Pierre Garreau
Most personal finance advice collapses into two camps: buy index funds and don’t look at it, or an endless stream of tactical opinions that aged poorly. What rarely gets done is a clean, side-by-side comparison of the canonical frameworks — the ones that have enough history and intellectual backing to be taken seriously.
This explorer covers six of them:
The data covers 2017–2024, with monthly returns, annualised CAGR, volatility, max drawdown, Sharpe ratio, and three-scenario projections to 2030.
A few things stand out when you put all six on the same ruler:
Buffett 90/10 wins on raw return (13.0% 10Y CAGR) but carries the most volatility and the deepest drawdowns. It is effectively a bet that US equities mean-revert upward — which they have, until they don’t.
Golden Butterfly has the best Sharpe ratio (0.85) among the six. Adding Small Cap Value to the Permanent Portfolio’s skeleton picks up the size and value factor premiums while keeping the all-season safety net intact. 2022 was -11.5% versus -24.3% for All Weather.
All Weather underperformed in the 2022 rate-shock scenario more than its thesis would suggest. Long-duration bonds and equities fell together — the exact correlation breakdown the strategy was designed to avoid. The 40% TLT allocation was the culprit.
The 60/40 is not dead, just cyclical. Its 8.5% 10Y CAGR with a 0.79 Sharpe still beats most actively managed funds after fees.
This portfolio explorer was my first ever project built with Claude Code — Anthropic’s agentic coding tool that runs in the terminal. The experience is worth writing about.
Session one was entirely about the explorer itself. I described what I wanted — six famous portfolios, historical data, interactive charts, a compare view — and Claude Code wrote the whole thing in a single session. The self-contained HTML file you see embedded above is that output, essentially untouched. Chart.js, the data, the CSS design system, the navigation logic, the scatter plots and drawdown charts — all of it produced in one go. I didn’t write a single line of that file.
Session two was this blog. This Hugo site had been sitting dormant since 2020, pinned to Hugo 0.70. I asked Claude Code to parse the repo, update Hugo to the latest version, set up Docker for local development, and then write a blog entry from the portfolio explorer I had just built. What I got back: a docker-compose.yml, a README.md, fixes for three breaking changes introduced across four years of Hugo releases — google_analytics_async.html removed, .Site.DisqusShortname gone, .Site.IsServer renamed — and this post.
I’ve been writing code for twenty years. I know what it feels like to pick up a project you haven’t touched in five years — the archaeology of it, the friction of getting it running again before you can do anything useful. That part, the part I like least, took about two minutes.
Whether that’s exciting or unsettling probably depends on where you sit. For me it was both.