Method

How the briefing gets made.

A rule engine, a curated library of phrasings, and one job: surface what matters today.

Four steps. Every one of them is auditable. None of them is a guess.

1 · Read
Read the state of your work.
The engine reads your Signal Tasks workspace. Tags become projects. Lanes become status. Assignees stay assignees. Nothing is invented, nothing is interpreted, the briefing only knows what your workspace already says.
Today the only source is Signal Tasks. Other sources will be added when they earn it.
2 · Detect
Look for ten patterns.
Ten rules run across the read model. Work that's blocked. Due or overdue dates. Too much in flight at once. One thing waiting on another. Work picking up speed. A run of things shipped. A project gone quiet. Everything resting on one person. A deadline creeping closer. A blocker that keeps coming back. Each rule is named, each rule is auditable, each rule fires only when its condition is met.
Every rule has a threshold. Every threshold is published. No detection is fuzzy.
3 · Compress
Keep three. Drop the rest.
Insights are ranked by cascade (does this slow other things?), irreversibility (can it be fixed in five minutes?), and proximity (does it matter today?). Each block is capped at three items. Anything below the cap is dropped silently, the briefing doesn't apologise for what it left out.
Silence is also signal. A short briefing means a calm day.
4 · Write
Pick a phrasing. From a library written by hand.
Every sentence in the briefing comes from a curated prose library, dozens of phrasings, each written by a person, slot-filled with the names and numbers from the rule that fired. The engine never generates language. It picks language.
No machine writes these lines. Not today, not in v1. The brand is in the writing.

The ten rules

Ten rules. Every threshold published.

Each rule has a name, a number, and a block it fires into. Nothing fuzzy. Nothing learned.

RuleThresholdBlock

blocked≥ 2 daysNeeds attention

overduepast due, not doneNeeds attention

overload> 8 tasks / one personNeeds attention

dependency-stall≥ 5 days, no activityNeeds attention

momentum-positive≥ 3 closes / 7 daysMoving well

streak≥ 5 closes / 7 daysMoving well

inactive-project≥ 8 daysQuiet risks

single-point-of-failure> 70% on one personQuiet risks

slow-burn-deadline≤ 7 days, < 30% closedQuiet risks

unresolved-recurring-block≥ 3 recurrencesQuiet risks

What the method refuses

Four things the engine does not do.

Not a dashboard.
Dashboards ask you to interpret. The briefing tells you what changed and what matters.
Not a recommendation engine.
The engine names things. It does not decide. The phrase "Suggested focus" is the strongest verb the briefing uses.
Not an AI workspace.
There is no model in the path. No agent. No copilot. A rule engine and a phrasing library, both auditable, both written by hand.
Not exhaustive.
Three items per block. Always three. If five things need attention, the engine ranks by cascade, then irreversibility, then proximity, keeps the top three, and trusts you to find the rest yourself.

See it in motion.

What's in a briefing·Thirty seconds

Read the state of your work.

The engine reads your Signal Tasks workspace. Tags become projects. Lanes become status. Assignees stay assignees. Nothing is invented, nothing is interpreted, the briefing only knows what your workspace already says.

Today the only source is Signal Tasks. Other sources will be added when they earn it.

Look for ten patterns.

Ten rules run across the read model. Work that's blocked. Due or overdue dates. Too much in flight at once. One thing waiting on another. Work picking up speed. A run of things shipped. A project gone quiet. Everything resting on one person. A deadline creeping closer. A blocker that keeps coming back. Each rule is named, each rule is auditable, each rule fires only when its condition is met.

Every rule has a threshold. Every threshold is published. No detection is fuzzy.

Keep three. Drop the rest.

Insights are ranked by cascade (does this slow other things?), irreversibility (can it be fixed in five minutes?), and proximity (does it matter today?). Each block is capped at three items. Anything below the cap is dropped silently, the briefing doesn't apologise for what it left out.

Silence is also signal. A short briefing means a calm day.

Pick a phrasing. From a library written by hand.

Every sentence in the briefing comes from a curated prose library, dozens of phrasings, each written by a person, slot-filled with the names and numbers from the rule that fired. The engine never generates language. It picks language.

No machine writes these lines. Not today, not in v1. The brand is in the writing.

Ten rules. Every threshold published.

Each rule has a name, a number, and a block it fires into. Nothing fuzzy. Nothing learned.

RuleThresholdBlock

blocked≥ 2 daysNeeds attention

overduepast due, not doneNeeds attention

overload> 8 tasks / one personNeeds attention

dependency-stall≥ 5 days, no activityNeeds attention

momentum-positive≥ 3 closes / 7 daysMoving well

streak≥ 5 closes / 7 daysMoving well

inactive-project≥ 8 daysQuiet risks

single-point-of-failure> 70% on one personQuiet risks

slow-burn-deadline≤ 7 days, < 30% closedQuiet risks

unresolved-recurring-block≥ 3 recurrencesQuiet risks

Four things the engine does not do.

Not a dashboard.

Dashboards ask you to interpret. The briefing tells you what changed and what matters.

Not a recommendation engine.

The engine names things. It does not decide. The phrase "Suggested focus" is the strongest verb the briefing uses.

Not an AI workspace.

There is no model in the path. No agent. No copilot. A rule engine and a phrasing library, both auditable, both written by hand.

Not exhaustive.

Three items per block. Always three. If five things need attention, the engine ranks by cascade, then irreversibility, then proximity, keeps the top three, and trusts you to find the rest yourself.