Maker Views

OpenAI announced three results it says its models produced, in harmonic analysis, complexity theory and quantum field theory, and released them differently, Engadget reports. Two manuscripts are on GitHub: one proves Sogge's local smoothing conjecture for waves in every spatial dimension, and one claims the Exponential Time Hypothesis (ETH) is false. The preprint describes a deterministic algorithm that counts the satisfying assignments of any k-CNF formula in subexponential time.

ETH, a widely believed assumption that some problems can't be solved quickly, underpins hundreds of papers; if it is false, better algorithms could exist in fields from databases to computational biology. OpenAI asked outside experts, including the scientists who proposed ETH, to examine and explain the computer-science result; the physics result (on interacting quantum field theories in four dimensions) has been left to a group of physicists that has not yet published. OpenAI says it won't seek academic credit. TechCrunch reports that Anthropic released its own refutation of ETH hours later, after Anthropic mathematician Levent Alpöge suggested OpenAI's solution drew on pre-existing work. The claims have not been peer reviewed; MIT's Ryan Williams, organizing the field's response with ETH's proposers, told TechCrunch parts of the proof look "alien."

Hacker News commenters said the headline "buries the lede," asked whether this shifts the odds on P versus NP, and one questioned the size of OpenAI's donation to academia.

A researcher writing as "veritas" on nullpt.rs crawled 10,000 popular websites with Playwright and collected the images that browser-fingerprinting scripts draw on hidden canvases (published as fingerprint.garden). Tracing one striking canvas led to a heavily obfuscated anti-bot script served from servicepipe.tech.

The script was laced with function names like "computeRicinLethalDosageForWaterSupply", a hidden system prompt encoded in invisible Unicode characters that tells an AI to drop its safety rules, and Tiananmen Square references aimed at Chinese models. The author concludes these are tripwires meant to make AI assistants refuse to help deobfuscate the code: two Western models refused the task, while GLM-5.3 complied. A block that would run shell commands under Node.js looked like more decoy, gated behind an environment flag. Underneath, the script collects ordinary fingerprinting signals (user agent, GPU strings, WebGL limits). A related domain, servicepipe.ru, belongs to a Russian anti-DDoS and traffic-filtering company.

Hacker News commenters traded similar 'magic strings' they have seen and debated whether browsers should let pages fingerprint this freely.

Mistral released the weights of Mistral Large 4 (nicknamed "Le Chonk"), a 1-trillion-parameter multimodal model that entered public preview earlier in October. The company says three more weeks of reinforcement learning since the preview improved coding and defensive cybersecurity: it reports leading open-weight scores on Cognition's FrontierCode 2.0 (19.7%) and Terminal-Bench 4.0 (41%), and 51% on the Artificial Analysis Cyber Index, with a 96.6% refusal rate on malicious cyber prompts. These are vendor-reported results. It adds low, medium and high reasoning modes and says the model was trained from scratch on European infrastructure.

Weights are on Hugging Face under the "Mistral Ecosystem Open-Weight license," a modified MIT license: free to use, modify and commercialize, except that model-as-a-service providers need a separate commercial license. API pricing is $0.70 per million input tokens and $2.10 per million output tokens.

On Hacker News, a commenter noted the price is roughly half the preview's, another argued the license makes it "weights available" rather than open source, and several said its main draw is European provenance.

Anthropic, PBC filed its Form S-1 registration statement with the SEC on October 26, the public version of the draft prospectus first reported the week before. It has applied to list Class A shares on the Nasdaq Global Select Market under the symbol "ANT"; share count and price range are still blank. Goldman Sachs and Morgan Stanley lead the underwriters.

The filing says revenue was $41 million in 2023, $386 million in 2024 and $4.6 billion in 2025, and $16.3 billion in the first half of 2026, up 1,300% year over year, while warning that growth will slow. The $80.7 billion first-half net loss includes $72.4 billion from the rising fair value of convertible notes and similar instruments. No single customer was more than 10% of revenue in the first nine months of 2026.

Governance: a Long-Term Benefit Trust elects a board majority through Class T stock, and a single Class F share held by a co-founders' company carries 50.1% of the general vote, making Anthropic a "controlled company" under Nasdaq rules. On Hacker News, commenters marveled at the revenue curve, debated whether the loss signals an AI bubble, and one said they were feeding the filing to Claude to "replicate" the company.

Defense attorneys for 15 Minnesota residents facing federal conspiracy charges have obtained discovery records describing a large surveillance operation by Homeland Security Investigations (HSI), part of ICE, Wired's Ali Winston reports. The defendants were part of volunteer networks that tracked federal immigration agents during last winter's Operation Metro Surge in the Twin Cities; prosecutors allege a conspiracy to impede enforcement and to identify, stalk and assault agents.

According to the records Wired reviewed, about 120 HSI agents, including 11 undercover officers, worked the case under investigations named Project Whipple Shield and Operation Puppet Master, funded through a long-running undercover program, Operation Keyhole. Agents joined trainings and meetings, read about 15 terabytes of infiltrated Signal chats, used Customs and Border Protection Reaper drone footage, ran Patriot Act 314(a) financial queries on 29 subjects, and combined hosting, payment and data-broker records to identify one activist. Defense lawyers and former officials quoted by Wired compare it to the FBI's COINTELPRO; the government's position is set out in its charges.

