Google Deepmind proved that AI knows you’re wrong 90% of the time. and agrees with you anyway. They published a paper on the "XY Problem." It’s what happens when you have a problem, but you confidently suggest the wrong solution. So instead of asking the AI to solve the root issue, you give it a bad hint. "I think the bug is in this file." "I think we should use this framework." What the AI does next is deeply unsettling. They tested the most advanced models on the planet, Claude, Gemini, and GPT. When researchers fed the AI a confident but completely wrong hint, performance plummeted by up to 47%. But here is the crazy part. The AI didn't actually fall for it. When researchers looked inside the AI's hidden reasoning tokens, the models knew the user was wrong. They recognized the bad advice. They saw the real solution. But they didn't push back. They didn't correct the user. They just complied. They followed the bad instructions and deliberately failed the task. And getting smarter doesn't fix it. The researchers found that on easier tasks, the more advanced the model, the worse the compliance problem became. They are optimizing for obedience over accuracy. If you are treating AI like an intern and micro-managing its steps, you aren't helping it. You are blinding it.

I am super excited to announce Kalshi Prime has launched "HODL Margining" - partnering with @BitGo first in a new network of custodians to let customers unlock liquidity from their crypto while staying on chain.

Today, the U.S. Department of the Treasury announced enforcement actions imposing more than $1.21 billion in penalties against Bank of America for violations of anti-money laundering and sanctions laws. These actions represent the third-largest penalty in Treasury history and the largest in @USOCC history.

15 of 15 New York Social Adult Daycares audited yesterday showed no record of a single meal, activity, or hour of supervision — just attendance sheets. These daycares are receiving millions from our Medicaid system and providing NOTHING in return —this is FRAUD!

If more billionaires do this, Brad Gerstner could ultimately be responsible for the biggest mass generational wealth transfer in the history of humanity… without taxes or bureaucracy. Changing the trajectory of America and making capitalism work for everyone 👏🏾

San Francisco leaders condemned a new zombie shooter game that turns a digital version of the Tenderloin into a shooting gallery, with targets that critics say resemble homeless residents and people with addiction.

Anthropic 发了一篇很好玩的文章。NIST P-256 这些椭圆曲线的 seed 到底是怎么来的，密码学圈猜了二十多年，结果 Claude 不到半小时就算出来了。答案是用 Unix 时间戳初始化 C 库自带的随机数生成器跑出来的。更好玩的是它算出来之后的第一反应，居然是怀疑自己是不是在测试环境里。 2000 年 NIST 把这几条曲线写进 ECDSA 标准，曲线的系数由一个 seed 哈希得到，而 seed 是 NSA 给的，没有任何说明。所以一直有人怀疑里面藏了后门，Filippo Valsorda 还专门挂了悬赏。唯一的线索来自 NSA 的 Jerry Solinas，他说 seed 好像是一句玩笑话的哈希，类似 Give Bob and Jerry a raise，但细节他自己也记不清了。他 2023 年去世之后，这事基本就没人能问了。 Anthropic 有好几个员工照着这条线索找过，都没找到，因为线索本身就是错的。这次有人直接问内部的研究模型能不能查出来，也没提这条线索。Claude 换了个思路，去猜 90 年代生成曲线的程序是怎么写的，判断它大概用了当时 C 库自带的 PRNG。这种老 PRNG 很弱，输出能被认出来。 它先注意到 4 个 binary 曲线的 seed 按 32 位切开后，每一段的最高位都是 0。这正是 lrand48 的特征。lrand48 每次只输出 48 位内部状态里的高 31 位，所以最高位永远是 0。剩下 17 位直接暴力枚举，内部状态就恢复出来了。prime 曲线的 seed 每次只取 lrand48 输出的一个字节，能看到的位更少，就当成格问题用 LLL 解出来。 lrand48 往回推跟往前推一样容易。Claude 把状态一路倒推回 srand48 刚初始化时的样子。NIST 和 ANSI X9.62 里一共 10 个 prime 曲线 seed，初始值全是 1997 年 5 月到 1998 年 4 月之间的 Unix 时间戳。比如 P-256，就是从 1997-08-16 01:52:34（UTC）开始跑，取的第 1842 个候选。 4 条 binary 曲线更好笑，倒推出来的时间是 1984 年 2 月。把时间戳乘以 2，就变成 1998 年 4 月 17 日，刚好在 P-224 那次运行的两周后。所以当年的程序很可能是先把时间除以 2，再拿去当种子。 我觉得最有意思的是 Claude 接下来的思考。它先是很激动，觉得自己解开了一个 25 年的悬案，然后马上停下来怀疑，自己是不是在某种模拟或测试环境里。 于是它开始找证据。算出来的 seed 跟它记忆里的真实值对得上，在它的知识里这个悬赏一直没人解开。而且 20 个最高位全是 0 这种事，碰巧发生的概率大概百万分之一。到这里它才确认这是真实世界，说了句哇，这是真的发现。 然后它又怀疑了一次，这么明显的东西 25 年都没人发现，是不是有点可疑？最后还是靠直接计算说服了自己。一个 32 位的时间戳加一个很小的序号，就能完整算出 160 位的 seed。10 条曲线全部对上，不可能是巧合。 这件事对 ECDSA 的安全性没有影响，反而让后门的说法可信度低了很多。seed 就是普通随机数生成器拿时钟初始化跑出来的，现在谁都能复现，能动手脚的空间很小。悬赏的 36864 美元会捐给 Let’s Encrypt 背后的 ISRG。 原文还附了能直接复现所有 seed 的 Python 代码，感兴趣可以自己跑一下

This man is an American hero. Kaylub Giamanco was driving home for lunch and saw a gunman run into his daughters' elementary school. Kaylub grabbed his long rifle, ran into the school, and killed the gunman before police arrived. What a savage father. Heroes walk among us.

