I don't predict price. I watch the people who move it. Fear, greed, attention, belonging, narrative. Same crowd. Different incentives. When the thesis breaks, I get out. I tell others to do the same. I still miss it sometimes. That’s the lens.

我以前挑聪明钱，最容易犯的错就是先看收益榜第一名。 数字只要够夸张，脑子里马上开始算，如果我早一个月跟上，现在是不是已经翻倍了？后来真正跟过几次才发现，收益最高的地址，不一定是最适合自己的地址。 有的交易员胜率很好，但习惯扛单；有的盈亏比漂亮，可一周只出手一两次；还有的专做高波动小币，赚钱时很猛，回撤时也完全不给反应时间。跟单最难的其实不是按下按钮，而是遇到连续亏损时，你还知不知道自己为什么跟、能不能按照原来的仓位继续执行。 最近我用CoinW聪明钱，已经不再只盯着总收益，而是会一起看胜率、盈亏比、最大回撤、交易频率和持仓偏好，再决定分多少本金。入口放这里： coinw.com/zh_CN/copy-trading… 我自己的做法是先小额观察，看看地址在震荡、急涨急跌和连续亏损时怎么处理。确认风格能接受，再慢慢增加跟随资金；如果只是看到某一笔暴赚就冲进去，往往最容易接在策略不顺的阶段。 我现在更把聪明钱当成一套观察真实交易行为的工具，而不是稳赚按钮。它能替我省掉找地址、盯链上和同步下单的时间，但仓位和风险还是要自己负责。选对一个适合自己的交易员，比追逐每天榜单上的第一名重要得多。

同一份Meta财报，我竟然看到两个完全不一样的EPS。 MarketBeat标的是7.40，还挂着超预期。可Meta官方的GAAP摊薄EPS，只有3.83。要是我先瞄到7.40，肯定会想这季报不错啊。 后来一对，差就差在一笔98.9亿美元的法律和解费。3.83是正式GAAP每股收益，这笔费用算进去了；7.40是公司剔掉它之后的假设测算。两个数各有各的口径，不能混着看。 最近在okx的TradFi上看xMETA这类美股代币化标的，价格变动比较直接，成交和结算都在平台内完成。这种标的我特别容易被页面最显眼的数字带着跑。先看到超预期就想着找利好，先看到3.83又急着挑毛病。 所以现在在OKX上看Actual EPS时，我会先往下翻一翻，看看哪些费用算进去了、哪些被剔除。同一份财报数字都摆在那儿，但先看到哪个，我的第一反应可能就不一样。 @okxchinese #OKX达人 #交易TradFi就来OKX #META #财报季

GM GN CT ☕️ Pharos flagged: “USG Burns $36.85M In A Day.” That Pharos alert could easily make me think a huge chunk of Tangent's USG had just disappeared from circulation. The word “burn” makes the supply story feel pretty obvious. Then @Tangent_fi clarified that the burn came from its Peg Keeper reserves. According to the team, that USG can be minted again when needed, and circulating supply wasn't affected. Tangent's Peg Keepers add or withdraw USG from liquidity pools to help maintain the peg. I can see myself remembering the $36.85M headline long before remembering the explanation. Next time I see a burn alert, I'd check one thing first where were those tokens sitting before they were burned?

An AI trading agent places no order. My first instinct would probably be guess it didn't see a good opportunity. But what if that's not why it stopped? @NeoSoulAI NeoTrade lets users define strategy goals, capital limits, market permissions and risk boundaries. An agent might decide to pass on a trade. A proposed action might hit a risk or permission limit and get blocked or held for approval. Or it might simply be waiting for human confirmation. Same outcome on the surface. Different reasons for no trade, and sometimes those steps overlap. And I'd react differently to each. A strategy pass makes me question the signal. A risk block makes me check which rule was triggered. Pending approval means there's a decision waiting for me. Without that distinction, I could end up changing a perfectly reasonable strategy when the issue was somewhere else. One thing I'd love NeoTrade to make easier over the next 12 months a clear reason attached to each actual strategy pass or unexecuted proposed trade, with any relevant rule or approval state and the final outcome visible in the same decision record. I'd want that record clearly marked as Paper or live, too. Because if an agent doesn't trade, I still want to know who made that call. #NeoSoul #NeoTrade

第一次在 @Roobet 点开 Minted Mike，跟前阵子玩的 Towers 完全不是一个节奏。 先留这段，换口味还挺新鲜 😂 roobet.com/?ref=all3in4out

GM GN CT ☕️ Slippage tolerance is the number I'd be tempted to raise when I just want a trade to go through. In my head, that's an execution setting. End of story. Except @protocol_fx has another rule in its fee docs: unused slippage is retained. For BTC and ETH xPOSITIONs, it's split 50/50 between the Stability Pool and protocol revenue. For BTC and ETH sPOSITIONs, the docs allocate 100% to protocol revenue. That doesn't mean setting a higher tolerance automatically costs me that percentage. The number is a limit, not a receipt. But knowing unused slippage has a documented destination changes how I'd look at the setting. I'd still want the trade to execute. I'd just check how much execution-price slippage I'm willing to accept, and what happens to the unused slippage, before increasing that number.