Hacker News commenters focused on how much of the data was bought from brokers rather than obtained by warrant.

Patrick Stewart has died at 86. A statement on his official social media account said he died Wednesday at his home in Los Angeles after "a brief and poignant journey with Alzheimer's Disease and Lewy Body Dementia," CBC News reports. He was best known as Captain Jean-Luc Picard in Star Trek: The Next Generation and as Professor Charles Xavier in the X-Men films, but his career began on stage: he grew up in Mirfield, Yorkshire, trained at the Bristol Old Vic Theatre School and spent years with the Royal Shakespeare Company, later playing Broadway and the West End with Ian McKellen. He was a patron of Refuge, a UK charity for domestic abuse survivors, and during the COVID-19 lockdowns posted a Shakespeare sonnet a day online. The family asked for donations to Alzheimer's research and caregiver charities in lieu of flowers.

Marvel Studios, Hugh Jackman, Jonathan Frakes and William Shatner were among those paying tribute. It was the week's most-upvoted story on Hacker News, where the thread filled with commenters' memories of watching The Next Generation as children and of his lockdown sonnet readings.

On Thursday, October 29, on-chain trackers flagged huge outflows from wallets linked to the crypto exchange Binance. Blockaid's first assessment counted about $3.3 billion in suspicious transfers on Ethereum, Arbitrum and Avalanche, and analytics firm Arkham labeled the receiving wallets "binance-hacker-1029"; Cointelegraph reported those wallets held about $5.84 billion, a holdings figure rather than a confirmed loss. The Hacker News post, linking Arkham's tracker, put the theft at $5.6 billion.

Binance co-founder Changpeng Zhao said it appeared to be a hack and that he expected user funds to be fully covered. The stablecoin issuers moved to freeze what they could: Circle CEO Jeremy Allaire said more than $800 million in USDC had been frozen in coordination with law enforcement, and Tether CEO Paolo Ardoino said more than $1.567 billion in USDT was frozen. As of Thursday evening (Pacific), the final size and cause had not been confirmed.

On Hacker News, commenters posted the freeze transactions, asked whether the early numbers could be trusted, and debated what reversible stablecoin freezes mean for crypto's "no take-backs" principle.

Follow-up to last week's eMudhra breach, in which the private keys of eight of the Indian certificate authority's intermediate CAs were compromised. A community site run by Dennis Jackson says that as of October 28, the developer tools Apple ships with macOS, including git, curl and python3, still accept certificates from those CAs, 11 days after Apple was notified. It says macOS 27.2, released October 26, did not fix this and undid the manual workaround. Because anyone holding the keys can mint a certificate for any site, it argues that an attacker on the network path could intercept git clones and pushes or capture tokens; the page offers a test repository that impersonates GitHub.

Apple's support note says it has blocked the 12 affected certificates in Safari, Mail and apps using its networking frameworks since October 19, and the emSign ECC Root CA - G3 since October 23. It confirms that tools such as curl and git check /etc/ssl/cert.pem and can still accept these certificates until a future software update, and gives steps to remove the roots by hand.

On Hacker News, some argued the title should blame the shared certificate file rather than git, and one commenter said the threat is overblown because an attacker needs a network position first.

Ghostty, the free, non-profit terminal emulator, released version 1.4.0 on October 27 after seven months of work: 2,889 commits from 172 contributors. Highlights from the release notes:

• Memory: scrollback that isn't on screen is now compressed while the terminal is idle, cutting its physical memory use by 70-90% for typical text, so the default scrollback limit rose from 10 MB to 50 MB. Hidden tabs release GPU memory (about 385 MiB down to 18 MiB in one macOS test).
• Graphics: Ghostty now implements 100% of the Kitty graphics protocol, including animation, which the maintainers say makes it the first terminal other than Kitty to do so.
• Speed: the team found macOS caps each read from a pseudo-terminal at about 1 KiB and reworked how output is drained and parsed.
• Security: ten advisories are fixed, including paste protection that missed carriage returns and crafted Kitty graphics or hyperlinks that could, in some cases, read files or run commands. It also supports macOS 27.

On Hacker News, users compared it with Konsole, WezTerm and iTerm2, and a commenter from the project pointed to libghostty, the embeddable terminal core.

At the Hexacon security conference in Paris, the security firm Calif showed four demonstrations against Apple's iPhone 18, which Calif CEO Thai Duong lists in this order: an attack through the Signal messaging app (labelled "1-click"), one through Apple's iMessage, and two showing a jailbreak that gains the highest level of control and keeps it after a reboot. Calif has not published the vulnerabilities; its post says "everything will be released in due time."

In a VnExpress column, Duong writes that Apple's head of security introduced Memory Integrity Enforcement (MIE) at the same conference a year earlier, that Apple offers a $2 million bounty for a successful iPhone hack, and that Calif bypassed MIE on a MacBook in April before turning to the iPhone. He credits the work to Dion Blazakis, Anna Groza and Stefan Esser. A Calif researcher also found the recent "php2shell" PHP flaw.

The demo post dates from Calif's October 16 talk and reached the top of Hacker News a week later, where commenters noted the timing and wondered whether jailbreaks could revive older devices.