Schools that disable Google AI will also lose access to Search
Starting next year, Google will disable AI tools for K–12 students by default, while allowing school administrators to turn them on. Schools that block AI entirely will also lose access to Google Search, which Google says is intertwined with its AI features.
Google will let schools disable AI tools for students, but schools that do so will also be unable to use Google Search.
02SECURITY
Anthropic warns AI-assisted research could accelerate progress
In an updated S-1 filing, Anthropic said it could soon reach the “AI-attributed acceleration” threshold in its Responsible Scaling Policy. It says none of its current models are considered to have crossed the threshold and that it is still evaluating the data; it is considering a policy update on human oversight.
The company says AI-assisted model development could outpace safety measures, and it may be nearing a progress threshold it set for itself.
03RESEARCH
Essay: Claim Astra improved Qwen by inspecting raw weights is misleading
The essay argues that it is misleading to suggest Astra improved Qwen3.5-0.8B simply by inspecting its raw weights. The Qwen experiment uses Python and runs the model on up to 50 sets of practice problems; the author says this differs from raw-weight experiments.
The essay says Astra’s Qwen result did not come from inspecting raw weights alone: it involved Python and running practical problems.
04INDUSTRY
Anthropic’s monthly personnel cost per employee is about $107,000
Anthropic’s amended S-1 says it recorded over $1.7 billion in personnel expenses for 5,300 employees from July through September, or about $107,300 per employee per month. Of that total, $1.1 billion was equity-based compensation not paid to employees in cash. The company says an IPO on September 30, 2026, would trigger $8.2 billion in expense that day for restricted-stock awards.
The company’s amended IPO filing shows more than $1.7 billion in personnel expenses for July through September.
05RESEARCH
Could modern Stockfish on a Pentium Pro beat Kasparov?
Nicholas Carlini argues that modern Stockfish adapted for a Pentium Pro from 1996 could be strong enough to beat Kasparov. Its estimated 3,300 Elo rating extrapolates from short games to classical time controls; in a 100-game comparison, it beat the older Crafty engine 78–17–5.
A writer estimates that modern Stockfish adapted for a 1996 Pentium Pro might be strong enough to beat Kasparov; this is not based on a match against him.
06MODELS
Mistral says it has released the weights for its Large 4 model
The company says its trillion-parameter model was released after additional training and showed gains on its own benchmarks. Large 4 also offers three reasoning modes to balance capability against inference cost; performance information comes from Mistral’s own claims.
Mistral says it has released the weights for its multimodal Large 4 model, which it says received coding and defensive security training.
07POLICY & LAW
Writer asks SEC to halt Anthropic’s IPO
Scherer argues that the SEC should halt Anthropic’s IPO. The essay says the company’s 2026 financials are unaudited and that it has at least $582 billion in non-cancelable spending commitments to five companies. The author calls for two consecutive audited quarters and disclosure of the commitments before an IPO.
An opinion piece argues that taking Anthropic public with its current financial disclosures could put public investors at risk.
NVIDIA’s Auto-Ontology gives AI agents context by turning database schemas into an ontology of business meanings, SQL metrics, and table relationships. Users can review and approve definitions in an interface; the tool also shows the SQL and steps behind its answers. The software is Apache 2.0-licensed and supports ten database engines.
Nvidia’s Auto-Ontology turns database metadata into a reviewable ontology to help agents generate queries over enterprise data.
✓
That’s it for today.The rest of your time is yours.
Your daily technology digest, every morning in English. Free. Unsubscribe anytime.
Sources, in the open.
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.
This edition covers articles shared on Hacker News on the date shown above.
Now on the App Store for iPhone
Your daily digest. Now in your pocket.
Read or listen to a new edition every morning with Kısa for iPhone, save stories for later and keep your streak.