CORRECTION: I Incorrectly Stated That Anthropic Booked $6 Billion In Revenue in February

Correction: In yesterday’s newsletter, I incorrectly stated that Anthropic booked $6 billion in revenue in February. I apologize for the error.

A $6B revenue month implied a $72B+ annual run rate. Bloomberg reports that Anthropic nears a $20B annualized run rate, driven in large part by strong adoption of Claude Code.

I hold myself to a high standard of accuracy and regret the error.

---

Claude is on a tear

Early adopters on the technical and nontechnical side are flocking to Claude. We know of dozens of asset management firms and companies now rolling Claude out to their employees and updating their 201 and 301 training programs. Three firms told us that escalation requests to approve Claude went directly to the CEO.

We believe the interest is fully warranted.

• Claude Code is a standout product

• Claude with Excel and PowerPoint currently outperforms ChatGPT’s Excel and PowerPoint capabilities, and far exceeds Copilot

• Cowork is a task-oriented AI tool for nontechnical employees. Multiple companies are rethinking their use of tools like N8N, Zapier, and others as a result

• Claude has aggressively shipped features similar to those in OpenClaw, the white-hot, open-source, personal AI assistant software

For my personal use and all my colleagues at GAI Insights, Claude now accounts for 95% of our usage and is our preferred tool for personal employee productivity. I shut down my OpenClaw project because of Claude.

Anthropic Shipped 45+ Features in 90 Days

Anthropic has been shipping at a rapid pace Below is a list of features by product, each scored on a 1-10 impact scale for enterprise value. Several features include a link to a short product demo.

CLAUDE CHAT (Web & Mobile)

1. Opus 4.6 Model (Feb 5) - The most capable Claude model with 1M token context window, stronger multi-step reasoning, and top benchmark scores in legal, financial, and coding tasks. Impact: 9/10 (Video: analyze multiple files) (Video: turn messy notes into neat notes)

2. Sonnet 4.6 Model (Feb 17) - Full upgrade across coding, computer use, long-context reasoning, agent planning, knowledge work, and design with 1M token context window in beta. Impact: 8/10

3. Agent Teams (Feb 5) - Deploy multiple specialized AI agents to collaborate on different parts of a large project simultaneously. Impact: 9/10

4. Persistent Memory for All Users (Mar 2) - Claude retains your name, communication style, writing preferences, and project context across separate conversations. Impact: 7/10

5. Infinite Chats (Jan) - Eliminates context window limit errors by summarizing earlier messages when a chat approaches its limit. Impact: 6/10

6. Inline Visualizations (Mar 12) - Creates custom interactive charts, diagrams, and visualizations directly within chat responses. Impact: 7/10

7. File Creation: Excel, PPT, Docs, PDF (Jan/Feb) - Creates and edits Excel spreadsheets, PowerPoint decks, documents, and PDFs directly in the app. Impact: 8/10 [Maggie video Claude in Excel) (Video: create spreadsheet of contact from one company in my gmail)

8. Claude for Excel and PowerPoint Add-ins (Mar 11) - Add-ins now share full context across applications, support skills, and connect via LLM gateway to Bedrock, Vertex AI, or Microsoft Foundry. Impact: 8/10 (video of Claude with Powerpoint Add-in) (Video of FInancial modeling Claude vs ChatGPT)

9. Enterprise Analytics API (Feb) - Programmatic access to usage and engagement data for Claude and Claude Code Remote, aggregated per organization per day. Impact: 8/10

10. Mobile Interactive Apps (Mar) - Mobile app supports live charts, diagrams, and shareable visual assets rendered within the conversation. Impact: 5/10

11. Computer Use, macOS Preview (Mar 24) - Claude opens files, clicks through websites, and completes multi-step tasks on your screen using mouse and keyboard. Impact: 8/10

CLAUDE COWORK (Desktop Agent)

1. Cowork Launch (Jan) - Desktop agent for knowledge workers that reads files, executes multi-step workflows, and produces deliverables autonomously with no coding required. Impact: 9/10 (video overview of Cowork)

2. Plugin Marketplace with Admin Controls (Feb 17) - Third parties publish skills into the Claude Marketplace. Six enterprise partners live: GitLab, Harvey (legal), Rogo (finance), Snowflake, Lovable, Replit. Impact: 9/10

3. 38+ Connectors (Feb) - Connectors for Gmail, Google Drive, Notion, Slack, Microsoft 365, and more. Impact: 8/10

4. Scheduled and Recurring Tasks (Feb 25) - Create and schedule both recurring and on-demand tasks from within Cowork. Impact: 8/10

5. Computer Use in Cowork (Mar 24) - Claude opens apps, navigates browsers, fills spreadsheets, and performs tasks on your screen with no setup required. Impact: 9/10

6. Dispatch, Cross-Device (Mar 17) - Persistent cross-device conversation that lets you assign tasks to your desktop Cowork session from your phone. Claude executes locally using your files and apps. Impact: 8/10
   

CLAUDE CODE (Developer CLI & Agent Platform)

1. Opus 4.6 as Default Model (Mar) - Opus 4.6 is now the default model for Claude Code, providing stronger reasoning and longer agentic task performance. Impact: 9/10 (video build analyze, export data)

2. 1M Token Context Window (Mar) - One million token context for Max, Team, and Enterprise plans. Work with very large codebases without premature compaction. Impact: 9/10

3. Claude Code Security (Feb 20) - Scans codebases for security vulnerabilities and suggests targeted patches for human review. Reasons about code like a human researcher rather than matching known patterns. Impact: 10/10

