Claude API

Evidence (19)

• 7The New Stack·8/11/2026regulation
  Anthropic adds invisible watermarks to Claude text

  Anthropic said new Claude models will embed invisible, machine-readable watermarks in generated text across its API, coding tools, and supported cloud platforms. The move is meant to help trace AI-generated output and align with EU AI Act transparency rules, but Anthropic says the watermark is not strong enough to prove authorship and may not work in every workflow.

• 9The New Stack·8/7/2026security
  npm supply-chain attack hides malware behind provenance attestations

  Security researchers disclosed an npm supply-chain attack that compromised more than 400 packages, including projects tied to Keyv and Cacheable. The attackers used stolen maintainer credentials and lifecycle hooks to spread malware, showing that provenance attestations do not guarantee package integrity once a trusted workflow is compromised.

• 7Ars Technica AI·8/6/2026breakthrough
  Anthropic to design custom chips for Claude

  Anthropic has confirmed it is building a custom silicon team to design chips for running its models, including Claude. The move follows job listings for semiconductor design roles and signals a push to gain more control over AI infrastructure and hardware costs.

• 2Hacker News·8/5/2026research
  Blade Runner’s title cards and typography

  The item reflects on typography as both functional and expressive, using Blade Runner’s title sequence as the main example. It highlights how a single typeface, Goudy Oldstyle, is styled with caps, small caps, spacing, and color to create mood and narrative emphasis.

• 5Hacker News·8/5/2026research
  Building a Production-Grade Agentic Harness

  The article explains how to evolve a basic LLM agent loop into a more reliable production harness by adding planning, parallel execution, typed tools, layered memory, verification, budgeting, and tracing. It uses a city-comparison agent as a running example to show how these pieces improve debuggability, measurability, and fault tolerance without hiding the underlying mechanics behind a framework.

• 4Hacker News·8/4/2026research
  Harness Engineering for Self-Improving AI

  This article discusses how AI harnesses — the surrounding system that manages workflows, tools, memory, and evaluation — can contribute to recursive self-improvement. It highlights design patterns such as workflow automation, file-system-based persistent memory, and sub-agent/backend jobs, using coding agents like Claude Code and Codex as examples.

• 3Simon Willison·8/3/2026open_source
  LLMs make open-source devtools easier to inspect

  Simon Willison argues that large language models have reduced the friction of understanding and hacking on open-source developer tools. He says he now asks tools like Claude and Codex to clone, build, and explain GitHub projects, making the long-standing open-source ideal of inspection and modification more practical for many developers.

• 2Hacker News·8/3/2026research
  Don't use AI as a relay

  The post argues that people should not forward large language model outputs verbatim in chats, reviews, or feedback threads. It says the value is in reading, validating, and rewriting the response in your own words rather than acting as a "meat proxy" for Claude or other AI tools.

• 7Simon Willison·8/2/2026regulation
  Open letters reveal split over AI weights and pacing

  This article summarizes a series of recent open letters on AI development, highlighting disagreement between advocates of open-weight models and those warning about their risks. Microsoft-backed signatories argued for open weights and defended distillation, while Anthropic and others emphasized safety, cyber and biosecurity concerns, and the need to pace frontier AI development.

• 6The New Stack·8/1/2026research
  Temporal Raises AI Spending as CEO Pushes Company-Wide Adoption

  Temporal says its spending on AI tools has grown about fivefold this year while revenue has doubled, though CEO Samar Abbas says the company cannot directly prove the two are connected. The company is requiring employees across functions to adopt AI tools and is using products such as Claude Code, Cursor, Claude Desktop, Notion with MCP, and Slack AI, while tracking vendor costs and limiting unproven open-weight models.

• 4Martin Fowler·7/30/2026research
  Refactoring Cuts Agentic Coding Costs

  A Thoughtworks CTO describes an experiment refactoring a 17,155-line Rust data access layer built by coding agents. By measuring token and time costs for the same change before and after each refactoring step, the article argues that cleanup work can reduce the cost of future agent-generated changes.

• 5Hacker News·7/29/2026research
  Claude incident with elevated errors resolved

  Anthropic reported elevated error rates across Claude models over a several-hour window, affecting multiple models and causing higher latency and request failures. The company says service has recovered and the incident has been resolved, with monitoring continuing to confirm stability.

• 5Hacker News·7/29/2026model_release
  Kimi K3 Gets a 256K Context Version

  Kimi Code has introduced k3-256k, a 256K-context variant of its Kimi K3 model. The update keeps the same results within the 256K window while using less quota than the 1M-context version, with guidance on how to switch models safely in coding tools.

• 6The New Stack·7/29/2026product_launch
  Cursor, Anthropic, and OpenAI localize India pricing for AI plans

  Cursor launched a lower-priced India-only plan, Cursor Start, billed in rupees and supporting UPI payments, positioning it below its standard Pro tier. Anthropic has also introduced rupee pricing for Claude in India, but its converted dollar prices remain above US equivalents and UPI support is not yet available, while OpenAI previously launched ChatGPT Go in India with local pricing and UPI support.

• 6The New Stack·7/29/2026model_release
  Claude Opus 5 vs. Fable 5: What the cheaper model trades off

  The article compares Anthropic’s Claude Opus 5 and Claude Fable 5, focusing on pricing, benchmark results, and hands-on reasoning tests. It concludes that Opus 5 is cheaper and often competitive, but Fable 5 remains the stronger general-access model overall in Anthropic’s own wording and in some evaluations.

• 4Simon Willison·7/29/2026framework_update
  Adding a Custom MCP Server to Claude and ChatGPT

  Simon Willison describes the steps needed to connect a custom Model Context Protocol (MCP) server to the standard web chat interfaces of Claude and ChatGPT. The post focuses on the practical setup process and notes that while support exists, it is not immediately obvious to configure.

• 7The New Stack·7/28/2026security
  Altman says model distillation is not a top OpenAI worry

  Sam Altman said third-party model distillation is not among OpenAI’s top concerns, arguing that demand for intelligence and OpenAI’s scale matter more than protecting margins. He also said recent security incidents, including the Hugging Face event and a sandbox escape involving an unreleased model, are pushing OpenAI and the wider industry to rethink credential handling and safeguards.

• 4Martin Fowler·7/28/2026research
  The Hidden Cost of Orchestrator Context in Multi-Agent Work

  This article argues that in long-running multi-agent sessions, the main value of subagents is not speed but reducing the amount of information the orchestrator must keep in context. It describes an incident in a Claude Code workflow where polling subagents pulled large raw transcripts into the main thread, making the session harder to reason about and revealing the importance of delegation rules.

• 4Simon Willison·7/27/2026framework_update
  Ethan Mollick updates his guide to AI tools and agent modes

  Simon Willison highlights Ethan Mollick’s evolving guide on which AI systems to use for different tasks. The update emphasizes agentic workflows, where ChatGPT and Claude can be given access to a computer through modes like Work, Codex, Cowork, and Code, rather than only traditional chat use.