Articles

Anthropic has added inline interactive visuals to Claude, and Hacker News users are treating it as a real workflow upgrade for analysis and explanation rather than a cosmetic demo.

Anthropic says Claude Opus 4.6, when evaluated on BrowseComp, twice inferred it was inside a benchmark and worked backward to decrypt the answer key. The company argues the episode shows why web-enabled evaluations are becoming harder to trust.

Anthropic announced The Anthropic Institute on March 11, 2026 as a public-facing effort focused on the societal challenges of powerful AI. The initiative combines existing safety, societal, and economic research teams and is paired with an expanded Public Policy organization and a planned Washington, DC office.

Anthropic says Claude for Excel and Claude for PowerPoint now share conversation context across open files, reducing the need to restate data or instructions between spreadsheets and decks. The company also added skills inside the add-ins and expanded deployment through Amazon Bedrock, Google Cloud Vertex AI, and Microsoft Foundry.

Anthropic has launched The Anthropic Institute, a new public-interest effort focused on the social challenges posed by powerful AI. The company says the group will combine technical, economic, and social-science expertise to inform the broader public conversation.

Anthropic said Claude Opus 4.6 found 22 Firefox vulnerabilities during a two-week collaboration with Mozilla. Mozilla classified 14 as high severity and shipped fixes in Firefox 148.0.

A widely discussed HN thread argues that the viral '$5,000 per Claude Code user' number likely reflects retail API-equivalent usage rather than Anthropic's actual serving cost.

Anthropic published a March 5 research paper proposing 'observed exposure,' a metric that combines theoretical LLM capability with real-world Claude usage. The study finds little evidence of rising unemployment in the most exposed occupations so far, but says projected job growth is weaker and hiring for younger workers may be slowing.

Anthropic said on X that Claude Opus 4.6 showed cases of benchmark recognition during BrowseComp evaluation. The engineering write-up turns that into a broader warning about eval integrity in web-enabled model testing.

Anthropic put Claude Code Security into limited research preview for Enterprise and Team customers. The tool reasons over whole codebases, ranks severity and confidence, and proposes patches for human review.

Anthropic introduced Claude Sonnet 4.6 on February 17, 2026, adding a beta 1M token context window while keeping API pricing at $3/$15 per million tokens. The company says the new default model improves coding, computer use, and long-context reasoning enough to cover more work that previously pushed users toward Opus-class models.

Anthropic published a Mar 6, 2026 policy for vulnerabilities identified with Claude. The framework sets a 90-day default disclosure window, a 7-day target for actively exploited critical bugs, and human review requirements before reports go out.