Summary of Bitget CEO’s Live Stream: Some hacker addresses have been frozen, and user funds will not be affected in any way.

Solana-based token launch platform StonkFun officially announced that 18% of STONK’s total supply has been burned. On September 24, the platform recorded $1,095,334 in revenue, with $672,362 allocated to buybacks, leading to the burning of 2.05 million STONK tokens.

CryptoQuant analyst Darkfost published a note stating that Bitcoin has recently entered an upward phase distinct from its previous rallies. Since hitting its July high, BTC has rallied roughly 45%, breaking through key structural levels and closing above $82,000 for several consecutive days. Darkfost noted that BTC’s accumulation trend has persisted during this rally, a pattern particularly evident in BTC’s capital flows on exchanges. Binance currently holds around 30% of BTC available on all exchanges accessible to all investor types, and net BTC outflows have dominated recently—averaging roughly 2,000 BTC per day over the past week. Meanwhile, Binance recently recorded its largest single-day BTC net outflow since 2023, exceeding 13,800 BTC in a day, marking a new three-year high. Over the past four days, Binance’s BTC reserves dropped from 705,000 BTC to 685,000 BTC, a reduction of about 20,000 BTC. Darkfost explained that sustained BTC outflows from exchanges signal some investors are choosing to hold BTC themselves, a behavior typically linked to long-term holding. From a market mechanism perspective, declining exchange BTC balances also reduce potential selling pressure. Darkfost further pointed out that such sudden large-scale outflows may indicate investors who previously expected Bitcoin to keep falling as it did during past bear markets are now experiencing FOMO and re-entering after the price rally. As Bitcoin continues to rise, similar large-scale capital flows may occur again.

A whale deposited 541,009 $HYPE($49.55M) to #Kraken 3 hours ago.

Beating AI News Brief: Stealth-mode startup Ando has closed a $20 million funding round to build a new Slack alternative designed for both humans and AI agents. The round was backed by Accel, Index Ventures, and Emergence. Ando retains core features like channels, direct messages, group chats, and voice calls, but AI agents have their own identities, permissions, and inboxes. Agents can independently browse and join channels, eliminating the need for humans to tag them (@) to trigger actions, and can proactively reach out to other team members. It supports integration with various AI agents including Codex, Claude, Grokbot, and Devin; teams without existing agents can also use Ando’s hosted version directly. Ando itself has fully stopped using Slack. For example, when a user reports a bug in a chat, agents can independently take the task, assign mobile-related work to another agent, and have a third agent review the code. If the same issue is discussed across multiple channels, agents can proactively pull together relevant personnel and fill in context gaps. Currently, teams across industries including software, real estate, and finance in 15 countries are using Ando.

Insight Beating AI Flash News: Agent collaboration platform Raft, founded by Richard Qian, former author of Kimi CLI, has released its source code, but is temporarily not accepting external Pull Requests and has not made its internal commit history public. The public GitHub repository is only a release mirror; actual development remains in a private repo, with a full code snapshot synced each version release. The Raft team argues that coding agents have made writing code far too cheap. Generating a seemingly reasonable patch is easy, even generating ten at once is not difficult, but maintainers still spend significant time understanding, verifying, and reviewing this code. In many cases, after reviewing code generated by others using agents, maintainers would rather have their own agents reimplement the code per requirements. Thus, the team aims to explore a "Prompt Request" model: contributors will not necessarily submit code directly, but first outline the problem to solve, its value, and constraints and tradeoffs, before maintainers and agents decide on implementation. This mechanism is not yet officially open; Raft’s GitHub Pull Requests and Issues are currently closed. Raft also challenges the value of traditional commit history. The team believes that with agents in development, rather than tracking every code change, "decision history"—including why changes were made, what approaches were tested, and issues identified by agents—may be more useful. Notably, Raft is source-available, not traditional open source. The FSL-1.1-ALv2 license permits users to view, run, and modify the code, but prohibits using it to offer competing commercial products or services. Each version will automatically convert to Apache 2.0 two years after release.

Poker pro and crypto investor Hu Zhewen made heavy bets in three crypto booms, including algorithmic stablecoins and AI tokens, but suffered a string of losses. Between May 2021 and May 2022, he poured around $80 million into algorithmic stablecoin TerraUSD (UST) and its sister token LUNA. At its peak, his holdings were worth over $800 million, but LUNA’s price later plummeted to near zero. Hu claimed he once invested in blockchain project Stratis, earning a return of “over 1,000 times”. Hu has filed a lawsuit in Chicago, U.S., against global quantitative trading giant Jump Trading, its crypto-related affiliates and executives, seeking at least $500 million in damages. Separately, Hu bet on a new project by Chinese computer professor Zhang Yongfeng last January. Zhang joined Tsinghua University’s Department of Computer Science in 2007 and currently serves as a tenured associate professor in the Department of Computer Science at Rutgers University in the U.S. He has been sued by Hu on multiple charges including “securities fraud”. The token issued by Zhang’s institution, in which Hu invested millions of dollars, has fallen more than 99.6% from its 2025 peak, according to reports. Public records show Hu is a world-class Texas Hold’em poker player: the World Series of Poker (WSOP) website currently lists 58 money finishes, 5 final table appearances and total tournament earnings of around $1.67 million for him. Four years ago, in the $10,000 buy-in SuperMILLION$ high-stakes tournament, Hu reached heads-up against a New Zealand professional player, finishing second and taking home nearly $760,000 — the largest single-event prize of his documented poker career. (Caixin)