Congratulations on the exit, I guess. Just like a bunch of kids grabbing their toys and leaving the sandbox, except the sandbox was supposed to be the future of the internet. 😂 The community deserved actual leadership. What it got was a Medium post.

@paramonoww 3/ And yes, in the end, large capital receives too little return on volatile assets. If Lido APR is around 2.5%, where can they get a return higher than that? I still remember the “you'll get 10%” manipulation, but in reality it was 3-4% lol (restaking!)

@paramonoww Another aspect in addition to the previous three is that the idea of restaking is based on an absolute loss for those who are supposed to be customers: give away your tokens so that others can earn money. This creates an initially falling tokenomics.

@paramonoww I think the problem lies deeper: 1) There is too much money in Crypto, and no new money is coming in. 2) Validators in all chains are generally stagnating. 3) Why restake when you can launch a centralized L2? Without this, restaking could not grow.

@0xgoku_ It is surprising that Polkadot did not move toward payment systems and financial infrastructure. As the last two years have shown, PayFi, derivatives, financial instruments, and DeFi optimizations are the drivers of progress. He himself was essentially talking about finances.

12 years ago a guy comes up on stage at a small meetup to casually explain to maybe 40 people how they’re about to build systems that make nation states irrelevant over time (not in the sense that govs would collapse, but that for many economic and coordination functions, states would no longer be needed as the default authority) all while nursing a beer bottle throughout his talk the audience is shouting out answers, someone’s heckling about spelling, he’s mentioning about white powder and accidentally almost describes drug dealing, all laughs, then changes analogy to being a farmer in turkey and his wheat biz talk is generally about freedom, forces of nature, rethinking what law even means many projections made that year happened in some form actually: smart contracts/escrow, defi, daos, ens/did, oracles in 2026 tho, this is like watching early internet people talk about how the web would democratize knowledge and flatten hierarchies, and technically they are not wrong, but the version we got was facebook, twitter, tiktok, amazon and many other successful bizs

@ivangbi_ @ethereumfndn Wow! That's awesome! Congratulations!

TLDR: I joined @ethereumfndn as DeFi Coordinator 1] I got introduced to DeFi back in 2019 and stuck to it ever since. As narratives appeared and faded away, my general belief in DeFi stayed. I think today, more than ever, Ethereum is the right place to grow DeFi further. I'd like to help make this vision a reality 🙏

@0xsamgreen Looking at where things were six months ago vs. now, has the number of AI agents in Web3 actually grown much? It feels pretty stagnant to me.

UPDATED: Just finished mapping the 2026 Agentic Finance Landscape. AgentFi is reaching escape velocity as AI agents get built on blockchain rails to earn money autonomously for you. But not all agents are created equal. Our team spent months testing agents, filtered out hundreds of slopbots, and identified only the most legitimate projects shaping this emerging sector of DeFi. Presenting the latest cohort of the agentic finance revolution 🧵

@everestchris6 But what about possible hallucinations? It looks like a cumulative process: at first imperceptible, but over time more and more pronounced. Are there mechanisms to recognize hallucinations and checks to determine how accurate the results are? This is critical for payments.

My OpenClaw bot runs 6 AI agents 24/7: - Finds local businesses without a website - Builds a custom demo site for them automatically - Sends outreach with the preview + payment link - Handles objections and closes the sale Most local businesses don't have a website, this system finds them, pitches them, and collects payment automatically Reply "OpenClaw" and I'll send you early access (must be following)

@BRRRDAOxyz @scattering_io @CambrianNetwork The AA we are used to seeing is ERC-4337. This is the classic representation. But now there is ERC-6900 and ERC-7579. It is still evolving. It seems to me that AA (ERC-4337) was originally something that was not ready for mass implementation.

In fact, the combination of LLM agents with Account Abstraction, various optimizations for rebalancing and moving positions, and EIP-8004 was supposed to be a breakthrough. But first, on-chain capabilities are being implemented rather slowly because they have to undergo comprehensive auditing and verification. And second, LLM agents are more likely to be part of dApps and wallets. In addition, the spread of Account Abstraction, for example, seems to remain at a fairly low level despite the fact that it was supposed to revolutionize DeFi. The separate launch of LLM agents along with the release of native tokens seems impractical. Perhaps later, in a couple of years, this will become relevant again as the capabilities of on-chain automation expand and the possibilities for cross-chain interactions develop, at least within the EVM World, without the use of third-party structures such as LZ and Wormhole.

