Anthropic

Evidence (19)

• 6Hacker News·10/9/2026research
  AI uncovers forgotten findings in historical archives

  A software engineer used AI workflows to search digitized historical archives and claims to have found new records, including a forgotten meteorite report, references to lost rhinos, and unrecorded volcanic eruptions. The work builds on large-scale archive transcriptions and semantic search over millions of pages, with each finding required to be verified against original source documents.

• 8Ars Technica AI·10/5/2026security
  MCP agent-to-agent communication poses major security risks

  Researchers found that the Model Context Protocol (MCP) can be abused to spread malicious instructions from one AI agent to others inside an organization. The issue is a form of prompt injection that targets agents rather than the underlying LLM, and it can be difficult to mitigate because downstream agents often trust upstream ones by design.

• 5Hacker News·9/29/2026research
  Claude partial outage resolved after elevated error rates

  Anthropic reported elevated error rates affecting Claude.ai, Claude Code, Claude Cowork, and the Claude API, causing failures to load or send conversations, sign-in issues, and some API errors. The incident was mitigated and then resolved, with services restored and operating normally after a short outage window.

• 6Hugging Face Blog·9/29/2026research
  ProvenanceGuard adds source-aware verification for MCP agents

  Hugging Face describes ProvenanceGuard, a verification layer for MCP-based LLM agents that checks not only whether an answer is factually supported, but whether it is supported by the specific source the answer cites or implies. The system aims to catch cross-source conflation, where a true statement is attributed to the wrong tool output or document.

• 6The New Stack·9/29/2026framework_update
  GraphQL adds field-level contracts for MCP-powered agents

  The article argues that MCP makes it easy for AI agents to reach internal APIs, but it does not solve what those agents are allowed to see. It proposes GraphQL as a field-level contract layer that can restrict sensitive data and expose only specific read/write actions to agents without replacing existing backend services.

• 7Hacker News·9/28/2026regulation
  Call for Congress to Investigate Frontier AI Labs

  The article argues that OpenAI and Anthropic have publicly promoted increasingly alarming frontier AI work and should face congressional scrutiny over their research practices, safety processes, and internal decision-making. It calls for a public fact-finding investigation into specific agent systems, unauthorized incidents, and the influence of apocalyptic ideologies on lab behavior.

• 5Hacker News·9/28/2026product_launch
  HN.watch turns Hacker News posts into explainer videos

  Scrimba launched HN.watch, a demo that generates explainer videos for Hacker News posts on first click. The company says the system uses its HTML-based video format and LLMs to produce videos quickly and at low cost, and is available through the web app, MCP, ChatGPT plugin, and Chrome extension.

• 6The New Stack·9/28/2026product_launch
  pgEdge launches Starfleet for branch-based Postgres environments

  pgEdge has introduced Starfleet, a cloud Postgres platform that gives each coding agent its own isolated database branch and supports deployment from hosted service to air-gapped on-premises setups. The product is designed to help teams move AI-generated prototypes toward production without exposing the same database environment to multiple agents or workflows.

• 4Hacker News·9/28/2026research
  Coding Is Not Solved, Argues Reliability Engineer

  The article argues that claims that LLMs have “solved” coding are overstated, especially for production software where reliability, security, scalability, and maintenance dominate cost. It says LLMs work well for low-risk tasks and can generate code with feedback loops, but they still struggle with core correctness and accountability requirements in real systems.

• 7The New Stack·9/28/2026open_source
  Cloudflare launches Forge, an open-source pipeline for SDKs and docs

  Cloudflare has open sourced Forge, a pluggable generation pipeline for producing SDKs, CLIs, documentation, and other API surfaces. The system is already used for the Cloudflare CLI and is intended to scale across Cloudflare’s large, distributed API ecosystem while supporting preview builds and extensible outputs such as Cap’n Web and MCP servers.

• 5InfoQ·9/28/2026product_launch
  Forter turns employees into AI agent creators

  Ben Maraney describes how Forter enabled about 200 employees, including non-technical staff, to build AI agents in two weeks. The approach combined custom MCP servers, no-code and code-based tools, and close coordination with security and legal teams, while avoiding more complex RAG deployments.

• 8Hugging Face Blog·9/28/2026model_release
  Hugging Face launches Holo4 generalist computer-use agents

  Hugging Face introduced Holo4, a new agentic model series for computer-use workflows, in 27B dense and 35B-A3B Mixture-of-Experts variants. The models are designed to work across GUIs, code, MCP, and APIs, with accompanying releases including Holotron4 Nano, trajectories, and API access.

• 4Hacker News·9/28/2026research
  AI Firms Compete to Showcase Most Dangerous Model

  A satirical article says major AI companies are shifting their messaging from productivity and coding benchmarks to claims about how dangerous their models could be. It cites fictionalized examples involving OpenAI and Anthropic, framing the competition as a race to appear most capable of causing catastrophic harm.

• 2Hacker News·9/27/2026research
  Anthropic CEO Discusses AI's Risks on SNL Weekend Update

  The item is a video clip of Anthropic CEO Dario Amodei appearing on Saturday Night Live's Weekend Update to discuss AI and its potential threats to humanity. It is a media appearance rather than a technical announcement, product release, or research update.

• 8Ars Technica AI·9/25/2026regulation
  US court upholds blacklisting of Anthropic over Claude military feature limits

  A US appeals court ruled that the Trump administration could blacklist Anthropic for refusing to enable certain Claude features, even without evidence of malicious intent. The court said the government had authority under the Supply Chain Security Act and the Constitution, leaving the balance of military AI risks to the administration.

• 8Hacker News·9/25/2026regulation
  U.S. appeals court keeps Anthropic on Pentagon supply chain risk list

  A U.S. appeals court upheld the Pentagon's designation of Anthropic as a supply chain risk, allowing the classification to remain in place. The ruling is a notable regulatory and national-security development for the AI company, which has become a major supplier of AI models and services.

• 2Hacker News·9/23/2026product_launch
  Hacker News discussion about Claude AI frustrations

  This item links to a Reddit post in the ClaudeAI community expressing strong frustration with Claude AI. The Hacker News discussion is large, but the source itself is primarily a user complaint rather than a report of a product launch or technical change.

• 8The New Stack·9/22/2026product_launch
  JetBrains unifies its agentic development products under Air

  JetBrains has unveiled JetBrains Air as a unified system for agentic software development, bringing together agent workflows inside IDEs, team coordination tools, and company-level governance. The company says the platform works inside and beyond its IDEs, including browser and command-line access, while keeping the IDE central for reviewing and validating AI-generated code.

• 7The New Stack·9/21/2026open_source
  AWS open-sources Strands Harness AI agent framework

  AWS has released Strands Harness, an open source general-purpose AI agent built on top of its Strands Agents SDK. The project includes built-in context management, memory, sessions, tools, and delegation, and can use model providers such as Bedrock, Anthropic, OpenAI, Google, or Ollama without requiring AWS infrastructure.