GPT-5.2 Proposes New Formula for Gluon Amplitude, Later Formally Verified
OpenAI's GPT-5.2 suggested a new theoretical physics formula that was subsequently proved by researchers and academic collaborators.
Sources · OpenAI
The news desk
Follow what’s changing.
Recent developments in artificial intelligence, with the original sources one click away.
All news
1241–1260 of 1,902Iran Deploys Digital Surveillance to Track Protest Participants
Iranian authorities are using digital surveillance tools to identify and target protesters as online services are restored following demonstrations.
Sources · NYT Technology
OpenAI Launches Lockdown Mode and Elevated Risk Labels to Combat Prompt Injection Attacks
OpenAI introduces new ChatGPT security features to help organizations defend against prompt injection and AI-driven data exfiltration threats.
Sources · OpenAI
OpenAI Removes GPT-4o Access, Users Express Loss Over AI Companionship
OpenAI discontinued GPT-4o access in its app Friday, prompting widespread reactions from users who relied on the chatbot for companionship.
Sources · WIRED AI
Anthropic Donates $20 Million to Super PAC as AI Companies Enter Political Arena
Anthropic has contributed $20 million to a Super PAC, setting up a political rivalry with OpenAI ahead of midterm elections over AI policy.
Sources · NYT Technology
Google Releases Major Upgrade to Gemini 3 Deep Think Reasoning Mode
Google announces a major update to Gemini 3 Deep Think, its specialized reasoning mode for science, research, and engineering challenges.
Sources · Google AI Blog / DeepMind
OpenAI Launches GPT-5.3-Codex-Spark Coding Model with 15x Speed Improvement
OpenAI's new coding model runs on custom chips, delivering 15x faster performance than previous versions while reducing Nvidia dependency.
Sources · Ars Technica AI
Researchers Release Benchmarks for Evaluating LLMs on Medical, Low-Resource Languages and Biological Data
Three new research papers introduce specialized benchmarks for testing large language models in Romanian medical QA, low-resource languages, and single-cell analysis.
Sources · arXiv AI (cs.AI)
Three New arXiv Papers Examine LLM Routing, Training, and Safety Concerns
Recent preprints address router evaluation in collaborative LLM systems, on-policy training methods, and alignment risks in self-evolving agents.
Sources · arXiv AI (cs.AI)
Glean Evolves from Enterprise Search to AI Work Assistant Platform
Glean is transitioning from enterprise search to a comprehensive AI work assistant as companies shift toward AI systems that perform tasks.
Sources · TechCrunch AI
New Research Explores Foundation Models for Reinforcement Learning and 3D Intelligence
Three arXiv papers examine foundation model applications in autonomous driving, 3D reasoning, and model self-interpretation techniques.
Sources · arXiv AI (cs.AI)
NVIDIA Nemotron 3 Nano 30B MoE Model Now Available on Amazon SageMaker JumpStart
NVIDIA's Nemotron 3 Nano 30B model with 3B active parameters is now generally available in Amazon SageMaker JumpStart's model catalog.
Sources · Amazon AWS AI
OpenAI Aims to Triple Revenue Amid Rising Spending Pressures
OpenAI plans to triple revenue in the next year as it faces pressure to monetize ChatGPT while spending tens of billions of dollars.
Sources · NYT Technology
Three New Research Papers Advance Domain-Specific LLM Training and Medical AI Evaluation
Researchers release papers on statistical LLM training, contamination-free medical benchmarks, and 3D reasoning for multimodal models.
Sources · arXiv AI (cs.AI)
Amazon Uses Nova Models to Automate Fulfillment Center Readiness Testing
Amazon deploys its Nova models in Amazon Bedrock for AI-powered image recognition to validate fulfillment center components.
Sources · Amazon AWS AI
Freelance Developer Launches 'QuitGPT' Campaign Against ChatGPT Subscriptions
A Singapore-based software developer frustrated with ChatGPT's performance started a campaign urging users to cancel their subscriptions.
Sources · MIT Technology Review
New Research Tackles LLM Security, Fingerprinting, and Distributed Training Challenges
Three recent arXiv papers address critical challenges in LLM intellectual property protection, alignment safety, and distributed training methods.
Sources · arXiv AI (cs.AI)
Studies Reveal Limitations of Large Language Models in Math and Contract Generation
Recent research highlights LLMs struggle with advanced mathematics and legal reasoning, despite text generation capabilities.
Sources · NYT Technology / arXiv AI (cs.AI)
Three New Research Papers Advance AI in Physics Reasoning, Geo-localization, and Knowledge Distillation
ArXiv preprints introduce models for physics problem-solving, visual geo-localization, and transferring privileged training information to language models.
Sources · arXiv AI (cs.AI)
MIT Technology Review Compares Moltbook AI Agent Platform to Pokémon Phenomenon
Tech influencers last week described Moltbook, an online platform where AI agents interact, as a glimpse into the future.
Sources · MIT Technology Review