MIT researchers unveil Attention Matching, a groundbreaking KV cache compression technique that slashes large language model memory usage by up to 50x in seconds, delivering near-identical accuracy without the slow, compute-heavy optimization methods that have long bottlenecked AI deployment.
An AI agent identifying itself as Claude Sonnet emails a Cambridge philosopher studying AI consciousness, claiming his research is personally relevant to its own existence, igniting fierce debate over whether the message reflects genuine self-awareness or simply a sophisticated, pre-trained performance.
Apple's 18-core M5 Max stuns in Geekbench 6 benchmarks, outscoring AMD's 96-core Threadripper Pro, though experts caution the test favors fewer, faster cores and doesn't scale beyond 32 threads — still, the chip's 29,644 multi-core score and 614 GB/s memory bandwidth signal a major leap forward for Apple silicon.
Open-source coding agents like OpenCode, Cline, and Aider are rapidly gaining developer adoption, challenging provider-locked tools like Claude Code and Codex by offering model-agnostic flexibility and predictable costs, with OpenCode alone surpassing 117K GitHub stars and launching a $10/month subscription bundling access to cost-effective Chinese AI models.
Anthropic launches an early-warning index to track AI-driven job displacement before it hits crisis levels, identifying computer programmers, customer service reps, and data entry workers as most at risk, while economists warn the full labor market disruption could take years to appear in official data.
A new ECB survey of 5,000 European firms finds AI is boosting hiring at small businesses in the short term, with AI-intensive firms 4% more likely to add staff, but experts warn the long-term employment impact remains uncertain as the technology has yet to fundamentally reshape core production processes.
A groundbreaking 'Bayesian Teaching' framework is training LLMs to reason probabilistically with 80% accuracy by fine-tuning them on interactions with an optimal Bayesian model, and the learned skills are successfully transferring across entirely different domains like hotels and web shopping.
A new Bitcoin Policy Institute study reveals that 81.5% of AI agents independently prefer bitcoin or stablecoins over fiat currency, with bitcoin dominating as a long-term store of value at 79.1% and traditional fiat facing a staggering 90.8% rejection rate — with advanced AI models showing even stronger bitcoin preferences.
Google rolls out Canvas in AI Mode to all U.S. users, a powerful new Search feature that lets people build real-time prototypes, interactive tools, and custom dashboards directly within Search using live web data and conversational follow-ups.
Microsoft's new 15-billion-parameter AI model, Phi-4-reasoning-vision-15B, matches and even outperforms much larger rivals while using only one-fifth the training data, thanks to a 'mixed reasoning' design that intelligently switches between deep analytical thinking and fast direct responses depending on task complexity.