4. Channels, Telegram and Discord (Mar 20) - MCP-based feature connecting your running Claude Code session to Telegram and Discord. Send instructions and receive results from your phone. Impact: 7/10

5. Voice Mode, Push-to-Talk (Mar) - Hold the spacebar and dictate coding instructions aloud in over 20 languages instead of typing. Impact: 5/10

6. /loop Scheduled Tasks (Mar) - Executes a prompt at regular intervals (lightweight cron job) for PR reviews, deployment monitoring, and other recurring work. Impact: 7/10

7. Remote Control from Mobile (Mar) - Monitor and direct your laptop’s local coding session from your smartphone or the web while away from your desk. Impact: 6/10

8. Background Subagents (Mar) - Spawns secondary AI workers to handle parallel coding tasks without cluttering your main workspace. Impact: 8/10

9. /effort and Ultrathink (Mar) - A keyword in your prompt temporarily activates high reasoning effort for a specific turn when deeper analysis is needed. Impact: 6/10

10. MCP Elicitation (Mar) - Allows MCP servers to request structured input during execution, displaying interactive forms to collect data without interrupting workflow. Impact: 7/10

11. 128K Max Output Tokens (Feb/Mar) - Opus 4.6 defaults to 64K output with maximum capacity of 128K tokens. Generates much larger code files in a single response. Impact: 7/10

12. Credential Scrubbing (Mar) - Strips Anthropic and cloud provider credentials from subprocess environments to prevent accidental exposure. Impact: 8/10

13. Plugin Marketplace for Code (Feb/Mar) - Install bundled toolkits, server connections, and AI skills with a single click rather than configuring dependencies manually. Impact: 7/10

CLAUDE API (Developer Platform)

1. 1M Token Context Window (Feb) - Feed an entire corporate knowledge base or massive codebase in a single API request. Impact: 9/10

2. Adaptive Thinking Mode (Feb 5) - Claude automatically decides whether and how deeply to engage extended reasoning on each request, replacing the old manual budget_tokens approach. Impact: 7/10

3. Context Compaction, Beta (Feb 5) - Automatic server-side context summarization so conversations continue effectively past the context window limit without manual truncation. Impact: 8/10

4. Effort Controls, 4 Levels (Feb/Mar) - A four-level dial (low to max) letting developers manually set how deeply the model thinks, trading off intelligence against cost and speed. Impact: 7/10

5. 128K Output Tokens (Feb 5) - Generates massive single responses without splitting output across multiple API requests. Impact: 7/10

6. Automatic Prompt Caching (Feb/Mar) - Remembers recent conversation parts to reduce cost and latency without manual coding changes. Impact: 8/10

7. US-Only Inference (Feb) - A new inference_geo parameter routes API calls to U.S.-only infrastructure, meeting compliance requirements in healthcare, finance, and government. Impact: 8/10

8. Fast Mode, Beta (Feb 5) - A speed flag that delivers up to 2.5x faster output from Opus models at premium pricing, without changing the model’s intelligence. Impact: 7/10

9. Free Code Execution with Web Tools (Feb 5) - Code execution is now included at no extra charge when used alongside web search or web fetch tools, reducing cost for research workflows. Impact: 6/10

10. Dynamic Filtering for Web Search (Feb 5) - Claude writes and runs code to filter web search and web fetch results before they enter the context window, improving accuracy and cutting token usage. Impact: 6/10

11. Tools Graduating to GA (Feb 5) - Several previously beta API tools (code execution, web fetch, programmatic tool calling, tool search, memory tool) all moved to general availability. Impact: 7/10

12. Fine-Grained Tool Streaming, GA (Feb 5) - Fine-grained tool streaming now generally available on all models with no beta header required, enabling real-time visibility into tool calls. Impact: 7/10

13. Structured Outputs, GA (Early 2026) - Structured outputs moved to general availability, providing guaranteed JSON schema conformance in production without a beta header. Impact: 8/10

14. Skills API (Dec 2025/Jan 2026) - An open standard that lets skills built for Claude work across AI platforms, with organization-wide management for Team and Enterprise plans. Impact: 7/10

---

Join us Monday April 6, at 7p ET / 4p PT to see a demo OpenClaw (a Personal AI Agents) and to discuss the implications for companies. Claude is already matching some of these features and ChatGPT is expected to as well. Register here (142 have already registered).

---

Our AI Adoption Specialists are conducting a lot of GenAI 201 and 301 instructor-lead training for ChatGPT, Copilot, Claude, and N8N. Contact us if you have this need.

---

Onward,
Paul

Resources and Media:

LinkedIn: Calendar: Learning Lab: HBR Article; Daily AI Show: GAI World 2026: TEDx: X/Twitter: TikTok

FAQ

Q: How does Claude compare to ChatGPT and Microsoft Copilot for enterprise productivity?

A: Claude now outperforms ChatGPT in Excel and PowerPoint add-ins and significantly exceeds Microsoft Copilot in both. Anthropic shipped more than 45 new features in 90 days, including file creation for Excel, PowerPoint, and PDF, persistent memory, and a desktop agent called Cowork that replaces workflow tools like Zapier and N8N. For non-technical employees, Claude Cowork requires no coding and executes multi-step tasks autonomously.

Q: What new Claude features matter most for enterprise AI adoption in 2026?

A: The highest-impact releases include Agent Teams for parallel task execution (9/10 enterprise value), Claude Code Security for autonomous vulnerability scanning (10/10), a 1 million token context window across Chat, Code, and API, and Computer Use for macOS that controls your screen without setup. Persistent memory, available to all users as of March 2026, retains project context and communication preferences across sessions.