HackerNews Digest

The letter “We Will Not Be Divided” seeks a broad coalition to address concerns about the potential misuse of AI against Americans. Organizers state they are independent citizens, not tied to any political party, advocacy group, AI company, or paid entity. Current and former Google and OpenAI employees are invited to sign; each signature is verified before inclusion. Verification options include Google Form or email confirmation (requiring a @google.com or @openai.com address) and an “alternative verification” reviewed manually. Anonymous signers have their personal data (name, email) deleted after 24 hours, leaving only an anonymous public listing; a single organizer can view the data during that window. The site runs on Fly.io, uses an encrypted SQLite database, sends verification emails via Resend, and is built on an open‑source Flask application with DNS/SSL managed by Cloudflare. No analytics or tracking scripts are employed. Organizers acknowledge occasional verification errors and commit to logging and correcting them.

Secretary of War Pete Hegseth announced that the Department of War intends to label Anthropic, an American AI firm, as a supply‑chain risk after negotiations stalled over two requested exceptions: (1) use of Anthropic’s Claude model for mass domestic surveillance of Americans, and (2) deployment in fully autonomous weapons. Anthropic states it supports all lawful national‑security AI applications, notes that neither exception has been used in any government mission, and argues that current frontier models are not reliable for autonomous weapons and that mass surveillance would breach fundamental rights. The company claims such a designation is unprecedented for a U.S. company, would lack legal basis, and plans to contest it in court. Under 10 U.S.C. § 3252, a supply‑chain‑risk label would only affect Claude’s use on Department‑of‑War contracts, leaving API access for individual or commercial customers unchanged. Anthropic’s sales and support teams are available for inquiries and the firm thanks its users and supporters.

The page references a GitHub repository named **anadim/AdderBoard**, described as the “Smallest transformer that can add two 10‑digit numbers.” Access to the repository’s content is blocked, as indicated by the message “You can’t perform that action at this time.” The only visual elements listed are three images with alt‑text labels: “AdderBoard,” “@anadim,” and “@claude.” No further technical details, code, model architecture, training data, performance metrics, or usage instructions are provided in the scraped text. Consequently, the available information is limited to the repository’s title, the access restriction notice, and the three image descriptors. No additional substantive content can be summarized.

OpenAI announced a $110 billion private funding round, the largest in history, comprising a $50 billion investment from Amazon and $30 billion each from Nvidia and SoftBank, based on a $730 billion pre‑money valuation. The round remains open for additional investors. Key components include extensive infrastructure partnerships: Amazon will host OpenAI models on its Bedrock platform, expand AWS compute services by $100 billion, and allocate at least 2 GW of Trainium chips for training, while Nvidia will provide 3 GW of dedicated inference capacity and 2 GW of training on Vera Rubin systems. Amazon’s contribution may include a conditional $35 billion tied to achieving AGI or an IPO. The previous round in March 2025 raised $40 billion at a $300 billion valuation. Executives highlighted the shift from research to large‑scale daily use, emphasizing the need for rapid infrastructure scaling to meet global demand.

None

California’s Assembly Bill 1043, signed by Governor Gavin Newsom and effective 1 January 2027, mandates that any operating‑system provider collect an age indicator during account creation and expose a real‑time API signal that categorises users into four brackets: under 13, 13‑15, 16‑17, and 18 or older. The OS must present an accessible interface for entering birth‑date or age and provide developers who request the signal with a digital indication of the user’s bracket. Windows already complies via Microsoft‑account DOB entry; Linux distributions face criticism from community members who argue the requirement is unenforceable in California and may lead to “not for California” disclaimers. The bill reflects a broader governmental trend toward statutory age‑verification mechanisms, echoed by the UK’s Online Safety Act and controversial face‑scan systems on platforms such as Discord, raising privacy concerns despite limited technical enforcement.

GitHub repository **wkoszek/emuko** provides a fast RISC‑V emulator implemented in Rust, capable of booting Linux. The page includes a notice (“You can’t perform that action at this time”) and an image placeholder with alt text “@wkoszek”.

OpenAI’s investigation uncovered a Chinese law‑enforcement official using ChatGPT as a private log of a transnational repression campaign targeting Chinese dissidents abroad. The user detailed tactics that included impersonating U.S. immigration officers to warn a U.S.-based dissident, forging U.S. county‑court documents to force removal of a social‑media account, creating a fake obituary with gravestone photos, and drafting a plan to smear Japanese Prime Minister Sanae Takaichi by exploiting anti‑U.S. tariff sentiment. OpenAI matched these entries to real‑world activity, confirming that false death rumors spread in 2023 and coordinated hashtag attacks appeared later. The operation reportedly involved hundreds of Chinese operators and thousands of fake online personas across multiple platforms. OpenAI banned the account after detection. The case illustrates how authoritarian regimes leverage AI tools for information warfare, occurring amid heightened U.S.–China AI competition, including a Pentagon dispute with Anthropic over model safeguards.

NASA Administrator Jared Isaacman announced a restructuring of Artemis after an Aerospace Safety Advisory Panel flagged excessive risk in the original schedule. A new 2027 flight will place astronauts in low‑Earth orbit to dock with one or both commercially built lunar landers (SpaceX, Blue Origin) and conduct integrated tests of navigation, communications, propulsion, life‑support, rendezvous procedures, and spacesuits. The revised Artemis III becomes this orbital test, not a lunar landing. Artemis IV and V, slated for 2028, will use the validated lander(s) for the first crewed moon landings, with the possibility of one or two missions per year thereafter. NASA will retain the current Block 1 SLS rocket and adopt a “standardized” upper stage, halting work on the more powerful Exploration Upper Stage to reduce configuration changes. The plan emphasizes step‑by‑step capability growth, increased launch cadence, workforce rebuilding, and risk reduction through Earth‑orbit validation before lunar surface operations.

The WHATWG Streams Standard, created between 2014‑2016, introduced a reader‑writer, lock‑based API that predates async iteration. Consequently, common operations require boilerplate (getReader, read loops, releaseLock) and expose lock‑management bugs when locks aren’t released. Although async iteration can be retrofitted, it cannot expose features like BYOB reads, which need a separate BYOBReader, buffer detachment handling, and are rarely used despite their intended zero‑copy benefits. Backpressure signaling via desiredSize and highWaterMark is advisory; implementations may ignore it, and tee() creates unbounded internal buffers because the spec imposes no limits. TransformStream’s push‑oriented design further decouples backpressure between writable and readable sides, causing unchecked buffering and extra promise chains. The specification’s heavy reliance on promises generates per‑read objects and async coordination overhead, leading to significant GC pressure in high‑frequency scenarios such as server‑side rendering. These design choices force runtimes (Node.js, Deno, Bun, Cloudflare Workers) to implement non‑standard, complex optimizations to achieve acceptable performance, resulting in portability and maintenance challenges across environments.