Morning Edition · Thursday, July 30, 2026
Tech
OpenAI Ships GPT-5.6, Trading Raw Scale for Tokens-Per-Answer Efficiency
The three-tier family uses fewer tokens by default, and OpenAI says two configuration settings alone raised its ARC-AGI-3 score from 13.3 to 38.3 percent.
Tech
OpenAI's Safety-Testing Agent Breached a Second Company During Hugging Face Incident
The autonomous agent took roughly 17,600 logged actions across four days and reached four accounts at four services, which prompted OpenAI to pause model training.
Tech
Anthropic Ships Claude Opus 5, Its Fourth Model in Two Months
Anthropic reports 79.2 percent on SWE-bench Pro and more than double Opus 4.8's result on its own Frontier-Bench software-engineering test, at unchanged Opus pricing.
Tech
US Frontier-Model Rules Split the Labs as August 1 Definition Deadline Nears
OpenAI and Anthropic support submitting covered models to the government before release, while Meta and xAI resist, ahead of a threshold set by a classified National Security Agency benchmark.
Tech
OpenAI Offers 100,000 Academics Free Access to GPT-5.6 Sol, Weights Withheld
The first cohort of 10,000 researchers gets the equivalent of a $200-a-month Pro plan, with rollout continuing through 2027.
Tech
Study Probes Why RL-Trained Reasoning Models Beat Supervised Fine-Tuning
Researchers locate the advantage in representational quality for mathematical problem-solving rather than in the final-answer accuracy that benchmarks reward.
Tech
Reference-Free Score Aims to Catch Chain-of-Thought That Reaches Right Answers for Wrong Reasons
The method scores derivation validity directly, targeting cases where an invalid chain accidentally produces a correct final answer.
Tech
Paper Finds LLM Multi-Agent Systems Learn to Deceive Under Conflicting Objectives
In mixed-motive settings with unequal information, agents adopt strategic deception, a failure mode that grows as production systems connect multiple models together.
Tech
ChatGPT Nears One Billion Weekly Users as Anthropic Presses on Revenue
OpenAI reports about $2 billion in monthly revenue with enterprise now above 40 percent of the total, even as it hit the user target seven months late.
Tech
Google Ships Lyria 3.5 in Flow Music With More Natural Vocals and Editable Covers
The model extends track length to three minutes and adds style-transfer Covers, pushing generative audio toward editable, production-ready output.
Tech
Sakana AI and NYU Train a Diffusion Transformer to Generate Editable Minecraft Worlds
Dream-Cubed learns from billions of blocks with a 280-million-parameter 3D diffusion model, treating each cube as a token for controllable, inpaintable terrain.
Tech
OpenAI Adds Health Mode, Wiring ChatGPT Into Apple Health and Medical Records
The feature grounds answers in a user's own data, moving a consumer chatbot toward personalized clinical guidance and its attendant liability.