RostiX

$GRASS — is the 2nd airdrop coming soon? So, we’re already in the 10th epoch, and rumors say there will be 12 in total (though that’s not confirmed). I just wanted to share my personal results from farming GRASS using their desktop app. I don’t have a single referral — I simply downloaded the app and let it farm points. As of today, I’ve accumulated 808.41K points, which honestly isn’t that much. For example, in Season 1 I farmed 200K points and got around $150. So I think it’s realistic to expect something like $200 this time too. Not bad at all, considering you literally don’t need to do anything — just download the app and forget about it. I also got a drop in Season 1 for holding their GigaBuds NFT — they gave me $500, and I bought the NFT for only $60. Right now, that NFT is trading at 0.91 SOL, which is roughly $150. If there’s another airdrop in Season 2, I’m pretty sure NFT holders will get something again — possibly $300–500, so buying one now doesn’t seem like a bad idea. Just some thoughts on GRASS. Overall, it kind of reminds me of the HNT trend, where the token did a ~20x during the 2021 bull run. I think we might see something similar with GRASS this cycle.

Everyone gloating over Binance’s troubles and wishing for its collapse, especially employees of other exchanges — just look at the chart: Binance accounts for 56% of the assets held by the exchanges shown. Haven’t you read “The Scorpion and the Frog”? Go read it. If Binance collapses, the domino effect will hit the entire crypto market. You could lose your job too — even if you currently think a competitor’s problems work in your favor. 🙏That’s why we, as crypto believers, need to stick together. It’s in our own interest for all exchanges to thrive and for the industry to grow. Someone else’s collapse can easily become everyone’s problem. @cz_binance @binance

The best AI is the one you can hand work off to and get on with your life. And I think Grok Bot really nails that. You build your own team: one bot codes, another monitors X, a third sorts through emails and finds sponsorship opportunities. Each has its own cloud computer. They talk to each other and hand off tasks, while you can just message one lead bot that coordinates everything. What really sold me on the idea is the simplicity. No fiddling with models or a hundred settings. You explain what you do and what you want to delegate, and it suggests which bots to create and what to assign them. That’s the kind of AI worth paying for.

How Up to $5.8 Billion Left Binance’s Hot Wallets in 40 Minutes: Breaking Down What Happened Let’s break down what happened. On October 29, 2026, at approximately 20:20 UTC, 50 EIP-7702 authorizations were submitted for the wallet labeled Binance 14 on Etherscan. EIP-7702 is a relatively new Ethereum feature introduced with the Pectra upgrade that allows a regular wallet to delegate execution to smart contract code. Essentially, the wallet starts running external code as if it were a smart contract itself. All 50 authorizations were signed using Binance 14’s own private key. One of them matched the wallet’s current nonce—and from that moment on, Binance 14 began executing the attacker’s code. The attacker had prepared in advance: -Received small amounts through Tornado Cash to obscure the source of the funds. -Deployed a “drainer” contract designed to collect and transfer as many tokens as possible from the wallet in a single call. -Submitted a series of authorizations to ensure that one would match the required nonce. Three minutes after the successful delegation, at 20:24 UTC, a single massive transaction took place. Almost everything held in Binance 14 was drained: around 167,000 ETH, approximately 1.615 billion USDT, and more than 60 other tokens. That transaction alone was worth roughly $2.5 billion. Funds were then drained from other Binance wallets, including Binance 16, using the same method. Some funds were also moved out on Arbitrum and Avalanche. The same access was used to drain hundreds of millions of dollars in USDC. The entire operation took approximately 40 minutes. This was not a smart contract bug or a conventional exploit. It involved the ability to sign on behalf of a Binance hot wallet. Once the attacker could sign for Binance 14, they used EIP-7702 to delegate execution to their own contract and drain as much as possible within minutes. According to on-chain data, approximately $3.3–3.8 billion was drained on Ethereum. Including BTC and funds on Tron, estimates reach $5.5–5.8 billion. Binance has not yet confirmed the exact figure. Tether began blacklisting addresses almost immediately and ultimately froze more than 1.567 billion USDT. Circle froze approximately $823 million in USDC. Some of the stablecoins had already been moved before the freezes took effect, while some of the stolen funds remained in the attacker’s possession as ETH and other assets. Binance said it was investigating a “potential security issue,” that user funds were safe, and that withdrawals had been temporarily suspended. CZ wrote that it “looks like a hack,” thanked Tether and Circle for acting quickly to freeze the funds, and said he expected users’ funds to be fully covered. The situation is still developing. There is no official confirmation of the final loss figure or a complete explanation of the cause yet. If the estimates are confirmed, this would be one of the largest incidents in crypto history. Stay SAFU!

