Docs · beyond memories

Three things that
are not memories.

A memory is true until something changes it. A task is owed until it is done. A measure is a number that only means anything as a series. Folding any of them into the memory table would corrupt ranking, so they sit beside it with their own rules.

Tasks #

When somebody asks to be reminded weekly, or to be chased on Friday, this is the answer rather than whatever scheduler the agent's host happens to have.

$ ns-brain task add "rotate the RAG box certificate" --when "every monday 9am"
$ ns-brain task add "ask about the context budget" --owner agent --when "tomorrow 9am"
$ ns-brain task list · --all # closed ones too
$ ns-brain task done 34 · drop 34 · snooze 34 --until "in 2 hours"
$ ns-brain task todo --out TODO.md

--when understands "in 3 hours", "tomorrow 9am", "every monday 9am" and "daily 17:00". A recurring task reopens on its schedule rather than being re-added. --owner user|agent records who owes it, which is the difference between a reminder for the person and a job the next session picks up.

How delivery works and what it cannot do

Delivery is pull, on the hooks that already run. A due task appears in the brief and in the prompt hook. So a task on a project nobody opens for a fortnight waits a fortnight. Say that when you set one, because the person may have meant an alarm and an alarm is not what this is.

A task nobody acknowledges is offered again and after a few times it says so and asks to be kept, snoozed or dropped. A reminder that keeps arriving with no way to answer it becomes a block people learn to scroll past, which is the failure mode every signal in the brief is protected from.

Measures #

A memory answers what is true. A measure answers what the value was and which way it is going.

$ ns-brain measure add p99 412 --unit ms --note "after the pool change"
$ ns-brain measure series p99 --from 01/06
$ ns-brain measure list

Readings sit outside the search index and outside ranking entirely and that is the point: a number must never outrank a fact. A p99 taken this morning is newer than every decision in the store and if it competed on the same axis it would sit on top of all of them.

Flags: --unit, --label, --date for a reading taken earlier, --note, --why, --source, --confidence, --ref, --scope and --from / --to / --n to bound a series.

Bands and the silence that is the default

$ ns-brain measure limit p99 --high 500 --better down --why "the SLO"

A key with no declared band is recorded and never judged. Most numbers have no right answer and a tool that invents one starts shouting about readings nobody asked it to police. Declare a band and anything outside it appears in every brief until it comes back inside.

The user model #

$ ns-brain whoami
$ ns-brain whoami --observe sources="reads primary sources" \
      --evidence "asked for the RFC rather than a summary, twice"
$ ns-brain whoami --forget sources

Every trait requires evidence. A model of a person built from inference with no citation is a set of assumptions nobody can challenge and this one can be challenged line by line because each line names what shows it.

It is shown on demand rather than volunteered. A brief that opens by describing the person reading it is unpleasant and the model exists to shape how work is done rather than to be recited.

Objections and the ladder #

$ ns-brain object 412 --why "measured the opposite on the staging box today"
$ ns-brain ladder 412

An objection records disagreement with a memory without deleting it and --why is the half worth keeping: the reasoning outlives the verdict. The ladder shows where a memory stands, from stated, through hardened by repetition and evidence, to law. It is the honest answer to "how sure is this, really" and it is computed rather than declared.

Capture and utterances #

$ ns-brain capture "no, I care about context poisoning, not cost"
$ ns-brain utterances

capture stores what the user said before the turn is answered, which is the only moment the exact words still exist. utterances plays them back with how each was classified. Between them they answer the question that otherwise needs a transcript: what did they actually ask for, in their own words, rather than what did the session decide they meant.

Leads #

A small sales pipeline for the projects that have one: stages, values and the next action. It is opt-in and it stays out of the brief entirely until a lead exists, because a CRM nobody asked for is noise in every session of every project that does not sell anything.