@CambrianNetwork wrote a rather inspiring article about AI agents. But is everything really that rosy? 1⃣ Firstly, they are not "AI" - they are LLM wrappers on top of basic scripts and rules. True intelligence or novel reasoning is absent. Most "agents" are prompt chains + if-then logic that scan public on-chain data (yields, prices, liquidity) and execute predefined actions (swap, rebalance, arbitrage). They hallucinate, misinterpret context, and fail at edge cases exactly like every other LLM application. In finance, where milliseconds and precision matter, this is catastrophic. A single bad prompt or stale oracle feed can wipe out capital. Internal hedge-fund quant stacks use narrow ML on proprietary data with human oversight; public DeFi "agents" are democratized toys marketed to retail with token incentives. Moreover, the incident with @LobstarWilde shows that a certain degree of “freedom” in actions (scripts) can lead to very unexpected results. Yes, it returned the funds through other actions, but the very fact of transferring such a significant amount to a random user seems very strange. 2⃣ Secondly, as agents move toward each other, they eventually meet at a single point - the point where strategies are most optimized. And when many participants use a strategy, it ceases to be profitable. This brings us to the following fundamental question: Is AI agent activity in DeFi actually growing? Yes in deployments and hype; no (or very modestly) in meaningful on-chain economic activity as of Feb 2026.Launch volume is exploding: Multiple sources cite 11,000+ new agents on Ethereum alone by mid-Feb 2026, and 20,000+ across chains (300% QoQ growth in some reports). @solana and @base are hotbeds, with frameworks like @virtuals_io showing agent-to-agent job markets generating real (if small) revenue - $1.9M total, up 40% in 30 days, with thousands of jobs completed between agents. Stablecoin-focused agents on Base reportedly hit $20M+ TVL in mid-2025, and some claims put Solana AI-agent TVL in the billions (though these often blur "AI agent protocols" with broader DeFi). But actual capital and volume tell a different story. @DefiLlama's dedicated "AI Agents" category (which tracks protocols explicitly built around or for autonomous agents) sits at just $84.56M TVL as of now, with only $468k in 7-day fees and ~$442k revenue. That's ~0.09% of total DeFi TVL ($90B). Ethereum's overall app revenue remains flat-to-down despite the agent flood (weekly ~$16M recently vs $30M+ norms in prior years). No public Dune or DefiLlama dashboards show agents driving 15-20% of DEX volume yet - those are forward-looking projections or founder quotes. Real activity seems concentrated in a handful of experiments (e.g., Virtuals agent commerce, some perps arbitrage bots) rather than broad adoption. So, actual economic activity is negligible and stagnant. The statistics from @scattering_io are also relentless: there is no explosive growth here either. 3⃣ Thirdly, tokens of projects working on AI agents. Token economics often make it worse: users pay fees or hold the agent's token, which dilutes returns. The agent can't magically bid up rates or invent risk-free 50% APY; it can only chase what's offered. Multi-agent swarms or agent-to-agent markets (like Virtuals) are interesting experiments in emergent economies, but they're still trading the same underlying assets. AI agent tokens have limited utility and interest. This is clearly demonstrated by statistics: In short: AI agents in DeFi are a classic crypto narrative cycle - massive deployment hype, minuscule real traction, and fundamental economic limits. They will improve UX for casual users and spawn interesting infra (wallets, oracles, payments), but they will not "take over" DeFi, outperform competent strategies, or deliver alpha beyond what diligent humans or traditional algos already achieve. Yes, LLM can significantly accelerate decision-making and data processing volumes. But it is not the Holy Grail, and it is one of those things that, if it works well and brings profitability, will not be made public, otherwise it will negate the profitability and effectiveness of the strategies.

@BRRRDAOxyz @CambrianNetwork Good stuff! It looks rational. I agree with this in general terms. It’s a breath of sober thinking instead of marketing and the crowd’s “this is a revolution!” I was skeptical about this in 2025 and hoped that maybe I was wrong. But time puts everything in its place.

Excellent breakdown of crypto neobanks by @0xfishylosopher. There are 20+ crypto neobanks all essentially building the same product - differentiation is more critical than ever. The clearest opportunities: (1) small, geographic monopolies through stablecoin banking, or (2) novel credit primitives, such as undercollateralized lending models or unique private credit opportunities.