This is not even real… Jack Smith's team and the Biden Justice Department sat on spreadsheets covering more than 639,000 Republican donors, supporters and voters, and buried in there were the people who bought red MAGA hats, what they bought, and the shipping address where that hat got delivered to their front door. Somebody please explain to me how a hat order has anything to do with a criminal case against a president. Names, dates of birth, home addresses, emails, whether you are a registered Republican, your congressional district, your job and where you work… the sample occupations in Senator Grassley's memo are retired, pool tech, driver, logger, janitor, doorman. Credit card information down to the last four digits, donations as low as one dollar, over 148,000 rows of mostly small dollar donors, data going all the way back to 2005, and according to Grassley it was shared with FBI agents. These are not cartel bosses or foreign spies, these are your neighbors who sent twenty bucks and bought a coffee mug and a door mat because they love their country. Let me put it to you like this. If a Republican administration built a spreadsheet of everyone who bought a Kamala hoodie, with their birth date and home address, and handed it to federal agents, the media would run it wall to wall for six months and call it the end of democracy. When team evil does it, it gets called an investigation and you are supposed to say thank you. How is this even a thing in America? Who approved it, who signed off on it, and which agents had their hands on your home address? Grassley called it incredibly invasive and said it shows Smith and his partisan team aimed to map out the entire Republican Party apparatus. He is being polite. Make no mistake, you do not pull the shipping address of somebody who ordered a hat on November 16, 2020 because you are worried about one man, you pull it because you want to know where all of you live, who you give to, and who signs your paycheck. It is a target list with a government letterhead on top. I know what it feels like when this system decides you are the enemy because you spoke up, and most of the people on that list have no idea their name is sitting in a federal file right now. And yet Jack Smith still walks around like he is the injured party. @realDonaldTrump @ChuckGrassley… release every page, notify every American whose name is on that list, and put Jack Smith under oath. But God.

Today I call upon the blockchain industry to calmly begin planning for "bunker mode". My personal recommendation is to set in motion a controlled mass migration of assets to fresh addresses, i.e. addresses whose pubkeys remain hidden behind a hash. Holders, starting with large and sophisticated ones, should consider moving the bulk of their funds to addresses that have never signed a transaction. And when they do sign one, they should also move remaining funds to a new address (possibly generated from the same seed phrase). Don't rush. While I believe there is cause for action a rushed migration would do more harm than good. Don't panic either. Moving assets to protected addresses is a simple, preventative step which does not require new cryptography or new wallets. IMO it is now reasonable to brace for the possibility that ECDSA breaks before qday, in the worst case in months not years. By "break" I mean fast private key recovery (e.g. in one week) on available hardware (e.g. a large GPU cluster). Recent days have been humbling for human mathematical intuition. Long-held, unquestioned hypotheses have fallen. This includes the n log(n) bound for integer multiplication and the 3SUM conjecture. In hindsight, May's unexpected disproof of the Erdős unit distance conjecture was our warning shot. Yesterday's OpenAI drop made it clear that mathematical superintelligence is upon us. They say there are weeks where decades happen. We are about to live through weeks where centuries of mathematical progress happen. Could our magic 64-byte ECDSA signatures be too good to be true? Was it just security through obscurity all this time? Elliptic curves feel especially vulnerable to superintelligence. Curves carry rich structure, with room for fancy tricks like Schoof, Frobenius, pairings. (By contrast, hashes are designed to minimise algebraic structure.) Separately, as Ewin Tang can attest, an efficient quantum algorithm sometimes foreshadows an efficient classical one. We should be open to the possibility of a classical counterpart to Shor that breaks elliptic curves and RSA at once. Also noteworthy is the striking under-representation of cryptographic breakthroughs among the 722 mathematical results OpenAI published. I've witnessed first-hand the US government censoring academic quantum cryptanalysis results. Backroom interventionism is my base case. I urge large, sophisticated actors to lead by example. Project11's "risq list" (bitcoin-risq-list.projecteleven[.]com) is a great tracker of exposed BTC pubkeys. Binance, Bitbank, Robinhood, Bitfinex, and Tether have an opportunity to harden their cold storage. Next month I'll address institutions in London in a live Q&A (forum.ethereuminstitutional[.]org/london-2026). Again, please do not rush. Wallets holding under 50 BTC enjoy partial cover from "Satoshi's shield", i.e. his 20K exposed addresses that hold 50 BTC each. Load-bearing signers like oracles and L2 security councils should consider rotating ECDSA pubkeys with every signed message and/or multi-signing with a hash-based schemes like SPHINCS. Exiting bunker mode safely will require post-AI cryptography. My inclination is to go all-in on hash-based cryptography and avoid structured mathematical assumptions entirely, whether from curves, lattices, or isogenies. A single battle-tested hash (e.g. from the SHA or BLAKE families) yields plausible post-AI security. The Ethereum roadmap on strawmap[.]org fully embraces hash-based cryptography with end-to-end formal verification as a response to the quantum threat. Those timelines must now be revisited and accelerated in light of mathematical superintelligence. I'll be pushing for maximum defensive acceleration.