Key Takeaways

  • Reverse the intake prompt: Farrar asked Claude what data to collect about his home rather than guessing, which led him to photograph every appliance serial number plate first.
  • Track device end-of-life alongside hardware: Farrar logs both appliance warranty dates and digital obsolescence dates, like when his child's iPad will lose iOS software updates.
  • Ground insurance shopping in personal asset data: By feeding raw insurance quotes alongside his existing home files into Claude Code, Farrar isolated specific policy trade-offs and generated precise questions for his broker.
  • Extract immediate blockers with open prompts: Asking Claude Code “what should I do next on this project” (play from 6:37) produced an exact task, like emailing agent Dana Whitfield, cutting through decision fatigue.

The Method

Farrar wanted a reliable system to handle the friction of homeownership. Instead of buying home management software that locks records in proprietary silos, Farrar combined local Markdown files in Obsidian with Claude Code running in his terminal.

The workflow begins by letting the model define the data model. “When I set this system up, I said to Claude, 'What information should I gather about my house?'” (play from 9:36) Farrar explained. “I asked it and I said, perhaps what about the refrigerator? And it said, 'Well, actually go around and take pictures of every appliance in your house.'” (play from 9:42)

Farrar photographed serial number plates on his range, refrigerator, and water heater, then used the AI to transcribe model numbers and retrieve warranty periods. He expanded the index beyond physical machinery to consumer electronics: “When is my kid's iPad going to lose iOS support software updates?” (play from 9:23) By storing these details in local Markdown files, he created a searchable registry without manual database maintenance.

When maintenance emergencies arise or policies renew, Farrar grounds new queries in that local repository. When his refrigerator ice maker failed or when he evaluated car insurance, he fed the raw PDFs directly into Claude Code alongside his home records. “With the power of AI, you can read it via that AI, and then I wanted to bring in my own home context and how that system meshed with what the insurance agent was giving me,” (play from 7:32) Farrar said. “And then I was able to ask way more detailed questions and email back and forth with the insurance agent and know exactly what I was asking and know exactly what they were telling me.” (play from 7:44)

To prevent projects from stalling, Farrar uses simple status queries. “I might say like, 'Well, what should I do next on this project?' And because I'm grounded on what I'm working on here, it comes back with the response says, 'Oh, your next step is email Dana Whitfield.' That's the blocker.” (play from 6:35)

Where This Breaks Down

This system requires catalog hygiene that is difficult to sustain. Taking photos of serial numbers is simple on a weekend, but updating text files when you replace an appliance months later is easy to skip. Once the local repository drifts from reality, the model gives recommendations based on hardware you no longer own.

Terminal tools also create friction for household members. Running Claude Code locally works if you are comfortable in a terminal, but sharing that context with a spouse who prefers standard note apps creates a sync barrier. If only one person updates the files, the household knowledge base quickly becomes an isolated silo.

What to Do With This

Walk into your kitchen tomorrow and photograph the serial number plates on your refrigerator, dishwasher, and oven. Feed those photos into an AI prompt asking for model specifications, expected lifespans, and warranty status. Save the output into a plain text file, then paste your latest auto insurance declaration page into the same chat to check for coverage gaps.