Everything gets in. Almost nothing gets closed.
190 observations have come in through ordinary conversation. Zero have been triaged. That is not a malfunction — the inbox is doing exactly what it was built to do. It is the honest shape of what happens once writing something down stops costing anything.
Capture and closure, by category.
Tasks close. Almost nothing else does, and the gap between them says something about what a due date is actually doing.
- Observations190 captured. 0 triaged.
- Decisions112 recorded. 0 outcomes ever revisited.
- Ideas51 captured. 2 acted on.
- Tasks506 created. 211 completed.
Tasks close at nearly half. Everything else closes at close to never — and it isn't because observations and decisions matter less. Some of the 112 decisions in there are the ones a task in this list actually traces back to. The difference is that a task has a due date sitting in front of it, and a due date is a forcing function that asks to be dealt with. An observation has none. It just sits, correctly recorded, waiting for a person to decide it's done being useful.
Friction used to be a filter.
Before this, writing something down cost something — opening the right tool, finding the right field, typing it out by hand. That cost was doing real work. It meant only the observations worth the trouble actually got recorded. Everything else stayed in your head and quietly evaporated, and that was fine, because it hadn't been important enough to survive the friction.
StratOps removes that cost on purpose. You say what you noticed, in the sentence you were already saying to Claude, and it lands. That is most of the pitch. But the filter that used to run at the moment of writing doesn't relocate when the friction goes away — it just stops running anywhere. Nothing has replaced it. So everything gets in: the observation worth building a decision on, and the one that was just thinking out loud in Sarah's direction. Both look identical sitting in the inbox. Telling them apart is still real, undone work, and now there is 190 of it instead of the dozen that would have survived the old friction.
What triage means, mechanically
Deciding an observation is done being useful — acted on, folded into something else, or no longer worth carrying. triage_inbox_item is the tool that does it, and reading the inbox back costs nothing: it never prompts for approval, only writing does. The mechanism isn't the obstacle here. Nothing makes it get used is.
Triaging something used to make it disappear.
Hiding a triaged item from the inbox is the point of triage. Erasing it from existence was not, and for a while that was what actually happened.
The inbox is built to hide items once you've dealt with them — an observation you've already handled shouldn't keep competing for attention with the 190 you haven't. That part is working as designed. What wasn't designed, and shipped anyway for a while, was that hiding was the only thing that happened. Nothing else in the system could read a triaged item back. Once you triaged an observation it wasn't archived, wasn't linked, wasn't searchable — it was just gone, in every sense that mattered.
That is worse than never triaging anything. An observation sitting untouched in the inbox is at least honestly visible as unfinished. A wrongly triaged one looked identical to a correctly triaged one, because neither showed up anywhere, ever again. There was no way to check your own work.
That's fixed now. A triaged observation still leaves the inbox — that's still the entire point — but it no longer leaves the record. It can be found again: by searching for it, by following whatever task or decision it fed into, by asking for it directly. Leaving the inbox and leaving existence used to be the same event. They aren't anymore.
The zero is real, not a placeholder for "not measured."
Elsewhere, StratOps prints "not measured" rather than a number it can't stand behind — calibration refuses to score anything until five decisions have closed, and says so instead of guessing. This page is the other case. 0 of 190 observations triaged isn't the system failing to compute something. It's a real count, checked, of a person who hasn't gotten to it yet — told to you plainly instead of rounded up to look tended.
The same shape shows up with ideas, where 2 of 51 have gone anywhere, and with signals, which don't even arrive unless somebody says them out loud.