More How I AI episodes

How the engineer behind Claude Cowork actually uses Claude | Felix Rieseberg (Anthropic) thumbnail

How the engineer behind Claude Cowork actually uses Claude | Felix Rieseberg (Anthropic)

Published 25 May 2026

Duration: 00:59:25

AI automates repetitive tasks to enhance human creativity, emphasizing user understanding challenges, DIY hardware innovations, versatile models like Claude, interface preferences, personalized workflows, generational adoption differences, and ethical design considerations.

Episode Description

Felix Rieseberg is the engineering lead for Claude Cowork and Claude Code Desktop at Anthropic. He previously spent five years at Slack building devel...

Overview

The text explores the evolving role of AI in automating repetitive tasks to free humans for creative work, emphasizing that challenges often stem from user understanding rather than AI limitations. It highlights practical applications of AI tools like Claude across coding, collaboration, and terminal interfaces, while stressing the need for user-driven design and adaptable interfaces to cater to diverse preferences. Hardware integration ideas, such as a $20 DIY device with Wi-Fi/Bluetooth for physical interaction with AI systems, are presented as low-cost solutions to enhance automation. The discussion also touches on generational differences in AI adoption, noting childrens natural engagement with tools like Claude for tasks ranging from creative projects to technical problem-solving, contrasting with older users adherence to traditional workflows.

Key themes include leveraging AI for structured problem-solving and creativity, such as generating 3D home plans from 2D data or organizing personal inventories via email integration. The text underscores the importance of abstraction in workflows, prioritizing high-level task delegation over manual input, and balancing model selection (e.g., Sonnet vs. Opus) based on complexity. It also addresses ethical considerations, such as empowering users through AI without stifling imagination, and the value of iterative development driven by user feedback. Examples of AI-driven dashboards, live data synchronization, and hardware-software collaboration illustrate the potential for AI to streamline productivity while emphasizing transparency and user control in automation.

What If

  • What if you built a hardware "claw" with Wi-Fi/Bluetooth integration to confirm AI task approvals?
    Concrete move: Design a $20 DIY hardware interface with a physical button and LCD screen to interact with Claude, allowing you to approve AI-generated actions (e.g., approving a document draft or initiating a task).
    Why now: The text emphasizes the need for simple physical interfaces to simplify AI interaction, and the $20 DIY hardware project is already feasible with existing tools.
    Expected upside: Reduces manual input friction in workflows, enables hands-free task delegation, and creates a tactile feedback mechanism for AI collaboration.

  • What if you used Claudes terminal interface to automate home planning from raw data?
    Concrete move: Create a workflow where you input a 2D floor plan (e.g., from a scanned document) into Claudes terminal, and let it generate a 3D model, suggest furniture layouts, and calculate unit dimensions automatically.
    Why now: The text describes a personalized example of using AI to derive floor plans and layouts, and Claudes terminal interface is already viable for such tasks.
    Expected upside: Saves hours on manual home design tasks, leverages Claudes versatility for creative problem-solving, and creates a reusable tool for property planning.

  • What if you built a live artifact dashboard to auto-refresh your personal inventory from email data?
    Concrete move: Configure QOD to create a live artifact that pulls furniture purchase receipts from your Gmail, auto-populates a virtual inventory, and suggests styling recommendations via Claude.
    Why now: The text highlights using email data as a source of truth for inventory tracking and mentions live artifacts ability to auto-refresh with real-time data.
    Expected upside: Eliminates manual data entry for household tracking, provides dynamic insights (e.g., decluttering tips), and demonstrates AIs role in personal accountability systems.

Takeaway

  • Leverage AI for Repetitive Task Automation: Integrate AI tools like Claude to handle repetitive, rule-based tasks (e.g., generating structured documents, updating dashboards) so you can focus on creative or strategic work. Use the "refresh" feature in live artifacts to automate data updates from sources like Notion or Spotify.
  • Build a Low-Cost Physical Interface for AI: Create a $20 DIY hardware "claw" with Wi-Fi/Bluetooth connectivity to interact with AI systems. Use it to approve tasks via a physical button, reducing manual input and enabling hands-free workflows (e.g., approving file creation or data requests).
  • Optimize Model Selection for Task Complexity: Use Sonnet 4.6 for routine, well-defined tasks (e.g., coding, generating floor plans) and switch to Opus for complex or nuanced problems (e.g., legal/medical reinterpretation). Avoid over-relying on AI for basic automation (e.g., cursor movement).
  • Design User-Friendly Prompts for AI Tools: Structure prompts to guide AI toward high-level task abstraction (e.g., "Generate a 3D home layout from this 2D floor plan"). Use clear instructions and avoid excessive detail to reduce confusion and improve output accuracy.
  • Automate Personal Productivity with Live Artifacts: Build dynamic dashboards using live artifacts that pull real-time data from email, calendars, or inventory systems. For example, automatically update a "personal closet" database with clothing purchases from emails to streamline planning or styling.

Recent Episodes of How I AI

6 Jul 2026 How I run autonomous coding agents from my phone with OpenAI Symphony + Linear | Alessio Fanelli (Kernel Labs)

AI automates small business tasks like inventory tracking and order management via tools such as "magic glasses," explores personal AI use cases (e.g., Codex for hobby tasks), delves into autonomous agent orchestration with cloud-based workflows and GitHub, addresses challenges like scalability and model behavior, and reflects on AIs potential to bridge physical-digital systems, reduce manual effort, and enhance productivity while highlighting underutilized automation opportunities.

30 Jun 2026 Sonnet 5 review: I ran 64 generations to find out if it's worth it

Anthropic's Claude Sonnet 5 offers Opus-level performance at reduced costs with enhanced agentic capabilities, while a new benchmarking framework evaluates its competitive edge against models like Gemini 3 Pro and GPT 5.5, highlighting the need for standardized, human-informed evaluations to balance objective metrics and subjective quality.

24 Jun 2026 GLM 5.2: why Im replacing Opus in Claude Code with this new model

GLM 5.2, an open-weight model from Z.ai, offers a 1 million-token context window, strong performance on coding and reasoning tasks, cost-effectiveness, and local deployment flexibility, though it lacks image support and struggles with modern frontend frameworks.

22 Jun 2026 How Claude Mythos found a 15-year-old bug in Mozilla Firefox | Brian Grinstead

Recommended: AI finds bugs

Firefox employs AI agents as "coding archaeologists" to detect and address security vulnerabilities in its massive codebase, leveraging models like Mythos and custom validation tools to identify and systematically fix nearly 500 bugs, while balancing automation with human oversight and open-source collaboration to enhance scalability and security.

More How I AI episodes