What Gmail Knows About Your Life (and Why Your Notes App Should Too)

Your email inbox knows more about your actual life than any app you use intentionally. Not because you planned it that way. Because everything found its way there anyway.

The lease renewal. The doctor's appointment from 2022 whose exact date you'd have to search to find. The salary negotiation. The dinner reservation from the night you decided to leave that job. A decade of flights, receipts, introductions, and follow-ups — all timestamped, still there, searchable in seconds.

Your notes app, meanwhile, has the outline for a project you eventually abandoned, forty-something untitled documents, and some book titles from a podcast you can't remember the name of.

The accidental archive

We didn't plan for email to become the authoritative record of our lives. It happened because email is where other people put things — confirmations, invoices, decisions — and you never quite deleted them. The inbox is the most comprehensive personal knowledge store most people own, and almost nobody treats it that way on purpose.

This is the part that bothers me. Email is a push medium. Things arrive because someone else sent them. Your lease renewal shows up in your inbox not because you decided that's where your housing history lives, but because your landlord's property management software uses SendGrid. You're the passive recipient of other people's record-keeping.

The places where you write — notes apps, journals, task lists — know almost none of this. They capture what you chose to record, in the moments you remembered to record it, and then they wait. There's no connection to the rest of your life. Ask your notes app when your last dentist appointment was. Ask it what you promised your colleague three weeks ago. Ask it what you decided about that project in November.

It doesn't know. It was waiting for you to tell it, and you didn't.

What AI is doing about it, and where it falls short

The obvious response to this fragmentation is to connect everything. That's what most AI tools are building toward now. Slack AI draws from your messages, calendar, and shared files. Microsoft Copilot reaches into Teams chats and Outlook. A 2022 Harvard Business Review analysis found that the average knowledge worker toggles between apps and websites nearly 1,200 times a day — the context-collapse problem is real, and everyone wants to solve it.

So they build connectors. Twenty-plus integrations: Gmail, GitHub, Notion, Jira, your CRM. AI synthesizing across all of them.

This works reasonably well for retrieval. Ask the AI what the status of a project is, and it can scan your emails and Slack and give you a passable answer.

But the write side is a mess.

When AI synthesizes context from six different sources, where does that synthesis go? Usually into a chat window you'll never find again. Sometimes into a document inside whatever app generated it, locked in its format. The intelligence surfaces for a moment and then disappears. Next time you ask, it reconstructs everything from scratch, which takes the same ten seconds and produces a similar answer, because nothing was remembered.

Pull from everywhere. Write to nowhere.

The useful version

The most useful version of "your inbox knows everything" isn't to make your inbox smarter. It's to let the intelligence flow out of the inbox into something you actually maintain.

Here's a concrete version: you had a difficult conversation with someone on Tuesday. The follow-up email arrived Thursday. Your AI reads both, notices a commitment you made, and proposes a task — visible, reviewable, stored alongside your existing notes on that person. You confirm it in two seconds. Now it's in your knowledge base, not in a search index somewhere that you'll have to discover by accident.

That's not a particularly ambitious scenario. The pieces technically exist. What's missing is the write target — a knowledge base structured enough to receive typed information from an external source. Something that knows a person is a person, a task is a task, a preference is a preference. Something that can be written to by a machine and read back by a human without the human needing to remember which folder they put it in.

Most notes apps weren't designed for this. They're optimized for human writing: flexible block structures, freeform layouts, nested pages. The same properties that make them comfortable to write in make them unreliable for a machine to write to. An AI that updates your Notion database will succeed about 80% of the time and produce something strange the other 20%. Over time, strange accumulates.

What "write to one place" actually requires

The knowledge base that receives AI writes needs structured types — not just text, but entities with known shapes. It needs every AI write to be visible and reviewable before it's saved, because a silent edit is just a new way to lose context. And it needs to be in a format you own, something that doesn't disappear if the service does.

These aren't exotic requirements. They're what you'd want from any system you trusted with important data.

Gmail got lucky. It became your life's archive by accident, because your landlord and your airline and your doctor all agreed that email was the right medium. You didn't choose it as a knowledge store. It just is one.

The thing worth building is a system that can receive what email accidentally collected — and return it to you as something more useful than a search result.

---

Asgeir Albretsen is the founder of Harbor.