The mechanism that priced Google's IPO has never worked on-chain. We're opening Raise Zero - the first public sealed-bid uniform-price auction on @solana Every founders today face the problem of how to price tokens fairly while avoiding bots or whales outbidding everyone. Sealed-bid uniform-price auctions are how Google IPO'd in 2004. How the U.S. Treasury clears $20T+ in bonds a year. They're the gold standard for credible price discovery - but they've never worked on-chain, because blockchains are transparent by design. Sealed bids need privacy that didn't exist... Together with @Arcium we're changing that. Their MPC network runs computation directly on encrypted data, without ever revealing it. If you've ever lost a fair allocation to a bot, or watched a "discovery" round clear at a number that had nothing to do with price discovery... Test the first Sealed Bid Auction on Solana:

A big pivot from Ken Griffin on AI: “Number one is, in the last few months, there has been a step change in the productivity of the AI toolkit. It is profoundly more powerful than it was just nine months ago. And for us at Citadel, that has allowed us to unleash a much broader array of use cases for AI. And it has been really interesting to watch, to be blunt, work that we would usually do with people with masters and PhDs in finance over the course of weeks or months being done by AI agents over the course of hours or days. These are not these are not mid-tier white collar jobs. These are like extraordinarily high skilled jobs being, I'm going to pick a word, automated by agentic AI. And I gotta tell you, I went home one Friday actually fairly depressed by this because you could just see how this was going to have such a dramatic impact on society. When you witness it in your own four walls, when you see work that used to be man years of work being done in days or weeks, it's like, wow, like that's the first time I've seen real impact in our four walls.” This echoes my own experience with agents and the conversations I am having with students, friends & clients. The toolkit has dramatically transformed and it feels like in finance, for the first time, AI is real.

Visa is quietly building the foundations for trillions of dollars in agentic activity I’ll explain: 1) Base is described in Visa’s own release as built for “agentic commerce”, that’s not accidental wording, Coinbase has been positioning Base as the chain where AI agents transact 2) Agents need machine-speed, machine-priced settlement: - traditional card rails settle in T+1 to T+2 with human-scale fees (I know Ramp/Tempo MPP are experimenting with “tokens” or cards for agents, but it’s early) - stablecoins on fast chains settle in seconds at fractions of a cent, this is the only viable rail for high-frequency agent-to-agent or agent-to-merchant payments 3) An agent doesn’t have a bank account, it has a wallet, Visa becoming the settlement layer across nine chains means an agent can pay any Visa-connected merchant from any supported chain without the merchant needing to care which chain or which stablecoin 4) Visa abstracts the chain, which is exactly what agents need, an agent shouldn’t have to negotiate “do you accept USDC on Base or USDP on Solana”, the settlement layer handles it 5) This pairs with the x402 / MPP (Machine payment protocol) momentum, the missing piece has been a trusted settlement counterparty that merchants already have integrations with, Visa is volunteering for that role 6) The $7B run rate today is almost entirely human-initiated B2B settlement, the agentic layer is the second curve, and it’s plausibly larger because agent transaction frequency dwarfs human frequency 7) Tempo’s inclusion fits here too, real-time liquidity routing is a requirement for agent commerce where you can’t pre-fund every chain There’s significant upside for those that build the global infrastructure of the new era of agentic finance It’s just a matter of establishing how to participate in the upside; understanding where the trillions of dollars in economic value will accrue

