Research

Keep AI research in one place.

You open an AI to find something out: to understand a topic, weigh a set of options, get to a conclusion. But real research is never one question. It’s dozens, spread across sessions, days and different tools. And the moment you need what you found, it’s somewhere you can’t reach.

Nothing about how you research with ChatGPT or Claude changes. Giraffe is where the findings pile up instead of scattering.

Research isn’t one question

A serious investigation happens in pieces. You ask, read, follow a link, ask again. You open a fresh chat tomorrow, then switch tools for the next stretch. Each session answers a little, but nothing adds up on its own.

Three weeks in, you have thirty conversations and no single place that says what you actually learned. So when you need the number, the source, or the reason you ruled something out, you scroll back, can’t find it, and just ask again. You end up re-running research you already did.

The shape of the problem is the same whatever you’re actually looking into: sizing up solar panels for the house, a regulation that affects the business, a market before you enter it, a category of product compared down to the details, or a family line traced back a hundred years. Findings pile up faster than one conversation can hold them, two sources disagree often enough that it matters which one you trust, and the conclusion you’d actually defend today keeps shifting as you learn more.

What Giraffe keeps

Giraffe holds the durable part of an investigation: the findings, the sources behind them, the conclusions you reached, and the questions still open. It’s organized under one project you can open any time. When you land on something worth keeping, you tell your AI to save it: “Add that to my family history project.”

Later, you open the project and re-read where things stand instead of re-running the search. One word, an archive reference, a surname, a date, finds it across every project you keep, not just this one, so you don’t have to remember which one it was in. Most of the time that’s the whole job, with no new AI session at all. When you do want to dig further, you pick it up in ChatGPT or Claude from exactly that state. Here is how Giraffe works, step by step.

A months-long investigation, in one place

Say you’re piecing together one side of your family. You work at it in bursts: a session to make sense of a census record, another to turn a scanned certificate into a clean timeline, a late one chasing a name that turns out to be a dead end. Different tools, dozens of small sessions, over months.

Into a single project you save as you go:

Nothing is captured automatically. Giraffe has no view of your conversations: only what you ask your assistant to save, or type in yourself.

Months later you pick it up again and don’t start over. You open the project, see exactly what’s proven and what’s still loose, and keep pulling the thread, reading on your own, or picking it up in ChatGPT or Claude from where you left off. The same holds for any long investigation: comparing options for a considered purchase, understanding a subject deeply, or gathering the evidence behind a decision you need to defend later.

Re-read what you learned. Don’t research it again.

The hard part of research was never getting an answer. It’s holding on to the dozens of answers you already found, with the sources that back them, so the next time you sit down you’re building on what you know instead of rebuilding it.

Written out in full, over eight months: a dozen conversations about solar panels that never became a decision, and the tax-credit number saved on the first evening, dated, is the one you go looking for at filing time.