Anthropic shipped 74 Claude releases in 52 days. This is what a company pulling ahead looks like.

I mapped every release since February 1st to a shipping calendar — dates, features, team attribution. Not by reading changelogs. By going through the feeds of the people actually shipping.

What stands out isn't any single feature. It's the velocity across every surface at once:

→ Developer tools (Claude Code) — 28 releases
→ Desktop automation (Cowork) — 15 releases
→ API and infrastructure — 18 releases
→ Models and core platform — 13 releases

No team is waiting for the other. They're all shipping in parallel:

For PMs, this creates a compounding gap. Every week you don’t explore Dispatch, Scheduled Tasks, or Computer Use is a week someone else does. The PM who builds a self-improving CLAUDE.md system today has an automation layer that gets better with every release. The PM who waits builds from scratch each time.

This isn't about picking the "best AI tool." It's about recognizing when one platform's shipping velocity makes the decision for you.

My Claude Guides

If you want to go deeper on any of these surfaces, I’ve written complete guides for PMs:

• Claude Cowork Guide for PMs — the desktop agent that works with your files, apps, and connectors

• Claude Code: The Complete PM Guide — the CLI/IDE agent for technical work

• The Self-Improving Claude AI System — how to build a knowledge system that compounds across sessions

• The Claude Dispatch Guide for PMs — the always-on persistent agent that runs while you’re away

Timeline & Everything You Might Have Missed

Feb 3 – Feb 9

CC Session Sharing: Share Claude Code sessions with teammates for collaborative debugging and pair programming.

Cowork Slack MCP on Paid Plans: Access Slack messages, channels, and search directly from Cowork on all paid plans.

Cowork GSuite Connectors: Connect Gmail, Google Calendar, and Google Drive to Cowork for cross-app workflows.

CC /insights Command: Get AI-generated analysis of your codebase’s architecture, patterns, and dependencies.

CC Agent Teams: Spawn multiple Claude Code agents that work in parallel on different parts of a codebase.

CC /debug Skill: Auto-diagnose failing tests and runtime errors with a dedicated debugging workflow.

Opus 4.6: New flagship model with improved reasoning, coding, and instruction-following across all Claude surfaces.

CC /rewind: Press ESC twice to summarize and compress your conversation context instead of losing it.

CC Fast Mode: Optimized response mode that trades depth for speed on routine coding tasks.

Fast Opus 4.6: Early preview of Opus 4.6 running 2.5x faster than the standard version.

Feb 10 – Feb 16

Cowork for Windows: Cowork desktop app now available on Windows alongside the existing Mac version.

CC --dangerously-skip-permissions: Run Claude Code with auto-approved tool use for fully autonomous, unattended workflows.

CC Contribution Metrics: Track how much code Claude wrote versus you across your project’s history.

Free Plan Expansion: Free users get file uploads, connectors, and skills in Cowork.

Cowork Folder Instructions: Drop a CLAUDE.md file into any folder to give Cowork persistent, project-specific context.

CC Slack App (Plan Mode): Use Claude Code’s plan mode directly from Slack to scope technical work before executing.

CC Desktop Local Plugins: Install and run local MCP-based plugins on the Claude Code desktop app.

CC Multi-Repo: Work across multiple repositories in a single Claude Code web session.

CC SSH Support: Connect the Claude Code desktop app to remote development machines over SSH.

Feb 17 – Feb 23

Sonnet 4.6: New mid-tier model balancing speed and quality, available across all Claude surfaces.

CC Sonnet 4.6 Support: Sonnet 4.6 available as a selectable model option in Claude Code with early benchmark results.

CC Figma MCP: Push Claude Code outputs directly into Figma for design-development handoff.

Claude in PowerPoint: Create and edit PowerPoint slides directly within the Claude interface on Pro plans.

CC Security (Research Preview): Automated security scanning that identifies vulnerabilities in your codebase before they ship.

CC Desktop Preview Apps + CI/PRs: Preview running applications and handle CI/CD and pull request workflows from the desktop app.

CC Worktrees: Built-in git worktree support so agents can work on parallel branches without switching context.

Feb 24 – Mar 2

CC Remote Control: Control Claude Code sessions from your phone — review, approve, and guide agents on the go.

Enterprise Customize + Marketplaces: Enterprise admins can customize Claude behavior and create private plugin marketplaces for their org.

Cowork Scheduled Tasks: Schedule recurring Cowork tasks that execute automatically on a cadence you define.

CC Auto-Memory: Claude Code automatically remembers your preferences, codebase patterns, and project context across sessions.

CC Claude for Open Source: Free Claude Code access for qualifying open source project maintainers.

CC /simplify + /batch: Refactor complex code into simpler patterns (/simplify) or run the same operation across many files at once (/batch).

CC AskUserQuestion Snippets: Claude Code shows formatted markdown previews with code snippets when asking you clarifying questions.

