I wanted a tool that would read the market for me and automate all the tedious stuff the finance bros do all day. So I built one, and gave it the most finance-bro name I could think of. It's three engines now, and every one of them grades itself in public.
The engines
Every engine publishes the same things: what it picked, why, how it scored against its own benchmark, and where it was wrong. Nothing here is behind a login. Benchmark-relative figures above are raw spreads, not risk-adjusted - each engine's methodology page explains what that leaves out.
Context
Economic Dashboard Not a scoring engine - a plain-English read on the US and Canadian economies. Official indicators from FRED, Statistics Canada and the Bank of Canada, each with what it means and where it sits in its own history. Reads the state, never predicts the date. Enter →What Brad is
One scoring system, pointed at three different corners of the market. Each engine is the same discipline applied somewhere new, and each one runs the same shape of sweep: scan the universe, cull the weak, deep-score the survivors on factors that measure genuinely different things, then send an AI research agent down the top of the list to read what a spreadsheet never sees.
What differs between them is the universe, the benchmark, and how long a position is held. Stocks are measured against the S&P 500 on a multi-month hold. Crypto runs its own faster clock and is benchmarked to Bitcoin, not the S&P, because measuring a coin against an equity index tells you nothing. Small caps hunt the part of the market that analysts have largely stopped covering.
They also all grade themselves the same way. Each publishes its own Information Coefficient - the academic test of whether a score predicts anything at all - alongside whether a strategy built on that score actually came out ahead. Those are two different questions, and most tools quietly mash them into one number that is easy to game. Brad keeps tracking the names it told you to sell, so the sell rule has to earn its keep too. It is built to try to prove itself wrong and show the receipts either way.
Meet Brad
I'm a software developer, not a finance person, and I want that clear from the start. About a year ago I decided to actually understand investing, and the thing about me is I learn by building. Code won't let you stay vague. It makes you answer the annoying questions out loud: what makes a stock worth a look? What's an instant red flag? What can you toss and never think about again?
So I teamed up with AI and went deep on the academic side, hundreds of research papers on what actually predicts whether a stock does well, some going back the better part of a century, and boiled the findings that hold up under real scrutiny down into a single scoring system. Every piece traces back to published work, not a gut feeling. Then came the part that ate months: tuning the weights against history and prying apart every spot where two factors were quietly measuring the same thing, because counting profitability in five places just fools you into thinking you have five signals when you've got one. That grind is what turned a stack of papers into a system that actually works, and it's the only thing I use to decide what I hold.
Questions, feedback, or a bug?
Tear it apart or reach out about anything, I read every message.
Get in touch