Signal Intelligence Basics

Signal Intelligence is the layer of River that decides what deserves your attention. Every email arriving, every calendar event approaching, every task aging — they get scored for relevance to you, and the things that matter surface before you ask.

This is what makes the difference between River as a reactive tool and River as a system that actively works for you.

The three pathways#

Signal Intelligence runs on three pathways at three speeds.

Reactive — fires within seconds#

When something new lands in your Dataspace, the reactive pathway scores it immediately for salience. If the score is high enough, it produces a memory and surfaces a signal — without waiting for any agent to process it.

Score factors:

• Who sent it (closeness in your relationship graph)
• How recently (timeliness)
• Whether it matches your active projects
• Whether it requires action
• Whether you're likely to care based on past behavior

A high-salience email from a known investor with the word "term sheet" hits the surface within seconds of arrival.

Consolidation — runs in batches every few hours#

Some patterns only appear when you look across many events. The consolidation pathway runs in batch mode every few hours and detects:

• Clusters of related emails that together mean something
• Stale follow-ups (you said you'd reply; the thread is silent)
• Overdue commitments
• Patterns across multiple sources (same person mentioned in 5 places this week)

Single events look ordinary. Patterns across them are where the system gets useful.

Synthesis — one LLM call after consolidation#

The synthesis pathway uses one bounded LLM call after each consolidation run to reason across the batch. It produces suggestions that pattern-matching rules could never express:

• "You have a board meeting tomorrow and this email chain has gone silent — you may want to surface a follow-up."
• "Three different signals this week point to risk in the Acme account — worth a check-in."
• "You haven't logged any calls with [advisor] in 60 days; you usually do every 3 weeks."

This is where the system goes from rule-based to genuinely intelligent.

Where you see signals#

Signals surface across four UI layers, each appropriate for a different urgency:

You can turn each surface on or off. The Tray is opt-in by default; the inline panel is on.

Why this exists#

Most apps make you triage. Inbox zero. Standup. Sweep through tasks. The unspoken assumption is that you are responsible for keeping context.

Signal Intelligence flips that. The system holds context for you and surfaces what matters at the moment it matters. You spend less time managing your attention and more time using it.

How it learns about you#

Signal Intelligence reads your System.UserProfile (which you can edit) plus your behavior — what you open, what you ignore, who you respond to fast. The longer you use River, the better its calibration.

You can also tell it directly: "stop surfacing things from this sender" or "always flag emails from [contact]." Those preferences are stored in your profile.

What it never does#

• Auto-send messages
• Auto-resolve tasks
• Mutate state outside your awareness
• Train on your data

Signals are recommendations. You're always the one who acts.

Tips#

• Spend a minute filling out your User Profile in the first week. The signal quality improvement is dramatic.
• Use the SuggestionPanel inline; it's the lowest-friction surface.
• Treat the InboxApp intelligence channel as your "morning triage of triage" — review it once daily, action what's worth actioning.