MIT just made every AI company's billion dollar bet look embarrassing. They solved AI memory. Not by building a bigger brain. By teaching it how to read. The paper dropped on December 31, 2025. Three MIT CSAIL researchers. One idea so obvious it hurts. And a result that makes five years of context window arms racing look like the wrong war entirely. Here is the problem nobody solved. Every AI model on the planet has a hard ceiling. A context window. The maximum amount of text it can hold in working memory at once. Cross that line and something ugly happens — something researchers have a clinical name for. Context rot. The more you pack into an AI's context, the worse it performs on everything already inside it. Facts blur. Information buried in the middle vanishes. The model does not become more capable as you feed it more. It becomes more confused. You give it your entire codebase and it forgets what it read three files ago. You hand it a 500-page legal document and it loses the clause from page 12 by the time it reaches page 400. So the industry built a workaround. RAG. Retrieval Augmented Generation. Chop the document into chunks. Store them in a database. Retrieve the relevant ones when needed. It was always a compromise dressed up as a solution. The retriever guesses which chunks matter before the AI has read anything. If it guesses wrong — and it does, constantly — the AI never sees the information it needed. The act of chunking destroys every relationship between distant paragraphs. The full picture gets shredded into fragments that the AI then tries to reassemble blindfolded. Two bad options. One broken industry. Three MIT researchers and a deadline of December 31st. Here is what they built. Stop putting the document in the AI's memory at all. That is the entire idea. That is the breakthrough. Store the document as a Python variable outside the AI's context window entirely. Tell the AI the variable exists and how big it is. Then get out of the way. When you ask a question, the AI does not try to remember anything. It behaves like a human expert dropped into a library with a computer. It writes code. It searches the document with regular expressions. It slices to the exact section it needs. It scans the structure. It navigates. It finds precisely what is relevant and pulls only that into its active window. Then it does something that makes this recursive. When the AI finds relevant material, it spawns smaller sub-AI instances to read and analyze those sections in parallel. Each one focused. Each one fast. Each one reporting back. The root AI synthesizes everything and produces an answer. No summarization. No deletion. No information loss. No decay. Every byte of the original document remains intact, accessible, and queryable for as long as you need it. Now here are the numbers. Standard frontier models on the hardest long-context reasoning benchmarks: scores near zero. Complete collapse. GPT-5 on a benchmark requiring it to track complex code history beyond 75,000 tokens — could not solve even 10% of problems. RLMs on the same benchmarks: solved them. Dramatically. Double-digit percentage gains over every alternative approach. Successfully handling inputs up to 10 million tokens — 100 times beyond a model's native context window. Cost per query: comparable to or cheaper than standard massive context calls. Read that again. One hundred times the context. Better answers. Same price. The timeline of the arms race makes this sting harder. GPT-3 in 2020: 4,000 tokens. GPT-4: 32,000. Claude 3: 200,000. Gemini: 1 million. Gemini 2: 2 million. Every generation, every company, billions of dollars spent, all betting on the same assumption. More context equals better performance. MIT just proved that assumption was wrong the entire time. Not slightly wrong. Fundamentally wrong. The entire premise of the last five years of context window research — that the solution to AI memory was a bigger window — was the wrong answer to the wrong question. The right question was never how much can you force an AI to hold in its head. It was whether you could teach an AI to know where to look. A human expert handed a 10,000-page archive does not read all 10,000 pages before answering your question. They navigate. They search. They find the relevant section, read it deeply, and synthesize the answer. RLMs are the first AI architecture that works the same way. The code is open source. On GitHub right now. Free. No license fees. No API costs. Drop it in as a replacement for your existing LLM API calls and your application does not even notice the difference — except that it suddenly works on inputs it used to fail on entirely. Prime Intellect — one of the leading AI research labs in the space — has already called RLMs a major research focus and described what comes next: teaching models to manage their own context through reinforcement learning, enabling agents to solve tasks spanning not hours, but weeks and months. The context window wars are over. MIT won them by walking away from the battlefield. Source: Zhang, Kraska, Khattab · MIT CSAIL · arXiv:2512.24601 Paper: GitHub:

Anthropic said Mythos was too dangerous to release. Then four random guys in a Discord gained access on day one by guessing the URL... This is pretty insane: → Group in a private Discord guessed the endpoint from Anthropic's naming conventions → They figured out the conventions from the leak in the Mercor breach three weeks ago → Used a contractor's legit eval credentials to walk in → Have been using it ever since to build simple websites The AI that finds zero-days in every operating system on earth was defeated by address bar autocomplete... big yikes

