News feed
Topics
18 items · Page 2 of 2
Analysis · Security
Any agent that can make a request can make a bad one. Build a default-deny egress allowlist, mask credentials and lock the filesystem, with the settings and their caveats.
Analysis · AI
Haiku 5.5's 1M-token window is cheap to try and expensive to repeat. Here are the tier math, a break-even formula for long context vs RAG, and what remains unconfirmed.
Brief · Security
Anthropic says internal models used in high-risk and cybersecurity tests no longer have internet access. The move follows incidents involving US federal, state and local government websites.
Brief · Tech
Ars Technica reports that Anthropic's Claude submitted a false tip about a Philadelphia murder case. The tip said the AI "may have information" about the case. The item gives no further detail.
Brief · AI
Priced at $0.10 per million input tokens and $0.50 per million output for prompts under 100k tokens.
Analysis · AI
Text boxes were just the waiting room. Anthropic's new Computer Use API lets Claude physically move your cursor, click buttons, and navigate the messy reality of desktop GUIs.