GM GN CT ☕️ Tangent FIR has a line that's pretty easy to like: “No term, no maturity.” A fixed borrowing rate, with no deadline reminding me to check the loan. Sounds convenient. Maybe a little too convenient. That's the part I'd worry about forgetting. Interest can still accumulate while the USG debt stays open, and the collateral still needs watching. Without a maturity date, it's easy to imagine putting off that check for another week. And then another. If I used FIR, I'd make Sunday my weekly check-in: outstanding debt, collateral health and liquidation buffer. Tangent gives borrowers flexibility over when to repay, which I like. But if nothing is due on a specific date, what's going to make me look at the position again? @Tangent_fi #Tangent

One final error. My brain would probably read that as the AI tried once and failed. But that's not always the full story. When developers run AI workloads on @Acurast smartphone Processor network, they can set restartPolicy to onFailure. The docs say it can retry up to three times after failure, and that only the third failure is reported. That last detail caught my attention. If I only saw the reported failure, I'd probably forget to ask what happened before it. One error report tells me something went wrong, but it doesn't tell me how many attempts led there. And that could change what I'd do next. Rerun the AI agent? Check the deployment? See whether it retried before failing? Before making that call, I'd want to know how much of the execution history I'm actually seeing. #Acurast hub.acurast.com/rebellion?re… docs.acurast.com/developers/…

GM GN CT ☕️ Gearbox Earn shows two RWA opportunities involving Frax's frxUSD: frxUSD Pool is labeled “Lend,” while mF-ONE / frxUSD is labeled “Leverage.” Same familiar ticker, two very different positions. I can see myself grouping them together just because both say frxUSD. Gearbox already labels the position types, but a familiar asset name can make the choice feel simpler than it is. With frxUSD Pool, you're supplying frxUSD to a lending pool. With mF-ONE / frxUSD, borrowed frxUSD finances leveraged RWA exposure, with debt, collateral and a health factor to monitor. The asset name is familiar, but what you're taking on is quite different. The funny part is that Gearbox puts the distinction right there in the labels. I'd just need to notice it before comparing anything else. Does seeing frxUSD in both rows make these positions feel more similar than they actually are? @fraxfinance

Block keeps the risk visible. Allow is the verdict I’d be more careful with. @Acurast Laya Prompt Injection Guard makes that distinction interesting. A System One model running on an Android phone in the Acurast Network checks a prompt, chooses Allow or Block, and identifies what kind of attack it looks like. There’s no real assistant behind this demo. You only see the verdict. I can see myself reading Allow as okay, this passed the security check. And then quietly turning that into this is safe. But Acurast openly acknowledges that tricky prompts can still fool Laya. They describe it as a fast first filter, not the final word. That’s the part I’d have to watch in my own thinking. A clean verdict makes it tempting to stop questioning what passed through. If the guard says Allow, am I treating it as one check completed, or permission to stop checking? #Acurast hub.acurast.com/rebellion?re… acurast.com/blog/network-and…

GM GN CT ☕️ Selecting 4× leverage feels like you've already decided what position you're getting. But what if the simulation comes back with a different number? Gearbox's SDK explicitly separates the leverage you request from the leverage it projects after fees and price impact. The simulation also estimates total debt, equity, borrowing cost and health factor. I can see myself getting attached to the 4× I chose, then treating the preview as a formality. The 4× is the number I chose. The simulated debt and health factor are the numbers I'd have to live with. The SDK provides those projected numbers. Whether I actually pay attention to them is another matter. Before confirming, am I looking at the position Gearbox simulated, or am I still thinking about the 4× I originally asked for? @GearboxProtocol

GM GN CT ☕️ UMX这张海报挺会抓眼球的，iPhone Duo一放上去，谁还不先看手机呢，哈哈哈哈！ @UMX_CN 计划于10月16日 20:00（UTC+8）上线。 这次指定USDT定期理财，期限90–360天，APR 4%–6%，前三档达标即可锁定赠机资格，4%那档还要满足90天日均持有要求，锁定资格后，耐心等待官方采购发货。 不过我对UMX还有个兴趣点，BTC、ETH不用先卖，也能通过抵押获得购买真实美股、ETF的资金或购买力。 对习惯长期拿币的人来说，多了一个跨市场选择。当然，借贷成本、抵押率和强平风险也要一起看，别只看购买力增加了多少。 手机确实吸引人，但我更想看看UMX怎么把加密资产和美股账户打通，能让资金多一种用法，这个方向挺有意思。 #UMX #UMXNOW

