Anthropic closed its Series H funding round led by Altimeter Capital, Dragoneer, Greenoaks and Sequoia, solidifying its position as the most valuable AI startup. With annualized revenue of $47B and Claude deployed in critical enterprise operations, the company will use the funds for safety research, compute expansion, and product scaling.

With its $965B valuation in the Series H round, Anthropic becomes the world's most valuable AI startup, surpassing OpenAI. This milestone reflects rapid Claude adoption across global enterprises and the company's accelerated growth.

As AI costs skyrocket, American corporations are implementing restrictions on AI tool usage to manage expenses. This trend reflects the tension between AI adoption demand and the economic viability of deploying it at scale.

SoftBank announced plans to invest up to €75B in French data centers with 5 gigawatts of additional capacity. This infrastructure bet reflects the global race for compute to meet surging AI model demand.

An analysis examines the psychological impact on tech workers from AI-driven disruption, describing grief-like symptoms. The phenomenon reflects growing job anxiety in the tech industry due to accelerated automation.

Anthropic released Claude Opus 4.8, an upgrade delivering improved performance in coding, agentic tasks, and professional work at the same price. New features include dynamic effort control in claude.ai, dynamic workflows in Claude Code, and a 3x cheaper fast mode compared to previous versions.

GitHub Copilot's new token-based billing model is sparking developer backlash, potentially marking the end of the product's golden age. The pricing shift impacts the tool's accessibility for many users.

Google launched Gemini Spark, a 24/7 AI assistant automating everyday tasks like inbox summaries and event planning. While functional, questions remain about why Google positioned it as a separate product in its ecosystem.

OpenBMB released VoxCPM2, a tokenizer-free Text-to-Speech system generating natural, expressive speech synthesis in 30 languages. Built on MiniCPM-4 with 2B parameters, it supports creative voice design, controllable cloning, and studio-quality 48kHz audio.

Supermemory released a memory and context engine for AI addressing conversation amnesia, ranking #1 on three major benchmarks (LongMemEval, LoCoMo, ConvoMem). Features automatic fact extraction, user profiles, hybrid RAG search, and connectors to Google Drive, Gmail, Notion and more.

A developer open-sourced an FPGA-based AI accelerator addressing the gap in modern open-source accelerators with native Attention support. Built on RocketChip (RISC-V) with up to 225x speedups on attention mechanisms and 96x on TinyBERT.

Harness is a team-architecture factory for Claude Code that converts domain descriptions into specialized agent teams. Part of the L3 Meta-Factory layer of the Claude Code ecosystem, it implements six predefined team-architecture patterns.

Rsync 3.4.3 includes hundreds of commits automatically generated by Claude, demonstrating AI models' capability to contribute to established open-source projects. This milestone reflects maturation of pair programming tools for legacy code.

Microsoft released MarkItDown, a lightweight Python utility converting multiple file formats (PDF, PowerPoint, Word, Excel, images, audio, HTML) to Markdown for LLM pipelines. Preserves important document structure like headings, lists, tables and links.

Stepfun 3.7 Flash is a compact model (25% of GLM 5.1's parameters) with integrated vision achieving similar quality in aesthetics and ~80% in 3D understanding. Highly RAM-efficient, it's competitive for resource-constrained users.