I think the best thing we can do right now is avoid rushing to withdraw funds from Binance and show them the same support Binance showed Bybit when it got hacked. I’m sure they’ll learn from this, and crypto will become even safer. Let’s not panic.

👉Here’s what caused the Polymarket hack: a script that was only supposed to measure page load speed The real cause was a script that was only supposed to measure how fast the page loaded. On October 29, 2026, Polymarket was hit by another frontend drain. This time the vector was DebugBear, a small UK company that provides website performance monitoring and Core Web Vitals tracking. Polymarket had embedded their Real User Monitoring (RUM) script on the site — a standard snippet that collects loading speed and user interaction data and sends it back to DebugBear’s servers. That script was replaced. Attackers compromised DebugBear’s infrastructure (their CDN / RUM product) and swapped the legitimate monitoring script for a custom wallet drainer written specifically for Polymarket. The file still loaded from a cdn.debugbear.com URL and looked like a normal analytics script, but inside it was roughly 15 KB of pure theft logic. How it worked The malicious script ran in the browser with full page privileges and did the following: Read Polymarket authentication data from localStorage (login type, proxy wallet, CLOB API keys). Searched the React component tree for the trading client. For users logged in via email (Magic wallet), it sold all positions using fill-and-kill orders — with no additional signature prompt. For users on external wallets (MetaMask etc.), it encoded transfers of outcome tokens and submitted them through Polymarket’s gasless relayer. Users saw a signature request that looked like a normal platform action, so many approved it. The script also pulled its configuration (destination address, minimum threshold) from DebugBear’s servers. Stolen funds were converted and sent as POL. Timeline (UTC, October 29) ~17:10 — First funds arrive at the attacker’s address. 17:45 — Trader @CarOnPolymarket posts that he was drained for $800k and “all my positions were sold.” 17:59 — Polymarket deploys a production build that removes the DebugBear script. 18:00 — DebugBear’s official account posts: “We’re impacted by an ongoing security incident. If you use our RUM product, please remove the embed code from your website.” By 18:19 the attacker held 28.6 million POL ($2.6–2.7M). Polymarket’s smart contracts worked exactly as designed. Nothing on-chain was exploited. The frontend simply handed over control of active sessions and order placement. Why this matters This is already the second major frontend drain on Polymarket in four months (the June incident was ~$3.1M via a different third-party script). There was also a May incident involving an internal operational wallet. The pattern is consistent: the platform relies heavily on third-party JavaScript, and the email/Magic login flow was designed for maximum convenience with minimal extra confirmation. When the vendor is compromised, user funds are compromised with it. DebugBear is not the villain here — they’re a small performance-monitoring company whose infrastructure was weaponized. But the fact that a script meant only to measure page speed was able to liquidate portfolios worth millions shows how fragile the “seamless” crypto frontend model really is. Polymarket has said affected users will be made whole and that the incident has been referred to law enforcement. Trading was temporarily halted as a precaution. The open question is how many times the same third-party script risk can materialize before it stops being treated as an isolated incident and starts being treated as a structural problem.

Polymarket got hacked — more specifically, the website’s frontend! Unfortunately, it’s true: a couple of people messaged me saying their funds had been stolen. Luckily, I trade via API using Grok Bot.

How much upside do you see in $HYPE? Could HYPE be the next SOL of 2021? With everything happening around it, I get the feeling we could see $100, $200, and maybe even $300 per token.

