Story cards stay below the sticky dock while audio, chapters, date, and brief navigation remain accessible.
MiniMax released M2.7 as open weights on April 12, a sparse mixture-of-experts model that matches GPT-5.3-Codex on the SWE-Pro benchmark (56.22%) and scores 57.0% on Terminal Bench 2. Its headline feature is self-evolution: an internal version autonomously optimized its own programming scaffold over 100+ rounds, achieving a 30% performance gain without human intervention. The model is available on Hugging Face and Ollama.
Anthropic's Mythos Preview model, announced April 7 under Project Glasswing, continues to dominate the news cycle. The model identified thousands of zero-day vulnerabilities across every major OS and browser, including a 27-year-old flaw in OpenBSD. Anthropic is restricting access to a consortium including Amazon, Apple, Cisco, JPMorgan Chase, and Nvidia rather than releasing it publicly, marking a significant moment in voluntary AI capability restraint.
Alibaba's HappyHorse-1.0, quietly submitted to the Artificial Analysis leaderboard, claimed the top spot for both text-to-video and image-to-video generation, beating ByteDance's Dreamina Seedance 2.0 by 115 Elo points. The model generates synchronized video and audio from a single text prompt using a unified Transformer architecture. API access launches April 30 with full open-source release planned.
Anthropic's annualized revenue reached $30 billion, tripling from $9 billion at the end of 2025 and overtaking OpenAI's $25 billion ARR. Enterprise customers spending over $1 million annually doubled from 500 to 1,000 in under two months. Claude Code alone is generating a $2.5 billion run rate, more than doubling since January.
Following the completion of SpaceX's $1.25 trillion acquisition of xAI, the merged entity filed for what would be the largest IPO in history at a $1.75 trillion valuation. Musk's stated vision is orbital AI data centers — "Grok-Sats" — that bypass terrestrial power grid and cooling constraints by combining Starlink infrastructure with xAI's Grok models.
Sam Altman's San Francisco home was targeted a second time on April 12, just days after a Molotov cocktail attack on April 10 that led to the arrest of a 20-year-old Texas man. The escalating threats against AI leaders underscore growing public anxiety around AI's societal impact. Altman posted a photo of his family appealing for restraint.
