Latest News
View all
New benchmark exposes gap between chatbot fluency and real reasoning ability
Researchers say current leaderboard-topping models still fail basic multi-step logic tasks when questions are reworded.
AI coding-agent startup hits $4B valuation after 8x revenue growth
The two-year-old company says enterprise teams are shipping production code with minimal human review.
Next-gen AI accelerator chips promise 3x performance per watt
New silicon architecture targets the growing gap between model size and available training compute.
Warehouse robots now handle 1 in 5 fulfillment tasks at major retailers
Operators report a sharp drop in injury rates alongside faster average pick times this year.
Cloud providers report AI workloads now exceed 40% of total compute revenue
Quarterly earnings show inference costs falling even as overall usage climbs sharply.
Regulators propose mandatory disclosure for AI systems used in hiring
Draft rules would require companies to publish bias-testing results before deployment.
Open-weight model release closes the gap with closed frontier systems
Independent evaluations put the new release within striking distance of top proprietary models on coding tasks.
Interpretability team maps how large models represent abstract concepts internally
The findings could help engineers debug unwanted behaviors before models reach production.
Memory bandwidth, not compute, is now the bottleneck for large model training
Chip designers are racing to ship high-bandwidth memory stacks to keep GPUs fed with data.
