Ideas. Evidence. Repeat. The third word is doing as much work as the first two. Reamer isn't built around a single pass from idea to result — it's built around a loop, run repeatedly, across many ideas, for as long as someone is doing research. What actually happens in one pass is its own story. This is about why it's a loop instead of a line.
A decision closes an idea. Knowledge closes the loop.
Every idea ends somewhere — kept or discarded. That's the decision, and it's specific to the one idea in front of you. But something else survives past the decision, whichever way it went: a discarded idea doesn't just disappear, it leaves behind a concrete, specific reason it didn't hold up. That reason is what a researcher actually carries into the next idea.
This is the difference between a loop and a dead end. A process that stops at the decision treats a rejected idea as wasted effort. A process that treats knowledge as the real output treats it as exactly the opposite — a confidently discarded idea, reached fast enough that the cost of finding out stayed low, is a complete, successful pass through the loop. The next idea starts a little sharper because of it, whether or not anyone remembers the specifics of what didn't work.
The same loop runs at two depths at once
Not every idea deserves the same amount of attention, and the loop isn't single-speed. Most ideas get triaged cheaply — implemented, run, checked against realistic costs, and ruled out inside an afternoon, sometimes faster. That's breadth: getting through many hypotheses quickly, most of which won't survive contact with real execution costs, and shouldn't need to.
A smaller number survive that first pass and earn something deeper — cost and slippage resilience checked more carefully, portfolio interaction across multiple instruments, a full robustness pass with Monte Carlo before anyone would seriously consider risking capital on it. That's depth, and it's the same loop, just continuing further because the idea has actually earned the extra scrutiny.
Both speeds matter, in the same process, for the same reason: an idea should get exactly as much attention as its own results justify, not more and not less.
Why this compounds over time
None of this is about any single strategy. It's about what happens across dozens of them, one after another. A researcher who runs this loop repeatedly isn't just accumulating a pile of past backtests — they're accumulating calibration: a sharper sense of which patterns tend to survive realistic costs and which ones tend not to, built from a growing set of ideas that were each actually checked rather than assumed. That calibration is the part that doesn't show up in any single result, and it's the reason the tenth idea through the loop tends to get resolved faster and more confidently than the first.
Repeat
"Ideas. Evidence. Repeat." isn't a closing flourish — it names the actual shape of ongoing research. One pass through the loop produces a decision and some knowledge. The next idea starts with that knowledge already in hand. There's no natural stopping point built into this, and there isn't supposed to be one — the loop is the practice, not a project with an ending.
Full reference: docs · The workflow this maps onto: reamerlabs.com