Claude Opus 5.5 tops writing benchmark with 2631 Elo

The global alliance launches first in Canada, pairing Cohere's North agentic platform and enterprise models with PwC's risk, regulatory and transformation expertise. Deployments span private cloud, on-premises and air-gapped environments for enterprise search, deep research and task automation.

A LocalLLaMA user reports that the strata-swift-iq3_xxs quantization, run through the Strata tool, delivers 2-3x performance versus Qwen3.8-27b-iq4xs variants but randomly injects unrelated information into its reasoning.

The Algorithmic Bridge essay contends Anthropic "might be screwed" because exponentials always hit a binding constraint. The piece frames the AI industry's expectation of endless growth in model size, benchmark scores, and intelligence as unsustainable.

Nadella succeeded Steve Ballmer in 2014 and turned Microsoft into a cloud giant; the piece examines whether he can repeat that reinvention in AI.

Sony Music's takedown notices for AI-generated deepfakes of its artists passed 260k, nearly double the 135k it reported in March 2026 and up from 75k in March 2025. The label told the FT it wants more policymaker support, calling enforcement "an uphill struggle."

Bruce Schneier reports another historic cipher has fallen to AI, this one from 1809 and written by Napoleon's nephew. It follows Claude Fable 5.1 solving a 370-year-old cipher in forty-four minutes, which Schneier attributes to AI strength at search-and-test problems.

a16z's seventh Top 100 Consumer AI Apps edition adds a new dimension: what consumers actually pay for. Editorial partner Elena Burger discusses the data with investing partners Olivia Moore and Josh Elman, who describe a striking power-user economy.