À PROPOS DE CET ÉPISODE
Claude spent more than three hours rebuilding our Subscribe & Save schedule while I spent roughly twenty minutes answering questions and checking decisions between sessions of Final Fantasy XIV. I wouldn’t have spent a day doing that work myself. I also built an email concierge, shopped for RV gear with Jason, rejected a campsite drawing that ran lines through parked RVs, and found three parts for a Roomba that wasn’t ready to die.
Claude maxed my Amazon subscriptions while I played video games
Claude compared 485 items in my Walmart history with my real Amazon prices, rebuilt 17 scattered subscriptions, restored missed items, calculated consumption, and worked around Amazon’s quantity and discount rules. I checked in when it needed household judgment—rejecting a weak Roomba kit, correcting dog-food assumptions, removing an 18-package Oreo result, and deciding which gaps to leave alone. Claude finished with 28 verified subscriptions and delivery bundles large enough for Amazon’s maximum savings tier.
What I got out of it: More savings and a better household delivery schedule for about twenty minutes of my attention instead of a day of research and arithmetic I realistically would never have done.
Why AI was useful: Claude could keep researching, clicking, calculating, correcting, and verifying for hours while I played Final Fantasy XIV. I stepped in for the decisions that required knowing our household rather than doing the repetitive work myself.
I said I never want to open my email again. AI said - I got you.
I hate opening my email app, but I still want the inbox at zero. The concierge now checks two accounts hourly, clears routine clutter, turns messages into tasks, routes selected notices to Jason, and keeps short-lived records outside the inbox. It also needed rules for its own mistakes, including a fallback for blocked deletions and a correction when it treated spam as useful.
What I got out of it: A quieter inbox and a brief containing the decisions and real-world actions that actually need me.
Why AI was useful: It can keep returning to both inboxes, apply the rules I teach it, and connect replies and confirmations with work happening elsewhere—even when I don’t open Gmail.
AI turned a patio rug into a ruler for our dogs’ campsite yard
AI searched for a supervised dog fence, portable pellet grills, and a lighted RV rug. When it described rug colors without showing them, I made it display the pictures inside the chat. I emailed the useful choices to Jason, and we used the 9-by-12 rug to picture how much yard the larger fence would create. We ordered the rug and fence and put the grill on the calendar.
What I got out of it: A shared decision with Jason and a campsite setup I could picture before the boxes arrived.
Why AI was useful: It kept the products, pictures, emails, delivered prices, and space calculations together while Jason and I supplied the taste and final judgment.
AI found the campground I couldn’t name and measured the only site left
AI found Desert Moon from my memory of a blue hotel and old buses, compared it with two Colorado campgrounds, and surfaced an age rule that made one a bad fit for our 1998 Holiday Rambler. The first Site 5 drawing was unusable, so I moved the work to Codex. It preserved the Google image, measured the 50-foot scale bar, and placed our 32-foot RV and 16-panel fence while labeling uncertain boundaries as estimates. We booked the site.
What I got out of it: A campground we chose together, a measured plan for the only open site, and a completed reservation.
Why AI was useful: It connected a strange memory, current campground rules, Jason’s input, and image measurements. I rejected its plausible-looking first picture and made it preserve and measure the actual Google image.
AI found three parts to save my five-year-old Roomba
Our five-year-old Roomba still cleaned well, so I asked what deserved replacement beyond brushes and filters. AI separated worn tires from working wheel motors, found a battery, genuine front caster, and replacement tires for the exact j7+ generation, and checked reviews from owners of the same model. I ordered all three.
What I got out of it: A focused refurbishment instead of replacing functioning modules or the entire vacuum.
Why AI was useful: It matched the exact model, separated replacement parts from clean-and-inspect work, and checked whether the parts worked for real owners rather than trusting “compatible” on the listing.
This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit handsonailab.substack.com
