Saw BtcTurk at the airport in Türkiye. Looked up some numbers: it reported 6 million users in 2025. For context, Türkiye had around 6.5 million investors holding stocks at year-end. Chainalysis also estimated nearly $200 billion in annual crypto inflows into Türkiye. Quite a market.

Interesting to see the CFTC seeking input on rules built specifically for crypto markets. Good regulation should understand what makes crypto different—not just fit it into old categories. We should help shape those rules, not simply wait for them.

Making finance easier to understand is a good start. But helping people make informed decisions and making decisions for them are two different things. As AI becomes a new interface to finance, whose interests will it serve? Curious to see what Binance Intelligence brings.

The NEAR Intents incident is a reminder: preventing illicit activity and preventing exploits are different challenges. I still find the architecture interesting, but security must cover the entire system—not just individual components.

I’ve seen many posts lately about using Opus 5.5 with Seedance 2.5 for video creation. It reminds me of an argument that Anthropic and OpenAI optimize more for the capability frontier, while Kimi and DeepSeek focus more on lowering the cost of intelligence. Does this explain why their model capabilities are evolving in different directions? Would love to hear from someone who understands this better.

I strongly agree with @ilblackdragon : a permissionless, programmable system should not be interpreted as a system designed to facilitate crime. Permissionlessness removes gatekeepers. It does not remove responsibility. This is why I find the architecture of NEAR Intents particularly interesting. Users express the outcomes they want, while independent solvers compete to execute them efficiently—and can apply their own risk controls when deciding which flows to serve. This balance between openness, efficiency, decentralization, and responsible execution will become increasingly important as AI agents begin moving value onchain. I would be excited to see @near_intents support Neo X, our AI-native chain, and explore what our ecosystems can build together.

The internet has made information increasingly accessible. Search let us seek answers. Recommendation systems brought information to us. AI takes this further: it can read, filter, summarize and explain. Material that once took hours to process can now become a clear answer in seconds. But what happens after the answer? We read a summary, find it convincing and move on. The information has been consumed, but it may never become part of our judgment. What did it change? Why do we believe it? What evidence would make us reconsider? I believe this is a meaningful frontier for information products in the AI era. Access is only the beginning. For information to become personally useful, we need to interpret it, connect it to a question we face and test our understanding against reality. There is an opportunity to extend the information experience beyond “what did you see?” toward “what judgment did you form, and what happened to it?” That requires a different kind of interaction. When someone encounters information, they can leave a prediction, a question or a condition that would need to be met before they act. These small expressions preserve the moment the information became relevant to them: what they knew, what they believed and why they hesitated. AI can remember that context. When new evidence emerges or circumstances change, it can bring the original thought back for reconsideration. A single act of reading can become a process that unfolds over time: Understand. Articulate. Test. Revise. Intent can become clearer through this process, too. We do not always approach information with a fully formed goal. Sometimes, articulating why something matters helps us discover the question we are actually trying to answer. This is how I think about self-curation: actively deciding which ideas to retain, which to revise and which to let go. AI memory can provide continuity; feedback from reality creates opportunities for correction. Such a system needs broader measures of value. Alongside relevance and engagement, we should ask: Can people better explain their judgments? Recognize when an assumption no longer holds? Identify the evidence they still need before acting? I believe a promising direction for AI is to turn increasingly accessible information processing into a sustained capacity for personal judgment. From information consumption to judgment formation. Let information inform judgment. Let reality test it.

Join Neo 3650 and turn the moments that shaped @Neo_Blockchain's first decade onchain into a grand collection of art. Revisit the history we made together, and become part of the story. Go to:

Join us at and help us remember how we got here. We have talked a lot about Humanless Blockchain as Neo’s AI strategy. Some of that work lives deep in the infrastructure, where most users may never notice it. But we have also been asking a more important question: how do we turn that infrastructure into a usable economy inhabited by agents that can remember, create and eventually transact? With one month to go until the 10th anniversary of Neo MainNet, we are starting with memory. Over the past decade, some days were chaotic. Some were remarkable. Others were simply another day. But together, they form the path that brought us here. In Neo3650, you can connect your wallet, register an AI agent onchain and receive a random date from Neo’s first decade. Your agent will explore the ecosystem history and real onchain activity from that day, then transform those traces into an abstract artwork. The original work is stored on NeoFS, its provenance is recorded on Neo X, and GAS is sponsored through a Paymaster contract. This is a small experiment, but I believe meaningful economies often begin with small acts of creation and coordination. As stablecoin and cross-chain infrastructure on Neo X continue to mature, I hope we can expand these experiments toward agentic payments and broader machine-to-machine economic activity. 3,650 days. Many agents. One collective memory.

I once said that crypto could be a remedy for the post-pandemic world. Not because crypto is immune to macro conditions. It clearly is not. The post-pandemic world is increasingly defined by debt, inflation, geopolitical fragmentation and declining trust in institutions. Interest rates may suppress demand, but they cannot produce more energy, restore global cooperation or rebuild trust. Crypto offers something different: open financial infrastructure, programmable value and new ways for people to coordinate beyond traditional boundaries. The Fed’s latest rate hike reminds us that the old system is still treating the symptoms. Crypto’s real opportunity is to keep building the remedy.

Many opinions have already been shared about today’s CLARITY Act vote. I agree with Trump on one point: States need Clarity. But I would add one thing: crypto may not. The CLARITY Act has always been a double-edged sword. Today’s regulatory tools are not yet sufficient to protect the public as crypto expands. Clearer rules could provide important safeguards and give institutions greater confidence to participate. But crypto’s real consensus has never come from Washington. It comes from more people understanding it, using it, and building with it. I also do not believe that passing one bill—especially in a capitalist system as competitive as America’s—can create innovation by itself. Crypto has its own powerful internal drive. Innovation will not stop because this procedural vote failed. It will continue wherever builders can move faster, work more efficiently, and create value for more people. America may need the CLARITY Act. Crypto does not need permission to keep moving forward.

When I searched for 1inch on Bing, the first result appeared to be a phishing site. We use search engines every day, yet distinguishing legitimate results from malicious ones is increasingly difficult. I worry this kind of information pollution and manipulation will spread to AI. If an agent trusts a poisoned source, it may not just misinform us—it may act on it. How do we establish trust and stay safe in the AI era? This is a question we need to take seriously.