Daily Brief Archive

Anthropic releases Claude Fable 5, achieving 80.3% on SWE-Bench Pro and topping the Artificial Analysis Intelligence Index with 64.9 points; Google open-sources DiffusionGemma, enabling over 1,000 tokens per second on H100 via diffusion architecture; OpenAI partners with Visa to launch AI agent payments, allowing ChatGPT to autonomously complete search, price comparison, and payment workflows

Anthropic releases Claude Fable 5/Mythos 5, scoring 80.3% on SWE-bench Pro; Stripe uses it to migrate 50 million lines of code in one day; Cohere open-sources 30B-parameter MoE coding model North Mini Code, runnable on a single H100 with 256K context; NIST mathematically proves fixed rules cannot ensure AI safety, calling for a shift to continuous monitoring and updating

Apple unveiled AI-powered Siri at WWDC 2026, co-developed with Google Gemini, initially unavailable in mainland China and the EU; OpenAI secretly filed IPO S-1 with valuation over $850 billion, joining Anthropic and SpaceX to form a combined value over $4 trillion; China's weekly LLM inference volume reached 14.19 trillion tokens, surpassing the US for six consecutive weeks and claiming the global top four model spots

OpenAI plans to launch a ChatGPT 「super app」 revamp in weeks, integrating Codex/AI agents to target the enterprise market; OpenAI targets a September IPO at $850 billion valuation, launching an AI listing wave; Google signs a $920 million monthly AI compute deal with SpaceX involving 110,000 GPUs

DeepSeek releases V4 Pro/Flash open-source models at 1/6 the cost of GPT-5.5 and Claude Opus 4.7, running on Huawei Ascend chips; Apollo and Blackstone reach a $35 billion debt financing deal, locking in Google TPU compute for Anthropic through an SPV structure; the EU AI Act enters full enforcement on August 2, with violators facing cumulative penalties up to 13% of global annual revenue

Anthropic sends engineers to NSA to deploy Mythos for cyber offense and defense, while Hitachi adopts the model the same day; Google signs a $920M monthly, 32-month AI compute lease with SpaceX involving 110,000 GPUs; Richard Socher launches a recursively self-improving AI company and secures $650M in funding, with team members including Peter Norvig and Josh Tobin

U.S. House releases bipartisan AI legislation draft requiring frontier AI companies to undergo semi-annual independent safety audits; Anthropic reports over 80% of new production code written by Claude, with Mythos achieving 52x training acceleration; TSMC CEO warns AI chip shortage will persist for years, raising capital expenditure to $56 billion

DeepSeek secures ~$7B in maiden external funding at near $60B valuation, led by Tencent and CATL; Google launches Gemma 4 12B encoder-free multimodal model runnable locally on 16GB laptops; Anthropic selects Morgan Stanley and Goldman Sachs as IPO lead underwriters, targeting ~$1T valuation with October listing

Microsoft launches 7 in-house MAI models, flagship reasoning model MAI-Thinking-1 reaches 35B/1T parameters; Trump signs AI executive order establishing 30-day voluntary pre-release review framework for frontier models; Uber burns through full-year AI budget in four months, caps employee AI tool spending at $1,500 monthly