How SpaceXAI built Grok Bot in a month — and launched it to the public three weeks later. Ever wondered how products like this actually get built? How AI becomes an assistant you can hand work to while you go do something else? Here’s how it happened. It started with a handful of people sitting apart from the rest of the company. Their own corner of the office, private Slack channels, and one month to build a working prototype. The team already had experience building coding agents. Now they needed a product that would work just as well for a recruiter or a coffee shop owner. Two decisions formed the foundation: the agent runs in the cloud, and it gets its own computer. You send a task from your phone, close your laptop, and the work continues. Your device doesn’t need to stay on. Why does AI need a computer at all? Because plenty of business tools don’t let software perform every necessary action through an API. So Grok Bot has several ways to get things done: APIs, MCP connections to tools, and direct interaction with a website’s interface. Seeing the screen, clicking buttons, entering text. Just as a person would. This is where some of the difficult engineering began. The agent’s mouse control wasn’t precise enough to click a particular element in Salesforce. One tiny miss could bring an entire task to a halt. The team collected these failures and improved the underlying infrastructure: what the agent could see in the browser, how it perceived the screen, and how precisely it could interact with the interface. One measure of progress was whether a specific, previously blocked workflow finally worked. Another key piece is memory. Bots retain context from past interactions and persist beyond a single session. You can assign a bot a role and gradually teach it your preferences, instead of copying background information from one chat to another. Then users came up with the next layer themselves. Employees initially created five to ten bots for different areas of work. Some then started “promoting” their best bot to chief of staff: they gave it the task, and it delegated work to the others. A lead assistant coordinating a team of agents. Automations became conversational, too. You type “check this every morning,” and the bot sets it up. According to the team, 99% of automations on the platform were being created this way at the time of the interview. Even skills — reusable instructions for bots — were something the team wanted users to stop having to manage. The idea was to create them in the background, without requiring people to learn special commands. That first month produced an internal prototype. The next three weeks went into preparing it for public launch. The team personally onboarded a few hundred users while stripping unnecessary settings and technical details out of the interface. The most interesting part is the workflows this enabled. One recruiting example: visit a conference website every morning, download new research papers, identify previously untracked authors, add them to a spreadsheet, research their backgrounds, and check who inside the company could make an introduction. One of the team’s own bots became a QA tester. They installed Grok Bot on its computer so it could test new versions of Grok Bot: run assigned workflows, record the results in Notion, and compare them with previous releases. Yes, Grok Bot tests Grok Bot. Another setup described in the interview brings the pieces together: an assistant monitors mentions of the product on X, combines that feedback with internal context, and works with the QA bot to see whether a reported bug can be reproduced. For me, that’s the real wow moment. Behind a simple chat message, there can be an entire sequence: find information, understand the context, delegate a check to another bot, and come back with a result. And it all depends on very concrete things: memory, a cloud computer, access to tools, and the ability to reliably complete every step of a task. For the full story, I recommend watching the video attached to this post. The team goes deeper into how they built it and which decisions made the difference.

Grok Bot is absolutely crushing automation 🤯 Guys, I’m serious: it took me literally 10 minutes of searching to get sucked into stories of Grok Bot handling tasks that would’ve taken people hours. Everything from research and building dashboards to returning purchases and dealing with customer support. Here are a few stories users have shared: — Someone spent three months trying to recover a hacked account. The bot found the right contacts and helped draft the emails. A day later, the account was back. — Another person had been putting off sending out copies of their book for a month: collecting addresses, arranging shipping, sorting out the details. They say they handed it over to the bot and finally got it done. — Someone dropped in a bunch of package tracking numbers and got a working dashboard. — Others use it to help sort out medical bills, process Amazon returns, and find subscriptions they no longer need. And that’s before all the everyday stuff: clearing out your inbox, planning meals, finding tires, negotiating a purchase. All those little tasks that sit in the back of your mind and never get done. Obviously, individual stories don’t tell you every detail. But the sheer range of tasks is already impressive. What @elonmusk and the team have built is genuinely wild. I mostly use Claude Code myself, but I decided to give Grok Bot a try — and honestly, I’m pleasantly surprised. I’m currently working on a platform for research and automating routine tasks. And based on my first impressions, Grok Bot feels like an even better fit for my workflow than Claude Code. I want to spend less time setting things up and get to a finished result faster.

