Tols (@0xTols) header
Tols (@0xTols) profile picture

只想賭博,然後守住,然後保持好心情

Joined April 2019
Photo by Tols (@0xTols)Photo by Tols (@0xTols)Photo by Tols (@0xTols)Photo by Tols (@0xTols)Photo by Tols (@0xTols)Photo by Tols (@0xTols)Photo by Tols (@0xTols)Photo by Tols (@0xTols)Photo by Tols (@0xTols)Photo by Tols (@0xTols)
神鱼 @bitfish 的这个认知太顶了,真正的“窃天之智” 核心观点: 1/ 算力决定你看到的世界:从原子 → 比特 → 向量 2/ 用 Token 购买「时间折叠权」 3/ AI 时代更需要「健脑房」+ 双轨阅读 4/ 向内觉察:Agent 成为最懂你的镜子 5/ 向内觉察:Agent 成为最懂你的镜子 6/ 人机协同的控制论:两个核心旋钮 7/ 超级个体放进传统组织会失效 8/ 人生本质上只有三件事:吃饭、睡觉、合道(运用算力对外探索世界结构、对内觉察灵魂) 结尾金句:在向量世界里,人人都在“窃天之智”
57
245
27
869
103,647
Replying to @BillyM2k
3,754
3,310
1,785
60,527
4,303,475
A record-breaking $6.3B breach demands a complete paradigm shift in cybersecurity architecture. 🚨 In our newest insight, GCSA examines the critical vector from Binance-related DevOps vulnerabilities to cross-border attribution. Dive into the report: media.gcsa.com/en-US/news-37… #GCSAInsights #CryptoSecurity #Binance
10
8
18
29
69,488
The attacker didn't know Binance 14's live nonce. So they sprayed 50 EIP-7702 auths (18914775-18914824) in one type-4 tx. Only 1 of 50 was a hit. A 7702 auth is valid only for the current nonce at inclusion. The hot wallet was still firing. One nonce in that window landed.
This EOA was 7702-upgraded to an unverified sweeper contract. Impl: 0x63e842b054bd624a465454d533c33f573e9c6fae Only hardcoded operator 0x7c8601d4645b93cea45f71e118a729332e6b5ca0 can drain it. sweep() pulls every ERC-20 + ETH (leaves 1 ETH for gas). Incoming funds stay stealable.
80
159
53
961
424,155
Tols retweeted
感觉Robinhood Chain Infra有望进入下一阶段 最近看到俩DEX,一个是 @metricxyz , 一个是 @notesystems 俩人共同点:用外部预言机真实价格定价,不是Uni那种 x*y=k的传统AMM 忽然让我想起DeFi Summer 发明PMM的 @BreederDodo 没经历过Defi Summer的老铁可能对这个东西有点陌生,先简单介绍下PMM - Proactive Market Maker PMM - 通过引入预言机(Oracle)获取外部市场参考价格,将大量的流动性主动聚集在当前市场价格附近。当池子里的某种资产被买走减少时,PMM 会自动调高该资产的报价,从而吸引套利者把缺失的资产套利补回。这种设计可以显著提高资金利用率、降低交易滑点,并减少无常损失风险 当年Dodo完败给Uniswap,原因应该有两个: 一是主流币的价格在 Binance等CEX,当年预言机慢于 CEX,套利者专挑这个窗口吃LP 二是DeFi Summer 的主战场是长尾币,很多没有CEX也没有预言机,只能靠AMM发现价格。PMM天然吃不到这块 但是 - 现在RWA 的情况不一样啊! 先说上一个版本 - 咱们发明了Meme配对Stock,还搞出了AMC之类的爆款,有过“让链上交易发过来影响显示世界”这么一个梦想 你看这个月一凉,是不是这梦想直接没人再提了? 因为本来就是个梦想,即便真的做成了,也只能影响几个无人问津大家听都没听过的小票股,7姐妹这种根本不可能 下一个版本可能是什么?AAPL的价格本来就在纳斯达克发现,链上DEX本来就是price taker。让池子自己去发现股票价格,属于第一性原理上的错误。链上AMM的价格只是噪音加套利成本。 DODO 那个设计放在crypto资产上当时有点错配,现在放在RWA 上,忽然感觉就对味了 最后说下两个Dex各自的特点,很多设计跟Dodo还不太一样 1. Metric - 流动性跟着价格跑 把"定价"和"流动性"拆开:报价交给预言机(Chainlink、Pyth),池子只负责放流动性 LP 的流动性不是挂在某个固定价格区间,而是挂在"偏离参考价 x%"的位置。参考价一动,整片流动性跟着平移,永远不会出区间。用过 Uni V3 或 Meteora的老铁应该懂,再也不用天天手动 ebalance 了 Robinhood 股票代币分红改乘数这种细节,都单独接了预言机来处理,否则每次分红、拆股后价格都会漂 一句话:DODO 是把流动性"聚"在预言机价格附近,Metric 是让流动性"跟着"预言机价格走。 2. NoteSystem的Swap - 价格旧了、休市了怎么办 引擎叫 S-DAMM,个人感觉有点像DODO PMM的完全体: 参考价 + 库存调整 + 单边 LP,这三样 DODO 都有 多了一个时间维度:价格每过约 5 秒没更新,置信度就减半,报价自动变宽、深度变小。DODO 当年就是不知道自己的价格有多旧,才被套利者吃掉的 专门为美股休市设计:收盘前 60 分钟撤回常规库存;夜间和周末只用专门的 Continuous 库存,报价更宽;完全没有新价格时进入 Discovery 模式;开盘先按真实价格重新估值,再接新单 LP 分层:Senior USDG vault,加上一个先承担亏损的 junior Backstop。DODO 时代的 LP 是裸奔 一句话:DODO 只知道价格是多少,Note还知道这个价格有多旧,美股开没开门。 Metric已经主网跑起来了,数据还不错,不少老外KOL在Farm。Note这个应该还是测试网阶段,不过可以重点关注下(话说之前重点关注的StonkBroker的期权还没出来,感觉要凉凉……)
25
19
1
83
13,863
Tols retweeted
