Princeton researchers introduced SCION, a hierarchical compositional scene representation that replaces millions of independent 3D Gaussians with a compact vocabulary of reusable primitives and lightw…
A review of recent research on recursive self-improvement (RSI) concludes that while large language models have made remarkable advances in scientific tasks, every demonstrated success relied on an ex…
Epoch AI launched Benchmark Reviews on September 17th and classified nine of its first 15 reviewed benchmark versions as Flawed, four as Verified, and two as Not Enough Info, flagging defects in tests…
A 2026 Nature study found Chinese state-coordinated media embedded in major AI training datasets, and when researchers added more of that material while training an open-weight model, its answers beca…
A historical survey traces the first physical and computational reinforcement learning machines built between 1948 and 1954, beginning with Norbert Wiener's 1948 Cybernetics and its negative-feedback …
OpenAI deployed roughly 10,000 AI agents simultaneously in August 2026 and solved the Navier-Stokes Millennium Prize Problem in 88 hours, according to The Verge. The result highlights a broader shift …
Meta is offering an average 95% discount on its Muse Spark model for users who share their prompts and outputs to help train future versions, with input tokens priced at $0.10 per million and output t…
Fei-Fei Li, co-founder of World Labs, has built a company that reached a $1 billion valuation within roughly twelve weeks of emerging from stealth in September 2024, after raising $230 million initial…
Foreign student applications to US colleges fell about 10% this year, the biggest decline in at least a decade, according to data released Thursday by Common App, the country's most widely used colleg…
A study conducted with Princeton and the UK AI Security Institute found that AI agents using Claude Opus 4.8 and GPT-5.6 Sol, given six days and $3,000 in API credits, produced research papers that or…
A study by Severin Field, a Visiting Fellow at the Institute for AI Policy and Strategy, found that 20 of 25 AI researchers interviewed from OpenAI, Anthropic, Google DeepMind, Meta, Princeton, UC Ber…
On July 7, 2026, Anthropic added Claude Cowork sessions to web and mobile, allowing users to start, steer, review, and resume the same session across Claude Desktop, web, and mobile, with sessions att…
Tokenless (YC S26) launched a drop-in API router that automatically selects the most cost-effective AI model for each request, claiming to cut inference costs by up to 57% while maintaining quality. T…
Yahoo Finance's product chief George Leimer says AI is collapsing the 'execute' layer of work while expanding the 'decide' and 'deliver' layers, a phenomenon Princeton computer scientist Arvind Naraya…
Twin sisters Danielle and Leah Cohen-Shohet, Princeton graduates and former Goldman Sachs employees, raised $44 million at a $1.15 billion valuation for their wellness company Genius AI, making it a u…
Only 5% of enterprises trust AI agents in production despite 85% piloting them, Amazon AGI lab's Bryan Silverthorn said at VB Transform 2026, urging a shift from benchmarks to reliability metrics like…
A developer completed an 8-year "100 Days of Algorithms" challenge, implementing algorithms by hand without LLM assistance. The project, now reviewed by an AI, contains broken implementations and stub…
Brown University economics professor Roberto Serrano switched to handwritten final exams after suspecting AI-assisted cheating, causing enrollment in his ECON 1170 course to plummet from 86 to a fract…
AI labs like OpenAI and Anthropic are moving up the technology stack to escape the commodity trap of undifferentiated model inference, pursuing vertical integration and enterprise lock-in strategies t…
A man walked into the White House in 2023 with an AI chatbot-generated recipe for a pandemic, and by 2026 the New York Times reported that leading AI models like ChatGPT, Gemini, and Claude can now gu…