Anthropic silently shipped 3 more features in Claude Code last week

My notes

Summary

A 59-second TikTok covers three real Claude Code features shipped in late March/early April 2026: Ultraplan (cloud-based plan drafting, v2.1.91+), /powerup (interactive feature lessons, v2.1.90), and /insights (session analysis report). The descriptions are somewhat embellished but the features are genuine. The video also correctly references Claude Mythos (cybersecurity research preview via Project Glasswing), though calls it “Ultron.”

Key Insight

• All three features are confirmed in official Anthropic docs and changelog (weeks 13-14, 2026)
• Ultraplan: real but overstated - it is cloud-based plan drafting in a single session, not “multiple parallel sub-agents on research, assets, and diagrams”
• /powerup: real - interactive lessons with animated demos covering 10 Claude Code topics, somewhat context-adaptive to your project
• /insights: real - generates an HTML report from ~30 days of session data with usage patterns, friction points, and suggestions
• Mythos is real (Project Glasswing, cybersecurity-focused, invitation-only) - the “Ultron” name is wrong
• This creator tracks Anthropic releases closely and provides useful signal despite TikTok-style embellishment