最近我一直在研究 local model,也在想要不要放一台機器在公司當備援。 原因很簡單,自己現在寫 code、做產品,真的太依賴 AI 了。如果哪天訂閱或 API 突然不能用,很多工作可能真的會直接卡住⋯ 所以我現在看 AI,會開始多想一層:模型背後的算力從哪裡來?工作可以放在哪裡跑?有沒有其他部署選擇? 自己做 TermHive 這種多 agent 工具,還有之前做素材生成工具時,我都有在處理不同工具、服務怎麼接起來的問題。像素材生成流程裡,就加了多個服務的路由跟 CLI fallback,讓流程有其他路可以走。 做這些東西的時候,就會開始在意整個流程能不能持續跑下去,以及每一步到底需要多少模型能力。 像讀懂一個 repo、處理複雜的程式問題,跟「把這筆資料分到哪一類」「找出相關文件」「決定接下來走哪個流程」,需要的能力其實差很多,所以最近小模型、裝置端 AI 的進展,我覺得滿值得看。 Google 10/6 推出的 EmbeddingGemma 2,完整模型是 740M 參數,可以把文字、圖片、音訊等內容轉成向量,做語意搜尋與檢索。只處理文字時,用 270M 的模組就能上場,定位就是讓這些工作能在裝置端完成。 Liquid AI 10/7 也釋出了 d1-3B 跟實驗版 d1-omni-600M,走的是 decision model:用一次 forward pass 輸出選擇,專門處理分類、意圖判斷這類工作。 看到這些進展,我會想到的是:一個 agent 工作流程裡,有多少步驟可以拆出來,交給更輕量的模型處理? 如果模型能把一件範圍明確的事情做好,那部署它需要的設備、成本和方式,就有機會跟著改變。 這也是我看 @Acurast 時,覺得有意思的地方。 它把分散在各地的手機接成運算網路,開發者提交任務,再由 Matcher 媒合任務需求跟設備資源,實際運算交給手機節點執行。 但要把工作交給別人的手機,光是有算力還不夠,設備的狀態也需要有辦法確認。 Acurast 會透過 Android Key Attestation、iPhone App Attest 這類硬體認證機制,檢查設備與 App 的狀態,也會持續排除認證被撤銷的節點。 我覺得任務調度、設備驗證、部署工具這些東西,才是手機算力能不能變成一項服務的關鍵。 它的開發文件也已經有 GGUF 模型的部署範例,示範在手機節點啟動 LLM 服務,提供相容 OpenAI 風格的 API。站在自己有在做 agent 工具的角度,這就給了一個可以研究怎麼接進現有流程的起點。 如果是我來挑第一批測試任務,我會先找能獨立執行、資料傳輸量不大、對即時性要求也沒那麼高的工作。 例如公開資料分類、批次文字處理、定期產生簡單摘要。先確認模型能把這些事做好,再去量從送出任務到拿到結果的時間,以及整體成本。 因為站在開發者的角度,最後一定會問:部署麻不麻煩?節點不在線時怎麼處理?失敗重試之後,時間和費用還划不划算? 這些問題能解決,我才會考慮把更多工作放上去。 Acurast 主網今年 1 月已經上線,目前官方公布的累計接入規模超過 33 萬支手機。ACU 則參與部署費用、算力獎勵、質押跟治理,是這套網路運作的一部分。 而我接下來會關注的,就是這些設備能承接哪些實際工作,以及開發者用了之後,會不會繼續用。 我的判斷是,隨著更多小模型能把特定任務做好,手機這類設備可以參與的 AI 工作會變多。對 Acurast 來說,把這些能力整理成方便部署、穩定執行的服務,就是值得看的機會。 以前想到抽屜裡的舊手機,大概就是備用機。 現在至少多了一個值得研究的方向:它能不能成為某個 AI 工作流程裡,持續做事的一個節點。 有在做 AI、agent,或關注 DePIN 的,可以看看 Acurast 跟 ACU  Always DYOR。
🤝 Paid partnership
17
4
42
4,839
I am not going away @vladtenev @DanGallagher AMC Entertainment does not consent to Robinhood creating Stock Tokens to replicate AMC shares in the United States (the SEC requires you to get our consent, or failing that, at least with us to stand down). They are fake, synthetic shares, which could impact our capital structure. AMC should control our financial destiny, not Robinhood.
1,132
1,626
347
5,596
449,600
Still hodl robinhood:0x385f4f8ae47651ce5f58f5265395a669f8281e18
Hello again @vladtenev and @DanGallagher. So what’s Robinhood’s answer? You claim to “support” the SEC’s new rules on Stock Tokenization efforts, as they apply in the United States. Internationally, are you going to respect those same principals as laid out be the SEC? Or shame, shame, shame, will your off-shore operation in far-off Jersey be a sleazy bastardization that makes a total mockery of what the SEC has demanded as being the right thing to do?
1
449
0xb1da35ddb716a2f66c612d15e6d3ce5b90afb447 菸狗檔一根
Hearing more and more AI researchers are taking up smoking … superintelligence is near
1
4
1,457
I don't recommend anyone scramble to move their funds to new wallets today. But we should take the risks to cryptography from AI-accelerated math seriously, and minimize our exposure to not just quantum-vulnerable cryptography, but also potentially AI-vulnerable cryptography. The core new area of risk from this viewpoint is, unfortunately, ML-DSA / FHE / lattices. (and it's also another reason, along with quantum, why ECDSA might fall even faster than expected, hence the "fresh address" recommendation) So far most people have been in the mode of thinking "elliptic curves broken, hashes safe, lattices safe". But there is a good chance that the concrete security of lattices will take serious hits from the next two years of AI math. The basic threat model is: factoring is something that naively takes 2^(n/2) time, but over decades smart people have found and optimized number field sieves, and degraded that to 2^O(n^(1/3)), which is why RSA keys and signatures need to be ~400 bytes (and not 64 bytes). What if there are skeletons in the closet like that, both for elliptic curves and lattices, that we are simply not smart enough to discover - but bots soon will be? This is a major part of the reason why for the past year ethereum's lean roadmap has been going in the "hash-only" direction: no lattices, no ML-DSA, no Falcon, no lattice-based commitments inside ZK proofs, etc. Signatures in lean ethereum are all hash-based, either WOTS or SPHINCS-. For signatures and proofs, we already know how to go hash-only. The bigger challenge is for *public-key encryption* - and this goes far beyond blockchains. Secure communication, anonymizing protocols, lots of things need public-key encryption. And unfortunately there are long-standing mathematical theorems showing why public-key encryption cannot be done with hashes alone. You have to have some kind of trapdoor object that has at least one form of usable "structure" - either group theory (incl. isogenies) or lattices or code-based or potentially in the future even more newfangled and spooky things (local mixing?). But for anything that has structure, you should assume that AI will make at least some progress in breaking that structure. Here, one reasonable inference is that if you want to make something plausibly long-term secure, multiply the key sizes by 10. To me that's a very plausible world and something not at all extreme to predict. If AI will bring us 50 years of math in 2 years, then that 50 years of math may very plausibly include a "naive factoring -> GNFS" level of improvement to our ability to break lattices. In that world, lattices will still exist, but they will have to be significantly bigger to guarantee the same level of safety. And at those new larger sizes, hash-based constructions will beat lattice-based constructions on concrete efficiency in every use case where hash-based constructions are possible at all. Theoretically, of course it's possible that hashes are broken too (eg. P = NP would imply that). But I think P = NP is very unlikely. And intuitively, it's much more likely that a mathematical object has exactly no exploitable structure (like hashes are intended to), than that a mathematical object has exactly ~3 forms of exploitable structure (for elliptic curves: associativity, Schoof, pairings) and not some secret fourth form of structure we have not yet discovered that greatly degrades its security (for elliptic curves, ECDLP and pairing security). Similar for LWE, SVP, RLWE and the zoo of lattice problems. For this reason, we do not yet see any reason to worry and start padding the byte size of hashes (if we start to worry more, we would pad the round count first before doing anything to the byte size). Concrete TLDR, my own personal views: * Hash-based > lattice-based, in