WhoBuilt

Automate Your Life with Claude Code in 40 Minutes

May 10, 2026· 43 min· 20K views·Watch on YouTube

Speaker

Moritz Kremb explains how he built a personal assistant system with Claude Code, comparing it with OpenClaw on Peter Yang’s channel and sharing examples from his workflows.

OpenClaw and Claude Code tradeoffs

Kremb says OpenClaw stands out for mobile access through apps such as Telegram and for its 30-minute heartbeat, which can check memory and to-do tasks. He finds cron jobs and subagents easier to manage there, while Claude Code feels more reliable for his work and offers access to a strong model. He uses OpenClaw for email and calendar tasks, and Claude Code to build things.

Building a personal assistant foundation

Kremb calls his Claude Code setup Claudia. It uses a CLAUDE.md file to reference instructions and context files for identity, user details, tools, and memory. He saves one-line daily memories and runs an overnight routine to compress them into long-term memory. He adds each new CLI, MCP, or skill to tools.md so the system can track its capabilities.

Connecting tools and repeatable workflows

Kremb connects Claude Code to tools through CLIs, MCPs, and APIs, and turns repeated workflows into skills. Examples include a grocery skill that refills a cart with regular items and adds new ones for approval, and a video upload workflow that creates folders in Google Drive for an editor. He also uses GWS CLI to create and manage Drive files.

Automating content production

Kremb’s content system captures ideas from Telegram, YouTube research, and Twitter messages, then organizes selected ideas into a weekly plan and drafts scripts. After reviewing and editing the scripts, he records videos on his phone and uses an automation to create upload folders. A CLI tool helps publish clips to YouTube, Instagram, and TikTok, while scripts can generate captions and Notion resources for audience downloads.

Start with one workflow

Kremb recommends setting up the folder structure first, then connecting tools one at a time. He says the manual work that matters most is reviewing and improving scripts, rather than accepting AI drafts as finished. His workflow supports posting roughly one short-form video per day.

Takeaways

  • Create a clear instruction and folder structure before adding integrations.
  • Track available tools in one file so the assistant can use its connected capabilities.
  • Turn workflows you repeat into skills, starting with tasks that consume the most time.
  • Keep human review in the loop for creative work such as scripts.
  • Choose tools based on the job: Kremb uses OpenClaw for communication tasks and Claude Code for building workflows.

Mentioned in this talk

Search inside this talk

What was claimed 20

Show all 20 facts

Related talks