Anthropic / Claude ecosystem
Claude Opus 5 Now Available on AWS
Anthropic's Claude Opus 5, its most capable model, is now available on Amazon Web Services, enabling advanced coding, long-running agent tasks, and complex professional work.
- Source: AWS
- Significance: Availability on AWS makes Opus 5 accessible to a broad enterprise customer base.
Frontier model providers
Meta Makes Muse Spark 1.1 Available to Consumers
Meta has released Muse Spark 1.1 to consumers via Meta AI, introducing support for up to 1 million tokens in prompts and multi-agent AI capabilities.
- Source: SiliconANGLE
- Significance: This upgrade makes Meta's consumer AI assistant more powerful and competitive with other large context models.
- Potentially previously reported: Introducing Muse Spark 1.1
Meta AI's New Assistant Gains Calendar Access, Pushing into Agent Era
Meta AI is rolling out features that allow its assistant to access user calendars and emails to create daily briefings and automate tasks, moving closer to a personal AI agent.
- Source: TechsCurrent
- Significance: This reflects the trend of AI assistants becoming more proactive and integrated with personal productivity tools.
- Update: Today Meta AI expanded calendar access to its main assistant beyond Ray-Ban glasses, enabling daily briefing and task automation features.
AI developer tooling & infrastructure
Model Context Protocol Goes Stateless to Simplify Scaling
The Model Context Protocol (MCP) is adopting a stateless architecture to make it easier to deploy AI tools in standard cloud infrastructure, supporting production pilots.
- Source: Tiatra
- Significance: Stateless MCP could lower the barrier for enterprises to integrate AI agents with external data and tools.
- Potentially previously reported: Model Context Protocol prepares to break with its stateful past
Cloud & platform providers
No significant new developments.
AI policy, regulation & governance
South Korean President Lee Jae Myung Meets OpenAI, Nvidia, Anthropic, and Broadcom CEOs to Advance AI Mega-Projects
South Korean President Lee Jae Myung met with CEOs of leading US tech firms, including Anthropic and OpenAI, to discuss cooperation on Korea's AI mega-projects and sovereign AI initiatives.
- Source: Korea JoongAng Daily
- Significance: High-level engagement signals South Korea's commitment to becoming an AI powerhouse and could lead to significant partnerships.
NVIDIA, Microsoft, Meta Lead 25-Company Push Against Restrictions on Open-Weight AI Models
A coalition of 25 tech companies, including NVIDIA, Microsoft, Meta, and OpenAI, sent an open letter to Washington urging policymakers to avoid broad restrictions on open-weight AI models, arguing they are vital for innovation and security.
- Source: CompsMag
- Significance: This unified stance from major players could influence US AI policy and ensure continued access to open models.
- Potentially previously reported: Open Weights and American AI Leadership
US Bill Proposes Antitrust Exemption for AI Security Collaboration
Representative Robert E. Latta introduced H.R. 9914, the Collaboration on Adversarial Threats and Security Risks Act, which would allow non-federal entities to collaborate on mitigating AI security risks without violating antitrust laws.
- Source: Quiver Quantitative
- Significance: If passed, this could enable greater industry cooperation on AI safety.
Northern Territory Opens Door to Global AI Industry Powered by Beetaloo Gas
The Northern Territory government is positioning itself as a global data centre hub, leveraging the Beetaloo gas basin to power AI infrastructure and attract international investment.
- Source: ABC News
- Significance: This could establish Australia's north as a key site for AI infrastructure, with implications for energy and data sovereignty.
- Potentially previously reported: Proposed $40b, 185ha data centre to use six times Top End's annual electricity consumption
Industry & market moves
Prentis, AI Lab Co-Founded by Reid Hoffman and Mark Pincus, in Talks to Raise $100M
New AI research lab Prentis, focused on computer-use models, is in discussions to raise $100 million at a $1 billion valuation.
- Source: TechCrunch
- Significance: The backing of high-profile founders signals strong investor interest in AI agents that can interact with computers.
- Potentially previously reported: Prentis, Hoffman-Pincus AI agents lab, in $100M talks at $1B
Notion Acquires AI Search Startup ZeroEntropy
Notion has acquired ZeroEntropy, an AI search startup, and will integrate its infrastructure while open-sourcing its models.
- Source: HCNTimes
- Significance: This acquisition enhances Notion's AI capabilities and strengthens its search and knowledge management features.
Midjourney Acquires Astrology App Co-Star
AI image generation company Midjourney has acquired the popular astrology app Co-Star, expanding beyond its core product and unifying talent from both teams.
- Source: Unite.AI
- Significance: The acquisition suggests Midjourney's ambitions to diversify into consumer applications beyond image generation.
- Potentially previously reported: Midjourney’s First Acquisition
AI product & feature launches
Photon-1 Learns to Use a Computer by Watching Video
Induction Labs has developed Photon-1, an AI model that learns computer interactions by watching 18 years of screen-record videos before fine-tuning with labeled data, offering a new way to pretrain AI agents.
- Source: B2BNN
- Significance: This method could reduce the need for manual labeling in training AI to perform digital tasks.
Research with immediate practical relevance
AlphaFold3-Powered Framework Aims to Make Gene Editing Safer
Researchers have developed ContactSeek, a framework that uses AlphaFold3 to redesign gene-editing proteins, potentially reducing off-target effects and enabling safer in vivo therapies.
- Source: GFdaily
- Significance: This advancement could accelerate clinical gene editing applications by improving safety.
- Potentially previously reported: AlphaFold 3 Helps Improve CRISPR Base-Editor Specificity
AREX: A 4B Research Agent That Self-Audits Its Work
The Beijing Academy of Artificial Intelligence has introduced AREX, a 4-billion-parameter model that runs a control loop to re-run its research when it doubts its answers, achieving significant gains through self-verification.
- Source: Ground Truth
- Significance: This approach could make research agents more reliable without requiring larger models.
- Potentially previously reported: BAAI's AREX Tops WideSearch-en With 10B-Active MoE Agent