those situations where hash-based is possible at all * For anything lattice-based, be much more paranoid on param sizes. Remember that blockchains are only a small portion of the cryptography story; this point goes far beyond blockchains and applies to eg. access to websites, secure messaging, Tor / VPNs ... * For privacy protocols, strongly favor NOT putting encrypted notes onchain. Instead, send them offchain through some third-party mechanism. * If it's not difficult for you, keeping your funds in addresses which have not yet been used to make a transaction is a good idea. If it's easy for you, do it. **But be careful about migrations; I personally have lost more money in botched migrations than I have lost in all hacks combined**. * For multisig wallets, doing confirmations offchain is better than onchain, because this way the signatures of signer wallets do not get exposed to the public, so if ECDSA falls to AI much faster than expected, at least the multisig "gracefully degrades" to a 1-of-1 where the 1 is whoever was gathering the signatures - a much better place to be than "anyone can take the money" firefly.social/post/x/210783…
1,246
1,910
892
11,143
3,964,685
家人们!去参加 BNB 活动才发现,原来我在中国这么有名 大家都叫我「西八姐姐」😂(「西八」= 韩语的「f*ck」) 中国粉丝一看到我就喊「西八姐姐!」,还学我的反应,太可爱了 谢谢你们的喜欢,下次去中国找你们玩! 你们最喜欢我哪个反应? (I just got back from a BNB event! Chinese fans call me "Shiba Sister" 🇨🇳 "Shiba" means "f*ck" in Korean)
257
55
34
1,188
163,046
九月底到這周, 台股ASIC板塊表現的很不錯 這邊我把6533晶心科拉出來, 稍微整理一下目前我對於他的看法 1. ai asic算是已經殺出一條血路,google以及其他巨頭們為了不受制於nv的gpu算力威脅,已經投入大量資本在自研asic上,且已經開始有跡象 (像是google這波tpu超級雄起, 外溢到聯發科、創意) 2. 承上, 現在ai asic仍大多使用arm架構, 但採用arm作為solution勢必也逐漸會受制於arm的高額授權金成本, 外加arm架構相較低的擴展性, 這種時候就相對利好零權利金+高擴展性的RISC-V架構 3. 那晶心科有什麼優勢? - 全球30% up RISC-V市占 - 向量處理很猛, 我認為跟SiFive有得拚 (雖然兩者主攻場景還是有點不同), 而向量處理正是ai運算的日常場景 - 相較低市值, 約190億元 4. 風險 - 因各巨頭各自都有養自己的RISC-V團隊, 假如大廠不選擇6533作為solution的話就GG, 尤其大腿meta也收購了rivos, 假如下一代MTIA不繼續採用6533而是選擇自研, 那我看就真的危 - 中國去ARM化, 養了一大狗票人尻RISC-V來滿足中國內需市場, 基本上價格戰應該是打不贏 - 近期有現金增資稀釋股本 結論: 我認為一項新的技術要很瘋狂地進入mass adoption時, 勢必會經歷一波成本壓縮的過程, 講白話一點就是東西太貴會沒人想用 AI經過去年到現在的瘋狂擴張, 雖然各巨頭們仍然是瘋狂燒錢競爭中, 但市場上已經慢慢地開始有一些主打省錢且立志扮演便宜螺絲釘腳色的solution, 就好比是近期的JEV, 在模型端沒有打算與通用LLM競爭, 只是想要作為低廉+快速的決策模型 而在AI產業的其他板塊上我認為也是會有相同的事情發生, 尤其在AI運算處理器上, 而事實上這也是現在進行式, 從一開始的GPU, 到這波ASIC by ARM, 到接下來我認為ASIC端可以玩的市場: ASIC by RISC-V, 其高擴展性+便宜的優勢應該會有其一席之地 (像是computing in memory, edge computing等等) 這並不是說舊有GPU、ASIC by ARM就不需要了,反而我是認為會逐漸變成一個相互合作的模式,不同的solution用在不同的場域上 而這種事情假如發生, 且市場願意青睞6533, 那麼現在6533的價格我認為是可以入手的, 雖然上面提的風險都不可忽視且都蠻致命的, 但買在分歧, 就賭看看囉 寫個自己, 給之後的自己驗證看看的, NFA
4
10
1,617
0x883cb7c36c09d8df481c461b50529b768d641e18 $STOKC
Introducing the community vault LP upgrade. The next step for LONG stock pairs: becoming the liquidity backbone for tokenized stocks on Robinhood by holding the largest external liquidity positions. Top LONG pairs already capture 20–40% of their underlying stocks’ circulating supply. They drive demand, concentrate liquidity, and generate substantial trading volume through main stock pools (STOKC/USDG pools for example) Now, community vaults can put their stock liquidity to work directly in those pools. By reallocating liquidity into USDG/STOCK pools, stock pairs can become the largest LPs supporting the same stocks their