I wanted 20+ years of tackle historical past. Gemini discovered it for me in minutes

I received my tackle historical past geese in a row ultimately, however there’s a neater solution to accomplish it—with the assistance of an AI agent.

For this job, we’ll be utilizing Gemini’s Spark agent and its Gmail connector. (Sure, you’ll want to permit Spark to comb via your Gmail inbox.) In the event you’re not comfy with that, no drawback—be at liberty to skip this tip. In the event you’re new to Gemini’s Gmail connector, see this text first for particulars about the most effective configuration settings and what you could learn about privateness.

Upon getting Spark’s Gmail connector configured, begin a brand new thread and provides it this immediate (which was crafted by Claude Opus 5.5):

I would like my tackle historical past for a kind that asks the place I’ve lived for the previous [X] years. Search all of my Gmail, together with archived mail, from [START YEAR] to at present.

1. Search for something that exhibits the place I lived and when: leases and rental functions, move-in/move-out and safety deposit emails, utility and web begin or cease companies, USPS change-of-address confirmations, DMV or voter registration confirmations, mortgage or closing emails, insurance coverage insurance policies, and modifications within the “ship to” tackle alone orders.

2. Create a brand new spreadsheet referred to as “My tackle historical past” with one row per tackle, latest first: full road tackle, metropolis, state, ZIP, earliest month I lived there, final month I lived there, how assured you might be (confirmed by a lease/closing/USPS electronic mail, or inferred from orders/payments), and hyperlinks to the two–3 emails that show every date.

3. Flag any hole or overlap of multiple month between addresses, and listing what sort of electronic mail would settle it.

Solely use addresses and dates you may hyperlink to an electronic mail. In the event you can’t pin a date down, write “unknown” reasonably than guess. Don’t embody anybody else’s tackle (items, individuals I shipped to). Don’t fill in or submit any kind, sign up wherever, or contact anybody.

I examined this immediate with my very own Gmail account, which has messages stretching again greater than 20 years. Inside a couple of minutes, Spark spat out a spreadsheet that seemed moderately correct, piecing collectively my rental historical past stretching again to 2006 utilizing every little thing from lease agreements and move-in notices to safety deposit returns and utility confirmations.

The spreadsheet additionally specified which dates Spark was assured about and which of them it wasn’t. For instance, it inferred the final month I lived at my oldest tackle primarily based on numerous orders and payments that it discovered. That’s not iron-clad proof that I moved out in a given month, however it’s a strong guess—and extra importantly, I know it’s a guess, not truth.

This agentic AI immediate may work with plenty of different brokers, too, together with a ChatGPT “dot” agent, Claude Cowork, and Meta Muse (and, actually, this immediate was impressed by this text about Muse). I examined the immediate with ChatGPT, Claude, and Muse brokers and received comparable outcomes from all of them, however Spark’s spreadsheet was essentially the most thorough.