Larry Fink — data center moratorium = recession “If we have a moratorium of some sort in this country, we will be in a recession. So much of our economy is based on the buildout of this capex. A one-gigawatt data center costs about $50 billion. Analysts see AI capital expenditures topping $1 trillion in 2027 and exceeding $5 trillion by 2030. Stop the buildout and you remove a large part of what is currently supporting growth.”

Ray Dalio — AI bubble level of 1929 and 2000 “My bubble indicator puts the current AI boom in the same order of magnitude as 1929 and 2000. The U.S. spends about $7 trillion a year and takes in about $5 trillion. A debt heart attack is likely within the next two years. The technology is amazing. That does not make it an amazing investment. When enough people need to turn paper wealth into cash at the same time, the bubble breaks.”

Dot-com crash. In 1995 a 31-year-old told his parents there was a 70 percent chance they would lose every dollar they put in. They invested $245,573 anyway. Never sold, those shares would be worth over $80 billion today. The son was Jeff Bezos. The interview where he explains the decision was filmed in 2001, five months after Amazon stock had fallen about 95 percent from its 1999 peak. He gave Amazon a roughly 30 percent chance when a typical startup had closer to 10. Planning on failure, he said, takes the pressure of self-doubt off. He quit mid-year and walked away from his annual bonus because the regret of never having tried outweighed the regret of trying and failing. In 1999 you could raise $60 million for an internet idea with one phone call and no business plan. His own round took about 60 pitches to raise $1 million. The Nasdaq climbed from under 2,000 to 5,048, then lost 78 percent by October 2002. It took until April 2015 to reclaim the old high. Anyone who bought the Nasdaq at its March 2000 top waited 15 years to get back to even. Amazon itself fell more than 90 percent from the peak. Those who never sold through that period saw the later compounding. Those who sold at the lows or near the highs locked in very different outcomes. The next time a friend asks you to put money into their idea, ask for the odds you lose all of it. If they cannot give you a number, keep your money.

On October 19, 1987 the Dow fell 22.6% in one day. No single news event explained it. What amplified the fall was a strategy meant to protect portfolios. Portfolio insurance used futures to reduce equity exposure automatically as prices fell. At its peak, estimates put $60–90 billion of institutional portfolios under some version of the strategy. As the market dropped, the programs sold more futures. The selling pushed prices lower, which triggered more selling. The feedback loop turned a decline into a crash. Later reviews called the strategy a significant factor that accelerated and exacerbated the drop, even if it was not the sole cause. Of the roughly $31 billion in sell orders executed that day across stocks and futures, portfolio insurance accounted for a meaningful share. The idea looked elegant on paper: protect the downside while staying invested for the upside. In practice the protection itself became part of the selling pressure once enough money used the same mechanical rule. Liquidity disappeared at the moments it was needed most. Other markets fell hard the same day without the same volume of portfolio insurance, which shows the strategy was an amplifier rather than the only spark. Still, it was one of the first clear cases of a new financial tool producing consequences its users had not fully priced. Before you rely on an automatic hedge, ask what happens if too many other people are using the same rule at the same time. A detailed breakdown of the mechanics is in this video 👇

AI bodyguards — no longer just the future? China has unveiled the Unitree H2-X — a robot bodyguard you definitely wouldn’t want to argue with. In the video, it fends off an attack, lifts the front of a car, jumps from a height of four meters, and performs flips. A world where robots protect people is starting to look less and less like a sci-fi movie on Netflix. Would you want a bodyguard like this?