communities are built around, capturing more of the upside from the activity they help create and any general stock activities. LP fees flow back into the vault, where they can be used to compound the same USDG/STOCK LP position and fund upcoming modules. From driving demand to providing liquidity, earning fees, and building deeper community-owned liquidity over time. The rollout starts with OG pair vaults, beginning with the highest-value vaults. We’ll gradually expand access to this feature to all pairs on LONG, including pairs without active vaults. LONG.
1
399
Tols retweeted
THE VLAD TENEV INTERVIEW. @vladtenev @RobinhoodApp @therollupco Timestamps: 00:00 Intro 00:21 New Markets Opening Soon For Everyone 02:27 $1B Daily Perps Volume On Robinhood Chain 04:30 #1 Chain For Developers In Two Months 05:08 Vlad Defends The Industry On CNBC Live
105
77
37
727
370,508
Tols retweeted
Robinhood Chain 不是只在炒股票上链,真正的 Alpha 才刚开始 $AI、$PONS 已经跑出第一波,下一轮资金开始往 RWA DeFi、AI Inference Market、Agent Economy 移动 Robinhood 尚未全面爆发
8
22
5
52
54,396
0x35985eb91a78bf5110d32af0db8f1256d19e7777
3
1
4
1,059
0x90077407ab60bf9675deaf9a480f2350e8e77777 pair marscoin-4:native
Heart-shaped crater on Mars
455
day 23 $MEME is a direct play at the $AMC holders who have been through the ringer for the last five years. These guys have watched the value of their stock get diluted many times over. They blame Robinhood and Vlad Tenev for disabling the buy button. The AMC investing public still feel like they were cheated out of generational gains. For $MEME to reach escape velocity, the crypto public had to show those holders the art of the possible. What happens when you unite behind a coin on a permissionless, distributed network, where nobody can disable a buy or a sell button. Crypto needs a narrative. A story that $MEME is an exciting new distribution mechanism for traditional equities that have been tokenised. A trip back to the halcyon days, when blockchain people were mocking TradFi and their antiquated ways and having fun whilst doing it. That pairing matters. $MEME is not AMC stock, and tokenised AMC is a listed share. They just share a pool. Buy $MEME and you are pulling tokenised AMC into that pool. That is the whole MEMEFI trick. Unfortunately on-chain trading is PvP, and the mentality is securing the bag. You cannot sell-shame someone for taking profit. But for something like $MEME to work over the long term, the collective had to get behind the story and show retail $AMC that dreams can still come true. As stories of the little guy winning emerge, the narrative perpetuates. Perhaps I am an idealist. I need to believe this is still possible, and that this place is not just a cesspit of the worst humans imaginable. Everyone wants to contribute to a winning team, but the individualistic nature of trading, and the intrinsic distrust of other market participants, is exactly why on-chain looks the way it does. The simple fact is whales have kept adding to $MEME. There have been some new mid-sized buys, and none of the top 20 holders have left. It has only been small holders selling, chasing green candles on other coins. The markets need a little cinema, and I am choosing to be part of the movie. 🍿
8
6
1
58
4,385