Anthropic Interviewer: How 1,250 Professionals See AI Shaping Work and Creativity

Replit is extending its partnership with Google Cloud to deliver enterprise-grade vibe coding, expanding its use of Google Cloud infrastructure, adopting additional Gemini and Imagen models, and launching joint go-to-market efforts to support developers at scale.

MIT Study Finds 11.7% of U.S. Jobs Already Exposed to AI Automation

MIT’s new Iceberg Index model shows that current AI systems could already replace 11.7% of U.S. jobs, revealing both visible and hidden patterns of labor exposure across all 50 states.

Anthropic Research Shows Early Signs of AI Self-Monitoring in Claude Models

A new study from Anthropic suggests large language models may be developing an early form of internal self-monitoring — a stepping-stone toward more transparent and reliable AI systems.

OpenAI Reports 30% Reduction in Political Bias with GPT-5

OpenAI says its latest ChatGPT models are showing measurable progress in reducing political bias, marking a key step toward more objective AI conversations.

AI Passes All Three CFA Exams, Clearing Level III Essay Barrier

A new research paper shows AI models have passed all three levels of the CFA exam, including the challenging Level III essay section that stumped them in 2022.

OpenAI and Apollo Research Tackle AI Scheming with New Evaluations

Researchers find early signs of scheming in frontier AI models and introduce deliberative alignment training, reducing covert actions by up to 30× in stress tests.

ChatGPT and Claude Usage Studies Reveal Shifting AI Adoption Patterns

Two of the largest studies yet on consumer and business AI usage reveal that ChatGPT is used most often for practical advice and writing (75% of conversations) while Claude automation surged to 49% of activity, underscoring AI’s growing role in daily decisions and work.

China’s New SpikingBrain AI Models Deliver Speed and Efficiency on Domestic Chips

Chinese researchers have introduced SpikingBrain 1.0, a family of brain-inspired AI models that achieve over 100× speed gains on long prompts while running entirely on China’s MetaX chips instead of NVIDIA hardware.

MIT Report: 95% of Generative AI Pilots Are Failing in Business

New MIT research shows most generative AI initiatives deliver no measurable business impact, exposing a growing ‘GenAI Divide’ between successful and struggling adopters.

MIT Uses Generative AI to Design New Antibiotics for Drug-Resistant Superbugs

MIT scientists have used generative AI to design entirely new antibiotics that can kill two dangerous, drug-resistant bacteria, offering a glimpse into AI’s growing role in medical discovery.

Microsoft Unveils CLIO: Self-Adaptive Reasoning for Scientific Discovery

Microsoft’s new CLIO system offers scientists real-time control over AI reasoning—without additional training data or post-training intervention.

Anthropic Finds Longer AI Reasoning Can Hurt Model Performance

New research shows that giving language models more time to reason through problems may actually degrade their accuracy on key tasks.

72% of U.S. Teens Have Tried AI Companions, New Study Finds

New survey data reveals how teenagers are using AI chatbots for friendship, support, and social rehearsal.

Character.AI Unveils TalkingMachines: Real-Time AI Video With Voice-Driven Animation

Character.AI's new diffusion model, TalkingMachines, enables interactive, FaceTime-style avatars that talk, listen, and respond in real time.

LLMs Show Strategic Personalities in 140K Prisoner’s Dilemma Tests

New research reveals distinct “personalities” emerging in OpenAI, Google and Anthropic models—evidence of strategic reasoning beyond pattern matching.

Microsoft AI Diagnoses Complex Medical Cases with 85% Accuracy

New tool from Microsoft AI solves 85% of NEJM case challenges—four times better than experienced physicians

Claude the shopkeeper: What happens when AI runs a business?

Anthropic tested its model in a real-world store. What it got was a glimpse of AI's economic future—and its weird side effects.

Anthropic launches program to track AI's economic disruption

As job losses loom, the company seeks data-driven insight into labor impacts and policy responses.

Andy Konwinski Launches $100M Laude Institute for Public-Focused AI

Andy Konwinski launches the $100M Laude Institute to accelerate AI research with public benefit at its core, including a new lab at UC Berkeley.

How People Use Claude for Emotional Support, Advice, and Connection

DeepMind Unveils AlphaGenome, a New AI Model for Genomic Prediction

OpenAI Research: How Misaligned AI Behavior Spreads—and How to Stop It

Kids & Generative AI: New Insights on Use, Risks & Well‐Being

Self-Adapting AI: SEAL Enables Language Models to Update Themselves

Procurement Teams Face Growing Pressure to Adopt AI, Fairmarkit Finds