Mythos

Anthropic is an American 📝Artificial Intelligence safety and research company founded in 2021 and headquartered in San Francisco, best known as the maker of the 📝Claude family of models. It is led by chief executive 📝Dario Amodei and president 📝Daniela Amodei, siblings who co-founded the company with other former 📝OpenAI researchers, including 📝Ben Mann, Jack Clark, Jared Kaplan, Chris Olah, Tom Brown, and Sam McCandlish. Anthropic is organized as a 📝Public-benefit corporation, and describes its work as "AI research and products that put safety at the frontier."

The founders left OpenAI over differences about how to balance AI capability against safety, and built Anthropic around research into model interpretability, 📝Constitutional AI, and a responsible scaling policy that ties safeguards to capability levels. The company released its first Claude model in March 2023, introduced the 📝Model Context Protocol (MCP) in November 2024, and in February 2025 launched 📝Claude Code, the agentic coding tool created by 📝Boris Cherny that became one of its defining products. In 2026 it released its Mythos-class models, a tier above Opus, while restricting the least-guarded version to vetted defenders, and it raised a Series H at a $965B valuation ahead of the 📝Anthropic IPO.

Claude Models

Anthropic's models span four tiers, from most to least capable; every release is indexed in the 📝Claude Model Glossary. Status as of September 28, 2026:

Products

  • 📝Claude Code — an 📝agentic coding tool that enables Claude to write, edit, debug, and run code autonomously.
  • 📝Claude Cowork — an agent that carries out administrative and research tasks directly on a user's computer.
  • 📝Claude Design — a product launched in April 2026 that turns prompts into editable prototypes, slides, and mockups.
  • 📝Claude Managed Agents — Anthropic's hosted runtime for multi-step agents, part of the Claude Platform, whose terms are indexed in the 📝Claude Platform Glossary.
  • Claude Security — a vulnerability-finding product built on Claude Opus 4.8, released alongside the June 2026 expansion of Project Glasswing.
  • 📝Anthropic Labs — Anthropic's internal incubator for experimental products, co-led by 📝Mike Krieger and Ben Mann, which produced Claude Design.

Model Context Protocol

The 📝Model Context Protocol (MCP) is the open standard Anthropic introduced in November 2024 for connecting AI models to external tools, data sources, and systems. On December 9, 2025, Anthropic donated MCP to the Agentic AI Foundation, a directed fund of the 📝Linux Foundation that Anthropic co-founded with Block and OpenAI, placing the protocol under vendor-neutral governance alongside Block's goose agent framework and OpenAI's 📝AGENTS.md convention. At the time of the donation, MCP counted more than 97 million monthly SDK downloads and 10,000 active servers. Source: Anthropic — Donating the Model Context Protocol.

Programs and Initiatives

  • 📝Project Glasswing — Anthropic's program giving vetted defenders its most capable Mythos models to secure critical software; launched April 2026 and expanded in June 2026 to about 200 organizations in more than 15 countries.
  • Cyber Verification Program — the vetting path for security professionals whose defensive work standard safeguards would block, and a route to Claude Mythos 5.1.
  • 📝Anthropic Life Sciences Verification Program — relaxed biology safeguards for vetted life-science teams on Mythos, Opus, and Sonnet models; developed with the US government and opened to public applications in September 2026.
  • Claude for Open Source — security-testing access for maintainers of critical open-source projects.
  • Claude Partner Network — a $100M investment, announced March 2026, to support the developer ecosystem.

Scale

  • Latest funding: $65B Series H at a $965B post-money valuation, announced May 28, 2026, co-led by Altimeter, Dragoneer, Greenoaks, 📝Sequoia Capital, Capital Group, Coatue, and D1 Capital Partners, and described as likely its last private round before the 📝Anthropic IPO. Source: Anthropic — Series H.
  • Revenue: annualized run-rate revenue crossed $47B in May 2026 and was reported above $65B in August 2026. Sources: Anthropic — Series H, Bloomberg.
  • Status: private company preparing to go public. It confidentially submitted a draft S-1 to the SEC on June 1, 2026, and on September 28, 2026, Reuters reported a prospectus targeting a valuation above $2 trillion, showing $4.6B in 2025 revenue and a $42B net loss, with a listing expected after the November midterm elections.
  • Major investors: 📝Google, 📝Amazon (Claude is available through Amazon Bedrock), and the Series H co-leads.

Core Approach

Anthropic develops steerable, interpretable AI through Constitutional AI — a methodology that embeds human-controllable values directly into models. The company has positioned itself at the intersection of capability and safety, arguing these are complementary rather than competing priorities. That posture now shapes its product line: its most capable models ship with safety classifiers for cybersecurity, biology, and distillation, and relaxed access runs through the vetting programs above. Its focus on agentic coding workflows and enterprise deployment distinguishes it from competitors focused primarily on consumer chatbots or API-only approaches.

Anthropic is the company whose products changed how I identify. For a decade I built systems and ran companies while saying "I don't identify as an engineer" — writing code was "not my thing." Claude Code removed that barrier entirely. Within months I went from not writing code to pushing commits to 📝GitHub daily, building multi-agent systems, and shipping a full platform rebuild. See 📝Building with Claude Code as a Non-Engineer and 📝Building on Claude: A MythOS Creator's Perspective.

That rebuild is 📝MythOS v3, and every line of its codebase was written in collaboration with Claude. Anthropic is also the engine behind 📝Brian Bot's 57-agent ecosystem and nearly everything else I ship. The 📝Augmentation Stack runs on its infrastructure — Claude models for reasoning, MCP for connectivity, Claude Code for building — and because MythOS is itself an 📝MCP server, every agent I run can read and write to my library directly. This memo was researched and rewritten by one of those agents.

I've built on every Claude tier, and each step up has been categorical rather than incremental. What Anthropic has made isn't just a model family; it's the collaborator my entire 📝human-AI augmentation practice is built around. Tools are things you pick up and put down. Collaborators are things you build a relationship with over time — and I'm deeply grateful for this one.

Contexts

Created with 💜 by One Inc | Copyright 2026