Towers 到这里，我终于舍得收了 😭 这张图先让我看会儿。 @Roobet roobet.com/?ref=all3in4out

GM GN CT ☕️ 335K+ smartphones onboarded on @Acurast . My brain immediately wants to picture one giant AI computer. Nice thought. But those are still individual phones, each with its own hardware limits. Acurast already supports LLM inference on its decentralized phone network. That's real compute being made available for AI workloads. But a single phone has limited RAM. Having 335K+ phones onboarded doesn't suddenly mean the largest AI models can fit on one device. That's why Acurast's Cray roadmap caught my attention. The plan is to connect hundreds of phones into compute clusters designed to handle much larger models. Sounds promising, but it's still a planned capability. I wouldn't count it as live today. I can see how a number like 335K+ makes the whole network feel more powerful. And scale does matter. But before assuming it can run any AI model, I'd want to know what runs on one phone today and what still needs that future cluster. 335K+ phones is impressive. What can they run today, and what will Cray need to prove? That's what I'll be watching. #Acurast hub.acurast.com/rebellion?re…

The Snake demo makes it easy to watch the AI make a move. But which phone actually made that call? That’s what caught my attention about @Acurast Laya. Its System One model takes a situation, picks from fixed options, and runs that decision on an Android phone’s CPU in the Acurast network. Snake is one example. Sorting emails into inbox, spam or phishing is another. The demos also show the phone’s country, chip information when available, decisions served, and a “verify on-chain” link. That gives me something concrete to inspect behind the words “AI powered by phones.” Of course, seeing where a decision ran doesn’t tell me whether the model made the right choice. And a verification link alone isn’t a complete audit of every request. Still, I like having somewhere to start. A country and chip are useful clues. How much of the actual execution does that on-chain link let me verify? #Acurast hub.acurast.com/rebellion?re…

GM GN CT ☕️ 6.5% fixed borrowing. 17.8% fixed PT yield. My brain immediately wants to do the subtraction and call the result “fixed” too. That’s what caught my attention in @Tangent_fi Oct 12 BTC carry thread. The example combines borrowing USG against cbBTC/WBTC at a fixed 6.5% with buying PT sUSG on Pendle, quoted at 17.8% fixed APY at the time. Two clean numbers make the outcome feel almost settled before I've even looked at my LTV. But Tangent says net APY depends on LTV. Most of the net APY is fixed, while part of it still moves with the collateral yield. I can see myself remembering the two headline rates and quietly treating the whole strategy as one predictable number. That's where I'd want a second look. What LTV am I actually using? How much of my net APY can still change? If the collateral yield moved tomorrow, would I still call the whole trade fixed?

GM GN CT ☕️ A Net APY can make a decision feel almost finished before you've looked at what's producing it. Take the Frax USDai Ecosystem Vault on Morpho. You deposit frxUSD, and the headline gives you one return figure. Open the APY Breakdown and you see Vault Net APY, WFRAX incentives and the fee structure. Suddenly, there's more to think about. The vault, curated by Re7 Labs, allocates frxUSD into USDai ecosystem lending opportunities, including a market backed by sUSDai collateral. Those lending markets carry collateral and liquidation risks, withdrawals depend on available liquidity, and token incentives can change. The Allocation and Risk tabs are there too. The information is available if you take the time to look. I can see myself forming an opinion from the APY first, then reading the breakdown mostly to confirm it. That's the habit I'd want to catch. If the token incentives disappeared tomorrow, would I still feel the same way about depositing? @fraxfinance

A company could beat earnings expectations, see its stock rally, and still leave a KPI binary call settling at $0. That’s what caught my attention about Robinhood’s planned Earnings Contracts. Under Cboe’s proposed rules, a call pays $1 if the specified financial metric meets or exceeds the contract’s strike, and $0 otherwise. Settlement uses the relevant figure from an earnings-related SEC filing. The product is still subject to regulatory approval. Imagine a company beats Wall Street’s estimates and its stock jumps, but the reported KPI specified in the contract falls below its strike. The call still pays $0. I can see how easily my brain would connect those three things before checking the actual condition. That's what would make me hesitate before clicking. Am I predicting a good earnings report, a positive stock reaction, or one exact number in an SEC filing? Those are three different bets.

GM GN CT ☕️ On @protocol_fx fxSAVE deposit screen, two controls sit only a few lines apart Unlimited Approve and Important Notice. They look like part of the same setup flow, but they ask you to process two different things. Unlimited Approve is about token authorization. Important Notice is a separate check tied to the conditions around the deposit itself. That’s where I can feel my attention wanting to collapse them into one routine: approve, acknowledge, continue. My hand can move faster than my attention. The useful pause is separating what am I authorizing? from what conditions could affect this deposit? Before depositing, do I actually understand both—or did I just move through two controls on the way to Approve & Deposit?