對 AI 代理(agent)現實面的驚人洞察。行動迅速很酷,但真正的價值在於避免災難性的鏈上錯誤。💎
ŘeGáL TraÐér
·
--
牛頓的真正考驗是在資金移動前的那一秒。
從稍微不同的角度思考牛頓。不只是把它當作 AI 交易基礎設施,而是把它看成車離開車庫前的那個小小檢查點…… 在加密領域,我們通常會為「執行」而興奮:更快的交換、更快的代理、以及更快的結算。很酷,但速度不一定等於智慧。我從牛頓身上理解的是,它想要放慢某一個非常特定的時刻:結算前的那一秒,在系統問:「等一下,這個操作其實是允許的嗎🧐?」
圖表又開始做那套了——橫向盤整、大家刷新 Twitter 等著下一段敘事。原本我應該要監控幾個部位,但最後我卻掉進了這個與 AI 代理(agent)相關的兔子洞:為什麼一半談論「代理經濟」的人聽起來像是少了某個地基。 我甚至本來就沒打算要去看 Newton Protocol。有個人在群組聊天裡丟了個連結,還抱怨上週有個濫用的機器人把測試錢包洗空了。說真的,出於無聊而不是別的什麼,我就點了。
No more blind trust. 🛡️ User-defined rules and cross-chain reliability are the real game-changers here. Spot on!
Meerab BNB
·
--
#newt $NEWT
I’ve been looking into @NewtonProtocol over the past few days, and honestly, the most interesting part isn’t even the AI itself. It’s the permission system.
Everyone loves to talk about what AI agents can do. But I keep coming back to what they shouldn’t be allowed to do. That feels like the real conversation.
Normally, giving any software access to your wallet is a leap of faith. It’s this binary choice—either you hand over the keys and hope for the best, or you just don't bother. Newton seems to take a different route. Instead of all-or-nothing, you can actually set ground rules upfront. Spending caps, specific protocols, exact transaction types—you draw the box, and the agent stays inside it.
That reframes the whole thing. It stops being about blind trust and becomes about controlled, pre-approved execution.
I’m still curious about a few practical things, though—like how the community-vetted strategies are going to be audited and how visible the logic is before I'd feel okay parking serious money in there. That kind of transparency matters way more than any hype.
If this actually holds up in live markets and works across chains, it might just remove one of the biggest mental blocks people have about letting AI near their funds. The tech is cool, sure. But the real test for me is whether users can truly stay in the driver's seat while the automation runs in the background.
Would you personally be okay using an AI agent if you knew, without a doubt, that every move had to fit inside rules you set yourself?
這周市場一直在橫盤震盪,老實說挺無聊的。我開着 CT 的頁面放在一個標籤頁裏,最後卻開始翻一些亂七八糟的項目文檔——這通常就是在暗示我,我在逃避做別的事情。於是我就看到並落在(<c-25/>,#Newt ,$NEWT )上。 我之前就見過這個名字——"AI agents for DeFi",又是那種。差點就把標籤頁關了。不過我注意到他們描述裏的某些東西,愣是讓我停了一下。 大家都在把它吹成“現在 AI 會替你做 DeFi”。聽起來很酷,當然了,我們從一百個智能體項目裏聽過一百次這種說法。但這並不是 Newton 在賣的東西。他們賣的是與“AI 自由”相反的東西——他們賣的是 AI 限制。整個架構(TEEs 加上零知識證明,再加上這件叫“zkPermissions”的東西)存在的目的,就是確保這個智能體不能做任何你沒有明確允許它做的事情。比如價格限制、時間窗口,甚至它被允許接觸哪些協議。
今天下午市場安靜得有點不對勁。圖表呈現那種平飛(flatline)的狀態:看起來什麼都沒動,但你又還是每五分鐘就刷新一次,彷彿白癡一樣。我本來是要去檢查一些持倉,結果卻掉進了這個在 X 上的兔子洞,關於另一宗 DeFi 漏洞的爆料。老故事了——某個聰明合約被掏空,原因是奇怪的權限設置或是漏掉了風險檢查。又一次。我想說,我們一直在打造這些美麗的開放系統,然後卻又假裝驚訝風是從裂縫中穿過來的。