We build the tools we wanted back when we were the ones trading.

AlgoBeam started in a one-room office above a bakery in Tallinn, with two monitors, a whiteboard nobody wiped, and an argument that ran for most of a winter. Maren had spent nine years running a systematic futures book and had rebuilt the same backtester three separate times — once in R, once in C#, and once in a spreadsheet so large it took forty seconds to recalculate a single parameter sweep. Tobias had written the execution layer for a crypto market maker and was tired of indicators that quietly behaved one way in research and another way in production.

The argument was about trust. The tools on the market split cleanly into two camps. On one side, charting products that were gorgeous and completely unfalsifiable: a ribbon turned green, and you were never told which bars made it turn or how often it had been wrong. On the other, quant stacks that were rigorous and close to unusable — a week of plumbing before you could look at a single chart. We wanted a third thing: a chart you can click through to the underlying bars, a backtest that warns you when you are fitting noise, and a language you can take with you when you leave.

So the language came first. AlgoBeamScript is deliberately small and deliberately boring — no hidden lookahead, no repainting by accident, one obvious way to express a rule. We shipped the transpiler in the spring of 2022 as AlgoBeamTS, MIT licensed, so that the part that computes your numbers could never be held hostage by a subscription. The platform grew around it: screeners, alerts, heatmaps, calendars and a strategy builder, all reading the same data and running the same indicator code. Quant, our research assistant, arrived last, and only once it could cite the exact candles behind every sentence it wrote.

Five years in, the rule we started with has not moved. If we cannot show you where a number came from, we do not print it.