Discover Anthropic's Record-Breaking Q1 Feature Launch: 120+ Innovations in 90 Days

Anthropic shipped 120+ features in 90 days.

How was your Q1?

The Claude teams shipped a feature every 18 hours. No other AI lab is shipping product at this pace. OpenAI has more users. Google has more distribution. Anthropic decided to win on tooling depth.

Look at the calendar closely. Agent Teams, Opus 4.6, and 1M context GA all landed in the same two-week window. That's not a coincidence. Opus 4.6 has a 14.5-hour task completion horizon, which is what makes Agent Teams work on real projects. 1M context means the whole codebase fits in one session. They shipped the model, the infrastructure, and the feature layer together.

Agent Teams lets a lead agent delegate to 10 parallel teammates. I tested it on three open issues (auth flow, dashboard layout, broken CSV export). All three had PRs open within 40 minutes. Two passed tests immediately.

Cowork is Claude running persistently on your computer with access to your apps, files, and screen. Before a 2-hour meeting, I told it to summarize my Slack DMs and update a revenue chart. Came back to both done. That's 30 minutes of scrolling I'll never do again on any meeting day.

/loop runs a task on repeat for up to 7 days. /loop 5m /babysit auto-addresses code review comments and pushes PRs to production. Chain it with Agent Teams and multiple features build and merge in parallel while you're in back-to-back meetings.

Connectors (Slack, Drive, Notion, Figma, 50+ others) are the quiet S tier. Each one you add makes every other feature more useful because Claude pulls context from more of your actual work. A session with three connectors active is a different tool than Claude with nothing connected.

The cost of this velocity showed up twice in Q1. A source map leaked 512,000 lines of TypeScript and 44 unreleased feature flags. Then the Mythos model leaked. Uptime has been rough. Speed without reliability is technical debt at the org level.

But the product philosophy underneath is worth studying. Boris Cherny's team built hundreds of versions of Agent Teams before shipping. ~30 prototypes of the condensed file view. 80% never shipped. They don't write better specs. They build more prototypes and kill more of them.

I tested all 120+ and ranked them. Here's everything:

1. Full tier list + PM workflows: https://www.news.aakashg.com/p/claude-q1-features
2. Cowork deep dive: https://www.news.aakashg.com/p/you-should-be-using-claude-cowork
3. My Claude Code setup: https://www.news.aakashg.com/p/claude-cowork-code-setup
4. Claude Code masterclass: https://www.youtube.com/watch
5. How to use Claude: https://www.news.aakashg.com/p/how-to-use-claude-for-work
6. 6 of my Claude skills: https://www.news.aakashg.com/p/steal-6-of-my-claude-skills

What's your favorite new Claude feature

#AIinnovation #Q1featurelaunch #Anthropic #technologyspeed #AItools