How I AI5 OpenClaw agents run my home, finances, and code | Jesse Genet
CHAPTERS
Discovering OpenClaw through Obsidian: “I need a second brain, but I’m missing my first brain”
Jesse explains how an Obsidian influencer comment led her to OpenClaw and why it clicked: she wanted a system that could organize homeschool life without requiring hours of manual upkeep. The core motivation is reclaiming time and cognitive bandwidth while parenting four young kids.
Building a Family Learning vault: turning daily lessons into structured data
Inside Jesse’s Obsidian “Family Learning” vault, she describes the structure she wanted but couldn’t sustain manually. OpenClaw enables capturing lesson details consistently so the system can later support planning and personalization.
Uploading entire books as curriculum sources (and why parents struggle with dense guides)
Jesse and Claire discuss photographing full curricula (like phonics/reading books) so an agent can transform them into usable, parent-friendly lesson plans. They brainstorm derivative outputs—extracting stories, creating printables, even physical learning aids.
From chapters to lesson plans: BFSU science curriculum auto-generated in Obsidian
Jesse shows how OpenClaw converts a science curriculum (BFSU) into organized lesson plans so she can teach quickly without reading every chapter in full. The plans include objectives, vocabulary, materials, and suggested activities.
Custom kid-friendly visuals on demand: book photo → watercolor printables
Using a lo-fi workflow (a phone photo of a book snippet), Jesse prompts her homeschool agent “Sylvie” to generate watercolor-style illustrations via an image model (Gemini/NanoBanana Pro). The result is high-quality, printable visuals with minimal prompting.
Why five agents: role-based personas, separated data access, and life partitions
Jesse explains she runs multiple OpenClaw instances with distinct responsibilities—homeschool, finance, scheduling/EA, coding—because mixing tasks feels inefficient and risky. Each agent gets only the files and tools needed for its role.
Hard truth on collaboration: Slack setup is painful and agent-to-agent chat isn’t solved
They dig into the technical friction of running agents in Slack, including creating custom Slack apps to add each agent as a bot. Jesse’s takeaway: existing communication tools are built for humans, and agent collaboration remains one of the hardest parts.
Security model: treat agents like new employees and build progressive trust
Jesse and Claire frame agent onboarding like hiring: start with limited permissions, avoid impersonation, and expand access gradually. They discuss separating communication channels from sensitive data to reduce blast radius.
Coding with “Cole”: building a kids-safe YouTube streamer app (Mira) in days
Jesse demos a custom app built with her coding agent to curate high-quality YouTube streams for her kids and avoid “AI slop.” The app constrains controls to simple navigation and history, acting as a parental-control layer with curated, endless playback.
Getting it onto the TV: pushing past “can’t” with agent-driven iteration
A key milestone is deploying the app onto a Google TV streamer so kids can’t easily escape into other apps. Jesse highlights the power of persistence: asking “try harder” until constraints are solved and the system works in real life.
The real lifestyle shift: asynchronous work for parents, voice workflows, and “no hands” reality
Both discuss how agents change time management for parents—work can happen in small windows, and tasks can continue while you’re away from the computer. Voice-to-text and agent autonomy matter most in postpartum and childcare contexts.
Bridging into the physical world: photo-based home inventory and just-in-time recommendations
Because agents can’t physically organize cupboards, Jesse uses photos to create a structured inventory of learning supplies, books, and materials. The agent then links inventory items to lesson plans and recommends what to pull out when it’s relevant.
The “killer feature”: agent-controlled printing for instant worksheets and materials
Jesse emphasizes how meaningful it is that her homeschool agent can print directly to a standard printer. The value is speed and reduced friction: a photo of a worksheet can become a printed handout in seconds without scanning/emailing workflows.
Managing the agent’s behavior: “decisions” files, SOUL.md edits, and self-diagnosis
In lightning-round mode, Jesse shares how she steers agents using system conventions: tagging something as a “decision” so it’s remembered, and asking agents to update their SOUL.md to adjust persona and behavior. She rarely edits files manually and instead has agents propose and apply changes themselves.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome