How the engineer behind Claude Cowork actually uses Claude | Felix Rieseberg (Anthropic)
Summary
Felix Rieseberg, an engineering lead at Anthropic, demonstrates diverse applications of Claude, emphasizing its role in automating tedious tasks to foster human creativity. He highlights that the primary barrier to AI adoption is users' limited understanding of its broad problem-solving potential. Rieseberg showcases personal workflows, including using Claude Co-work to generate 3D house models from 2D floor plans and leveraging email as a "source of truth" for personal inventory. He explains the heuristic for selecting between Claude Sonnet (for well-scoped problems) and Opus (for ill-defined problems), noting Sonnet's general efficacy. The discussion also covers "Live Artifacts" for dynamic, data-driven dashboards and a unique \$19 hardware "Claude Buddy" that provides physical interaction for approvals. Rieseberg stresses the importance of asynchronous design for managing AI latency and the uninhibited creativity children exhibit when interacting with AI.
Key takeaway
For AI Product Managers or AI Engineers seeking to expand user adoption, recognize that the primary hurdle is not AI capability but user awareness of its broad applicability. Focus on designing interfaces and workflows that encourage you to abstract tasks and trust AI for background automation, even for seemingly complex personal or hardware-related problems. Consider integrating physical interaction points or live, data-driven artifacts to make AI more tangible and delightful, fostering a mindset of "what if AI could do this?"
Key insights
AI's true potential lies in automating tedious tasks, freeing human creativity, and solving problems users don't realize are solvable.
Principles
- AI should automate tedious tasks, freeing human creative energy.
- The biggest AI adoption gap is user understanding of its broad applicability.
- For complex, ill-defined problems, use more capable models like Opus.
Method
Claude Co-work can analyze documents (e.g., floor plans), build interactive 3D models, and create live, self-refreshing dashboards by connecting to data sources like email, calendar, and Spotify.
In practice
- Use email as a "source of truth" for personal inventory.
- Build a physical "Claude Buddy" for task approvals.
- Ask Claude "how can you help?" for problem discovery.
Topics
- Claude Co-work
- Live Artifacts
- AI Workflow Automation
- Hardware Integration
- LLM Model Selection
- User Problem Scoping
Best for: Machine Learning Engineer, NLP Engineer, AI Engineer, AI Product Manager, Director of AI/ML
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by How I AI.