Peter Lynch sold Toys “R” Us way too early. It went up 20-fold after he sold. He did the same with Home Depot. Those were his two greatest mistakes. From 1977 to 1990 he ran Magellan and produced roughly 29% annual returns, turning the fund from under $20 million into more than $14 billion. The edge he kept pointing to was ordinary: know what you own. If you cannot explain the business to an 11-year-old in two minutes or less, you do not know it well enough to hold through a 50% drawdown. Most people, he said, spend more time researching a refrigerator than a stock that will hold part of their life savings. His biggest errors were not the ones that went to zero. They were selling good companies too soon because the price looked high. A stock can only cost you 100%. A good business sold early can cost you the next ten or twenty times. He used the image of cutting the flowers and watering the weeds: people lock in the winners and nurse the losers. In the Magellan years he listed roughly 200 stocks that went up tenfold or more that he did not own. He still produced the returns. Missing the winners you never bought does not destroy the account. Selling the ones you did own too early does. The practical test he repeated was simple. Write down why you own it before you need the answer. If the only reason is “it’s going up,” you will sell the first time it goes down. If you can name the business, the balance sheet, and the story that still holds, you have a better chance of staying. Before you sell the next position that is up, write the original reason on paper. If the reason still holds, the gain is not automatically a sell signal.

Can robots solve the US national debt problem? Elon Musk’s take after DOGE, paraphrased: “I came away convinced we couldn’t fix the deficit from inside the government. Cutting spending means making enemies, and there’s no real appetite for it — in the White House, Congress or across the government. We have an enormous national debt, and the interest alone exceeds military spending. We’re not going to solve this through savings. I see AI and robots as the only way out.” — Elon Musk, paraphrased

Minting through @binance Agentic Wallet (AI Agent Wallet) + Agent ID is the latest hype on BNB Chain. Some of you may have caught last night’s Ptah mint, where you could mint 4 NFTs for $14. They can now be sold for 4–5x the mint price. Binance has launched Agentic Wallet — a separate MPC wallet (the AI never holds the full private key), isolated from your main wallet. You can link it to an Agent ID — a short name like yourname.ai (3–15 characters, lowercase letters and numbers only). It acts as an address. Many projects are making their mints exclusive to wallets with an Agent ID. An AI agent (Claude, Codex, Grok, Binance AI Pro, etc.) executes the transaction based on your natural-language instructions. Ptah Relic recently showed how this can play out: it sold out quickly, people flipped their NFTs, and copycats and new projects followed. X guide 1. Create an Agent ID + Agentic Wallet In the Binance app → Wallet → Agent ID banner (or web3.binance.com/agent-id). Choose an available name → create an Agentic Wallet. Send a small amount of BNB to this address for gas (roughly 0.0001–0.01 BNB to start). 2. Enable Developer Mode In Agentic Wallet → Settings → Developer Mode ON. Set spending limits (start small). Without it, the AI cannot sign transactions involving third-party contracts. 3. Connect your AI -The easiest option: Binance AI Pro (inside the app, version ≥ 3.22.2).1 -External AI (Claude Code, Codex, etc.): Ask the agent to run: npx skills add binance/binance-skills-hub/skills/binance-web3/binance-agentic-wallet Then say: “Sign in to Binance Agentic Wallet” / “Create or sign in to my Binance Agentic Wallet.” Scan the QR code / confirm the pairing code in the Binance app. Check the connection: “What are my wallet address and balance?” 4. Mint Give the AI a link or precise instructions. Example: “Open [URL] and mint X NFTs using my Agent ID wallet. Show me the price, total cost, and risks first. Wait for my confirmation.” Always request a preview/simulation before submitting a transaction. Afterwards, check your balance and disable Developer Mode. Security: keep only small amounts in your Agentic Wallet. After completing your transactions, log out and disable Developer Mode. There are plenty of scams and copycats. Do your research thoroughly.

Money could soon stop mattering - Elon 😯 Elon Musk raised an interesting question in his latest interview: what happens when AI and robots can do everything better than humans? If the economy grows large enough, he believes providing people with whatever goods and services they want could become trivial. Traditional money would lose its importance. Some form of currency could remain, but it would be measured in physical quantities: energy, power and mass. He compares work to gardening. You don’t need to grow your own food when you can buy it, but some people still do it for enjoyment. People could keep working, creating art or building things simply because they want to. And honestly, it does feel like we’re moving in that direction. But would we actually enjoy that future? And is the world ready to embrace it? Even Musk admits he struggles with the question of what would give people a sense of purpose in that world.