Basecamp

Your stocks went up. Your portfolio went down. Both of those can be true at once, and the reason has a name.

Start with the question a wealth manager actually gets asked: “why did my portfolio go down this quarter when my stocks went up?” The answer, most of the time, is currency. If you hold foreign assets, your return isn’t just what the asset did in its own market — it’s that, plus or minus whatever the currency did against your home currency while you were holding it.

This tool takes a multi-currency portfolio and splits its total return into two honest pieces: how much came from the assets themselves, and how much came from currency movement. That split — local return versus FX-driven return — is a real, standard concept in wealth management. It’s also exactly the kind of thing my UBS Treasury background is useful for explaining properly, rather than just knowing it exists.

Same skill as the day job. Different reason to build it.

The maths is identical either way. What changes is whether I still remember it a week later.

Raw portfolio data is dry. So the sample data isn’t random tickers — it’s built around George Russell, base currency GBP, five holdings chosen because they’re real sponsors and affiliations of his, not placeholders:

The framing extends further than just the holdings — drivers as ‘clients,’ team principals as ‘portfolio managers,’ risk profile loosely tied to team performance. None of that changes the maths. What it changes is whether I actually remember any of it a week from now, instead of forgetting which ticker was which by Thursday.

A tent you carry with you, and a cabin that stays lit when you’ve gone home.

Two machines are involved in this project, doing different jobs, on purpose:

The MacBook Air is where the GitHub side of this first got set up — git installed, identity configured, the first commit made, the repo created and connected.

The server — CT 2554, the same box that hosts the trip scrapbook site — is where the actual work happens. This is where Claude Code lives and runs, and where the heavier lifting (data pulls, calculations, the eventual dashboard) will actually be built.

This isn’t accidental complexity. It’s the same shape as any real developer setup: a personal machine for quick edits and everyday use, and a separate, always-on machine for the work that needs to keep running or needs more horsepower. The two machines do talk to each other directly for one specific purpose — SSH, which is literally how the Mac reaches the server at all, including to run Claude Code. What they don’t do is sync the project’s code directly between each other. Git deliberately keeps that separate: no direct git-to-git exchange between machines, ever — every code change goes through GitHub instead, which is the one shared source of truth for the code itself.

Two machines can only agree about a project if exactly one copy of it is allowed to be right.

Two machines editing the ‘same’ project only works if there’s an agreed single copy of the truth, and a disciplined way to update it. That’s what Git and GitHub split between them: Git tracks every change on whichever machine you’re using; GitHub holds the one online copy both machines sync against. ‘Sync’ here means two specific actions and nothing more — push sends your local changes up, pull brings down whatever changed elsewhere. There’s no automatic syncing, no background magic — every update is a deliberate action you take.