🚨 I don't think people realize how bad things are at @aave right now. All core markets are at 100% utilization, that includes $3 bil in USDT and $2 bil in USDC stuck! That means you CAN'T WITHDRAW your money! A long post on why and how we ended up here. When the rsETH exploit happened and AAVE incurred bad debt, whales like Justin Sun, MEXC exchange, and others immediately withdrew billions from AAVE. This instantly drained all available liquidity in key core markets like ETH, USDT, USDC and so on. Those first to withdraw got out, laggers got trapped. Initially, the ETH market hit 100% utilization, meaning you could not withdraw your ETH from AAVE. Worse, this also means the protocol can't process ETH liquidations should ETH price fall/crash. If you can't sell any ETH, you can't liquidate to cover debt obligations. That means the risk of more bad debt incurred by AAVE is increasing the longer its markets remain stuck. Nevertheless, users can still sell at a minor loss the aETHwETH tokens on Uniswap or similar aggregators. That exit door is the last one remaining for ETH depositors on AAVE. The same cannot be said by depositors of USDT and USDC. They are stuck. That's because AAVE lost over $6 billion in liquidity in the past 24h. As whales took out their money, USDT and USDC also hit 100% utilization. These markets are now also stuck with money locked. Panic is spreading and desperate times call for desperate measures. Some users decided to borrow against USDT/USDC and exit via other markets at a 10-25% loss (90-75% LTV). Basically you borrow GHO/DAI/USDe against your locked USDT/C. But as more liquidity leaves AAVE, more markets get to 100% utilization and get locked/stuck due to low liquidity. This is quickly cascading across all available markets. Luckily the crypto market was rather flat today so liquidation risks were marginal, but if things change there are billions in stablecoins and other assets locked on AAVE that can't process liquidations = more bad debt for AAVE. If users or related protocols that are stuck need access to their money to prevent liquidations or other critical function, they have a huge problem on their hands. Plus, nobody wants to deposit (or provide liquidity) in these markets now since your ETH, BTC, USDC/T could be stuck there for who know how long. As soon as any available liquidity is made available, it is instantly taken out by bots fighting to get out. As I wrote this I saw 250k in liquidity on USDC vanish in seconds. Then there is the bad debt question. There's over $200 mil in bad debt incurred by AAVE via rsETH that's like a hot potato. Nobody knows who will eventually pay this bill. If you didn't remove your assets from AAVE, you risk receiving at least part of that bill in some form. Not having access to your money is part of that risk too. Contagion is also extremely high. Many protocols and apps rely on AAVE for their earn mechanics. These protocols and their users are stuck too and may be forced to incur bad debt with no fault of their own. October 10th was a CEX driven crash, this is a DeFi risk mitigation failure of epic proportions. AAVE should have never onboarded rsETH as a collateral asset, at least not to the size of hundreds of millions that allowed the hacker to walk away (i.e. borrow) over $200M in ETH after posting fake collateral. Rumors on X are saying rsETH was onboarded by AAVE due to a conflict of interest (lobbying) by a given service provider. If true, this is a major failure of its governance structure (nothing new). The folks at @KelpDAO who manage rsETH also have a tough decision to make on who will actually pay for the $200M exploit. AAVE users? L2 rsETH users? Everyone affected gets a haircut to account for the loss? The AAVE team and its founder, Stani, have been quiet for over 20h since the exploit after initially announcing the rsETH market freeze. They have a pretty big problem on their hands since the whole protocol is at risk right now. Trust is already lost as AAVE is bleeding billions in TVL to the level of hitting 100% utilization on all core markets. Maybe some key actors in the space will step in to provide liquidity to stabilize the markets on AAVE before this gets even worse. I got lucky to get out of AAVE early when I first saw this. I also removed all assets from DeFi and will not touch any protocol in the next few weeks. Too much risk for a few percentage points in yield. If you found this informative, like, share, and follow @duonine