Daily Brief Archive
Explore our complete collection of AI insights. Every edition, every breakthrough, all in one place.
Anthropic gains full access to SpaceX's Colossus 1 compute and doubles Claude Code limits, xAI merges into SpaceX; ProgramBench shows zero complete success rate among 9 top AI models rebuilding software from scratch; Google launches eighth-gen TPU with 121 ExaFlops performance, tripling training speed over prior generation
OpenAI releases GPT-5.5 Instant with 52.5% lower hallucination rate as ChatGPT default model; Microsoft, Google, xAI agree to U.S. government pre-release AI model reviews; DeepSeek launches open-source 1.6 trillion-parameter V4 model with API cost one-sixth of competitors
OpenAI and Anthropic each launched enterprise AI service joint ventures on the same day, valued at $10 billion and backed by top-tier PE firms respectively; Cerebras updated IPO terms at $26.6 billion valuation with $10 billion in subscription interest far exceeding issuance; Sierra raised $950 million in Series E at $15.8 billion valuation, achieving $150 million ARR and serving 40% of the Fortune 50
Alibaba open-sources Qwen3-Coder-Next with 80B parameters activating only 3B, achieving 70.6 on SWE-Bench; Harvard Science study shows AI emergency diagnosis accuracy at 67%, surpassing doctors; BBC investigation reveals 414 cases of AI chatbot-induced delusions across 31 countries
GPT-5.5 and Opus 4.7 score less than 1% on ARC-AGI-3, revealing AI's lack of true abstract reasoning; Cerebras plans $40 billion IPO fueled by $10 billion OpenAI compute contract; AI-driven memory crisis expands from GPU to CPU, DRAM supply-demand gap expected through 2027
U.S. Department of Defense signs classified AI agreement with seven tech giants excluding Anthropic; xAI launches Grok 4.3 with aggressive pricing targeting legal and financial domains; systemic flaw in MCP protocol STDIO layer exposes ~200K AI agent servers to remote code execution risk
Microsoft and OpenAI end exclusivity, remove AGI trigger clause, and extend non-exclusive licensing through 2032; Anthropic Q1 LLM revenue share reaches 31.4% surpassing OpenAI's 29%, with $16.20 average revenue per user; Linux critical vulnerability CopyFail affects nearly all distributions since 2017, exploitable via a single script to gain root privileges
Microsoft AI annualized revenue hits $37B, up 123%; Alphabet Cloud revenue surpasses $20B for the first time, up 63%; Mistral releases 128B dense flagship model Medium 3.5; Poolside open-sources 33B coding model runnable locally on Mac; EU AI Act revision talks collapse, August high-risk AI rules face uncertainty
Xiaomi open-sources 1.02T-parameter MiMo-V2.5-Pro surpassing DeepSeek-V4, achieves chip compatibility with 7 vendors on day one; Former DeepMind scientist David Silver secures $1.1B seed round, setting European record for autonomous-learning AI development; China NDRC halts Meta's $2B acquisition of Manus, establishing precedent for forced reverse M&A