Stay informed. Keep your time.
7 stories · ~3 min readEstimated reading time for the entire edition.
Bloomberg says OpenAI has raised $11 billion so far in a new funding round valuing the company at $1.25 trillion. The round remains open; one source said the total could reach about $20 billion, below the previously reported $30 billion target.
Bloomberg reports that OpenAI has raised $11 billion so far in a funding round that has yet to close, valuing the company at $1.25 trillion.
The New York Times reports that the address of New Matrix, approved by the US to buy H200 AI chips, leads to an empty office. The newspaper says the company’s owners are linked to a Chinese investment firm blacklisted by the US, and that other companies with similar ties were also approved to buy chips.
The New York Times reports that New Matrix, approved by the US to buy H200s, has an empty office address and links to blacklisted entities.
In an 11-week review, AI safety group METR found Anthropic’s AI-assisted R&D accelerating. Model development was about twice as fast as last year, and employees said they were 12 times more productive with its strongest internal model. Anthropic redacted most evidence, making independent assessment difficult.
METR’s report says Anthropic’s models are developing about twice as fast; as of October, most of the acceleration is attributed to AI use.
Crusoe sold its unfinished Wyoming data-center project to Google, which will take over development and invest $17 billion in the campus. The sale price was not disclosed. On other projects, Crusoe offers GPU rentals and inference services for AI models.
Google took over Crusoe’s unfinished Wyoming data-center project and said it would invest $17 billion to develop the campus.
Researchers examining macOS 26 say two background services build a knowledge graph from people, communications, photos, calendar and location data. XGBoost models classify family relationships; when Apple Intelligence is enabled, an on-device LLM can also infer relationships from messages. The databases are plaintext but protected by Full Disk Access permissions.
A security study says macOS’s knowledge graph infers relationships from personal data and uses models that classify family members.
The authors changed instructions or starter code in 89 tasks, adapting reference solutions and tests. Claude Opus 5.5 scored 76.4% on MaxxBench, versus 53.4% for Gemini 3.8 Flash. Some agents understood new requirements but reverted to old test expectations. The paper has not been published.
MaxxBench modifies DeepSWE tasks to evaluate how well coding models adapt to new requirements.
The article describes Claude’s cross-session memory: extracting notes from session logs and organizing them during idle “dreaming.” In tests, wiki-assisted retrieval raised performance on multi-document questions from 73 to 87 compared with retrieval alone; it showed no notable gain on a single-document task.
Claude uses session logs to produce, verify and organize memory notes, then link them in a wiki format.
That’s it for today.The rest of your time is yours.
↑Stories are selected from Hacker News; accessible original articles are summarized and translated with AI. Summaries may contain errors; check the source beneath each story for details. Preparation starts daily at 06:55 Türkiye time (UTC+3), targeting publication before 07:00.
Stories are read aloud by an AI voice.
This edition covers articles shared on Hacker News on the date shown above.
Read or listen to a new edition every morning with Kısa for iPhone, save stories for later and keep your streak.
