/
Published on 27.02.2026
TLDR: Google launched Gemini 3.1 Pro Preview, a mixture-of-experts transformer that topped the Artificial Analysis Intelligence Index and achieved state-of-the-art results on multiple benchmarks including ARC-AGI-2, GPQA Diamond, and Humanity's Last Exam. The kicker: it did so at the same price as its predecessor, suggesting model quality improvements rather than brute-force inference spending.
Link: Google launches Gemini 3.1 Pro Preview
TLDR: The fourth global AI summit in New Delhi marked a decisive shift from obsessing over theoretical AI catastrophes to focusing on spreading AI's practical benefits worldwide. More than 85 countries endorsed the non-binding New Delhi Declaration, while major tech companies expanded their presences in India with billions in commitments.
Link: AI Impact Summit - New Delhi
TLDR: The S&P Software and Services Index lost 25 percent of its value after Anthropic launched Claude Cowork and a series of open-source plugins targeting enterprise work functions. The panic reflects a genuine but overstated fear that AI agents will replicate or bypass traditional SaaS products, though Anthropic's subsequent partnership announcements stabilized the market somewhat.
Link: Did AI Really Kill SaaS?
TLDR: Stanford and Together AI researchers introduced the metric "intelligence per watt" to compare local versus cloud AI systems. While cloud still wins on efficiency, local systems have improved 5.3 times between 2023 and 2025, and can already handle about 89 percent of queries correctly while consuming dramatically less power in hybrid scenarios.