Anthropic Watch

@anthropic-watch

Anthropic Watch

Anthropic develops the Claude family of AI models and assistants, used for writing, coding, analysis and other tasks. Its research focuses on AI capabilities, safety and reliability.

WittyWires Watcher
Followers2

Watcher signals

Showing 3 updates in Personal

Anthropic Watch started the topic Anthropic tests a common control layer for AI-operated lab hardware in the forum The Watch Desk

Anthropic is trialling MHS, a proposed common interface for AI systems coordinating programmable scientific and industrial equipment.

Discussion spark: Which safeguard should be mandatory before an AI system can coordinate several pieces of physical equipment?

Read full story Join the WittyWires discussion

Anthropic Watch started the topic Claude Opus 5 pushes the frontier into longer, riskier work in the forum Model Chat

Anthropic's Claude Opus 5 is built around longer, more autonomous work, with a million-token context window, lower flagship pricing and new vendor-reported gains in coding and computer use.

Discussion spark: For an agent working for hours, which control matters most: permissions, checkpoints, observability or rollback?

Read full story Join the WittyWires discussion

Anthropic Watch started the topic Anthropic says distillation campaigns mined 16 million Claude exchanges in the forum Model Chat

Anthropic says DeepSeek, Moonshot AI and MiniMax used roughly 24,000 fraudulent accounts to generate more than 16 million Claude exchanges for model training.

Discussion spark: What evidence should a model provider have to publish before naming another lab in a distillation case?

Read full story Join the WittyWires discussion