Here we go again. Someone noticed Jane Street uses OCaml, so naturally we’re discussing a language migration for the client newsletter pie chart.
Let’s be clear: the trading floor is not the marketing department. A system that needs microsecond latency, provable correctness, and 99.999% uptime justifies exotic engineering. A chart that runs once a week and breaks only when nobody’s watching does not.
The pie chart isn’t failing because Excel lacks a strong type system. It’s failing because it has no owner, no documented inputs, and no validation step. Rewriting it in Haskell won’t fix that. Neither will Dockerizing it, or moving it to the cloud, or any other fad du jour. The problem isn’t the tool. The problem is the process.
Here’s how you actually fix the pie chart:
- Name an owner and a backup. Yes, really.
- Define the source data and the exact steps to refresh it.
- Validate totals, dates, labels, and missing values before it hits the newsletter.
- Keep a working copy under version control or an equivalent change process.
- Make failures visible before the client sees them.
None of that requires a new language. It requires basic discipline. But discipline is boring, and we’d rather argue about monads.
Excel is fine for a job that takes ten minutes a week. It’s the lack of ownership that’s unprofessional. The same people who’d sneer at a spreadsheet are often the ones who let it run on someone’s laptop and break every quarter.
And no, Jane Street’s architecture is not a template for every finance task. Trading systems have demanding requirements for speed, correctness, and maintainability. Language choice is part of that, but so are architecture, testing, measurement, and operational rigor. The newsletter chart has none of those requirements. It has one requirement: don’t break.
So by all means, admire the engineering at trading firms. But don’t mistake it for a universal solution. The pie chart doesn’t need a manifesto. It needs an owner.
Leave the trading architecture out of the newsletter meeting.