Memory on Free Plan: Persistent memory across conversations now available to all free-tier Claude users.

Mar 3 – Mar 9

Voice Mode: Talk to Claude Code using push-to-talk voice input instead of typing commands.

CC HTTP Hooks: Fire HTTP webhooks on Claude Code events — connect sessions to Slack, CI pipelines, or any external system.

CC Local Scheduled Tasks: Schedule recurring tasks that run automatically on your machine while your computer is awake.

Enterprise Marketplace: Enterprise customers get a curated marketplace of approved plugins and integrations for their org.

CC /loop: Set Claude Code to continuously iterate on a task — building, testing, and fixing in a loop until your tests pass.

CC Auto Mode: Claude Code automatically selects the right execution approach without requiring manual mode switching.

CC Code Review (Multi-Agent): Multiple Claude agents analyze your pull request in parallel, each reviewing different aspects of the code.

Mar 10 – Mar 16

CC /btw: Send Claude a side note mid-task without interrupting its current work — corrections, context, or new info.

CC /effort: Control how deeply Claude thinks — from quick one-shot responses to extended multi-step reasoning.

Cowork Excel + PowerPoint Sync: Edit Excel spreadsheets and PowerPoint decks in Cowork with cross-file synchronization.

Interactive Charts: Claude can now build interactive charts, diagrams, and visualizations directly in chat. HTML-based, editable, available on all plans.

Rich Text Output: Claude no longer defaults to plain text — responses now include interactive charts, formatted tables, and rich media.

CC Setup Scripts: Add a setup script that runs automatically when Claude Code initializes in your project directory.

Voice Mode on Desktop + Cowork: Push-to-talk voice input now available on the Claude desktop app and in Cowork sessions.

1M Context Window GA: One-million-token context window generally available across all paid Claude plans.

CC Opus 1M Default: Opus 4.6 with 1M context is now the default model in Claude Code.

CC Session Spawning: Spawn new local Claude Code sessions directly from the mobile Remote Control app.

CC /color: Customize your Claude Code terminal with color themes and visual preferences.

CC Review Manual Trigger: Type @claude review in any PR comment to trigger a multi-agent code review on demand.

CC MCP Elicitation: MCP servers can now ask Claude Code for additional context during tool execution instead of failing silently.

2x Usage: Double usage limits during off-peak hours for all paid plan subscribers.

Mar 17 – Mar 23

Cowork Dispatch: A persistent, always-on Claude conversation that runs on your computer — continues working even when you close the app.

CC Output Tokens 64K: Maximum output token limit increased to 64K for longer generated code and responses.

Opus 1M on Desktop: Opus 4.6 with 1M context now available on the desktop app for Max, Teams, and Enterprise users.

Dispatch + CC Sessions: Dispatch can now launch and manage full Claude Code sessions alongside its standard Cowork tasks.

CC Shell in SKILL.md: Embed live shell command outputs directly in skill files using !command syntax for dynamic context.

Claude.ai Speed Boost: Claude.ai and desktop apps load significantly faster after migrating from SSR to Vite + TanStack Router.

CC VSCode Remote: Use Remote Control to manage Claude Code sessions running inside VS Code from your phone.

CC Skills over Remote Control: Run skills and slash commands from your phone via Remote Control — full command access on mobile.

CC Channels (Telegram + Discord): Connect Claude Code to Telegram and Discord — interact with your coding agent from messaging apps.

CC Cloud Scheduled Tasks: Schedule recurring Claude Code tasks that run in the cloud — no local machine required.

Projects in Cowork: Organize Cowork conversations into named projects for better context separation and management.

CC DOM Element Selection: Select DOM elements directly from a running application to reference them in Claude Code conversations.

CC Effort in Skills: Set effort levels directly inside skill and slash command definitions to control reasoning depth per task.

CC /loop 7 Days: Loop sessions now run for up to 7 days instead of the previous 3-day limit.

CC /init v2: Redesigned project initialization that interviews you about your codebase before generating CLAUDE.md configuration.

CC /schedule Cloud: Schedule cloud-based tasks directly from the /schedule command — tasks run on Anthropic’s infrastructure.

Computer Use: Claude can now control your computer directly — mouse clicks, keyboard input, and screen reading for end-to-end automation.

CC Channels Permission Prompts: Channels now support permission prompts so you can approve sensitive actions from Telegram or Discord.

Mar 24 – Mar 30

CC Channels for Teams + Enterprise: Channels (Telegram, Discord integration) now available for Teams and Enterprise plans.

And the week continues.

[Edited] Here we go, not included in the infographic:

CC Auto Mode (Research Preview): Claude autonomously handles permission prompts during coding. A pre-execution classifier reviews every tool call, blocking destructive operations and prompt injection. Enable with `claude --enable-auto-mode`.

P.S. Glad Claude Team noticed this. Noah is the legend:

Have a great rest of the week,
Paweł