At minute nineteen of a *FocusBottle* session I reached for my phone. The light in the bottle dimmed before my hand got there. Which is funny, because I had spent the whole week building a system whose only real purpose was to throw things away.
A dashboard full of noise
Last week I built a screening pipeline. It watches six sources of market noise at once — trending themes, limit-up leaders, volume spikes, foreign flows, the broker seat lists, the news mood — all free, all refreshing every few minutes, all of them insisting that something is happening right now.
I built it as a solution. The first time I opened the finished board I understood it was a problem, and a familiar one: more signals than a person can read, arranged neatly so the whole thing looks like progress.
Every stage exists to say no
The design that fixed it has three stages, and only the first one is generous. Discovery is broad on purpose, because being stingy at the entrance is how you miss the thing you did not think to search for. Everything after that is narrowing.
The second stage turns a name into something usable — a theme becomes the actual list of tickers inside it, fetched fresh each time instead of kept in a hand-maintained table that would rot within a month. The third stage is where the system earns its keep: a handful of numeric conditions applied without sentiment, most of the pool gone by the end and nobody thanked for it.
I had to write that down before I believed it. A stage that does not remove most of what enters it is not doing work, it is decoration. And caching the events for ten minutes rather than streaming them was the same lesson in miniature — nothing here changes on the scale of seconds, so the appetite for real-time was a feeling I wanted, not a requirement the problem ever had.
The part I cannot automate
So I have a machine on my desk whose entire job is discarding things, and I still could not stay with one task. At minute nineteen of a FocusBottle session I reached for my phone — a reflex, not a decision — and the light in the bottle dimmed before my hand got there. Twenty-five minutes and go, no account to create, nothing to set up; the only warning it gives is the one that lands exactly when you need it.
It was doing to my attention what the pipeline does to the market: keeping the inputs few and the exit narrow. Mine is easier to configure and considerably harder to obey.
Few inputs, chosen on purpose
Both systems ended up the same shape. A small number of inputs, each picked deliberately rather than inherited from whatever shouted loudest. A hard filter that throws most of it away and does not apologise for the pile it leaves behind. And a refusal to be real-time about things that were never urgent in the first place.
The dashboard took a week to get right. The twenty-five minutes took a week to stop arguing with. Same lesson, different desk.