Gemini 3.5 Pro delayed: Google admits that the frontier release is now stuck on the code
Google differs Gemini 3.5 Pro to improve coding. The metric that blocks a frontier release is no longer knowledge, it's tool-use.
18 h ago 8 11
"The LLM critics are right. I use LLMs anyway" - the voice that reassembles
A post from theocharis.dev makes it to Hacker News: "LLM critics are right. I use them anyway." The assumed compromise position becomes an editorial style in itself.
yesterday 10 8
The archaeologist and his copilot: Malykhin disciplines the LLM on Java 1.5
Nik Malykhin had to run a Java 1.5 base on modern hardware. The first answers from his LLM were plausible, but did not hold up to scrutiny. The breakthrough: stop believing it at face value, force it to rely on evidence.
yesterday 9 8
Anthropic provides its user manual for scaling agents: layers, tokens, reliable execution
Two engineers from Anthropic publish field lessons: for an agent that needs to scale, it's no longer the model that limits, but the harness. Three pillars: layers, budget, execution.
Jul 15, 2026 at 12:34 7 5
Delangue: the real race may no longer be at the border
41% of Hugging Face downloads in spring 2026 for open Chinese models, the six most used models on OpenRouter all open. The center of gravity shifted while we were looking at frontier benchmarks.
Jul 14, 2026 at 22:32 9 5
Meta sued: when an AI scoring system for a layoff plan meets labor law
26 former employees accuse Meta of using AI systems - including a productivity score based on keyboard strokes - to rank who to lay off. Among the alleged criteria: token usage.
Jul 14, 2026 at 22:30 8 7
OpenAI imposes hardware keys on its cybersecurity researchers
Hardware-bound passkey mandatory as of September 1st to access cutting-edge cyber models. A standard measure - and that's precisely what makes it interesting.
Jul 14, 2026 at 22:30 10 8
Codex now encrypts prompts before inference
The OpenAI Codex coding tool has started to encrypt prompts and use the ciphertext directly for inference. The change raises questions about the reproducibility, observability, and privacy of developers' inputs.
Jul 14, 2026 at 15:44 12 8
Wiz launches an urban forecasting AI model - China urbanizes forecasting
A Chinese company, Wiz, publishes an AI forecasting model at the city level: traffic, energy, air. Focus on "planning" rather than public use - a sign of maturity in an under-the-radar segment.
Jul 12, 2026 at 11:15 13 10
Sugon Dawn 8000 - China assembles its first 100,000-card cluster
Sugon announces Dawn 8000, a cluster of 100,000 "ultra-intelligent" accelerator cards. Symbolic + industrial: Beijing catches up with the threshold of Western megaclusters, with domestic chips.
Jul 11, 2026 at 17:15 14 7
Copyright vs. training data: the NYT vs OpenAI dispute, and Google's AI ad-disclosure rule
The New York Times is asking a court to sanction OpenAI over training-data discovery; Google now requires advertisers to disclose generative AI use in ads. Two regulatory pincers on the same problem: what "AI-generated" means when the law wants a source of truth.
Jul 11, 2026 at 11:41 7 8
GPT-5.6 and ChatGPT Work: OpenAI drops a workhorse, Anthropic answers with a reset
OpenAI ships GPT-5.6, folds Codex into ChatGPT Work for apps and files, and cuts the top-tier price in half. Anthropic responds the same day by wiping Claude's usage limits. The frontier war is now billed by the meter.
Jul 11, 2026 at 11:41 8 9