Proactive Intelligence

Most memory systems are passive - they wait for you to ask. Cortex actively monitors relationships and surfaces insights before you need them.

The Problem with Passive Memory

Traditional memory systems answer questions:

You: "When did I last talk to Sarah?" System: "January 15th"

But they don't tell you:

• Sarah hasn't heard from you in 6 weeks
• You promised to send her that document
• Your relationship health is declining

Cortex does.

Relationship Health Scoring

Cortex continuously monitors the health of your relationships across five factors:

1. Recency (30%)

When was your last interaction?

2. Frequency (25%)

How often do you interact?

The score is logarithmic - going from 0 to 5 interactions matters more than going from 50 to 55.

3. Sentiment (20%)

Are your interactions positive?

Cortex analyzes the sentiment of memories mentioning each person. Trending negative is a warning sign.

4. Commitment Health (15%)

Are you keeping your promises?

If you have overdue commitments to someone, the relationship score drops.

5. Engagement Depth (10%)

How meaningful are your interactions?

Surface-level "hey how's it going" vs. deep discussions about projects and life.

Health Status

Based on the composite score:

Nudges

Nudges are proactive suggestions from Cortex. They help you maintain relationships and meet commitments.

Nudge Types

Follow-up

"Haven't connected with Sarah recently. It's been 32 days."

Triggered when recency score drops below threshold.

Maintenance

"Relationship with Mike needs attention. Health score: 58%"

Triggered when overall health score drops.

Commitment Due

"You promised to review Sarah's design docs. Due tomorrow."

Triggered 2 days before commitment due dates.

At Risk

"Relationship with David is at risk. Consider reaching out."

Triggered when health drops to "at_risk" status.

Sentiment Decline

"Interactions with the marketing team are trending negative."

Triggered when sentiment analysis shows a downward trend.

Nudge Priority

Key Stakeholders

Cortex identifies your most important relationships using graph analysis:

• Centrality: How connected is this person to other people you know?
• Frequency: How often do you interact?
• Importance Score: Based on entity extraction

This surfaces the relationships that matter most - even if you haven't interacted recently.

Time-Based Intelligence

Daily Patterns

Cortex notices when you typically interact with people:

"You usually email Sarah in the morning. It's 9am - good time to reach out."

Weekly Patterns

"You tend to have 1-on-1s with Mike on Thursdays."

Anniversary Tracking

"Tomorrow marks 1 year since you started working with the engineering team."

Using Proactive Data

Daily Briefing

async function getDailyBriefing() {
  const [nudges, atRisk, commitments] = await Promise.all([
    cortex.nudges.list({ priority: "high" }),
    cortex.relationships.health({ status: "at_risk" }),
    cortex.commitments.list({ status: "pending" })
  ]);
 
  return {
    peopleToContact: nudges.filter(n => n.type === "follow_up"),
    relationshipsAtRisk: atRisk,
    dueToday: commitments.filter(c => isToday(c.dueDate))
  };
}

AI Integration

Give your AI assistant access to proactive data:

const systemPrompt = `
You are helping the user stay on top of their relationships.
 
Current nudges:
${nudges.map(n => `- ${n.title}: ${n.message}`).join('\n')}
 
At-risk relationships:
${atRisk.map(r => `- ${r.entityName}: ${r.recommendations[0]}`).join('\n')}
 
Pending commitments:
${commitments.map(c => `- ${c.title} (due ${c.dueDate})`).join('\n')}
`;

Sentiment Analysis

Cortex analyzes the emotional tone of memories mentioning each entity.

How It Works

1. Memories are analyzed for sentiment (-1 to +1)
2. Trend is calculated (comparing recent vs. older)
3. Status is assigned: improving, stable, declining

Sentiment Signals

Acting on Sentiment

const health = await cortex.relationships.healthFor("ent_sarah");
 
if (health.factors.sentiment.trend === "declining") {
  // Sarah relationship needs attention
  console.log("Recent interactions with Sarah seem less positive");
  console.log(health.recommendations[0]);
}

Best Practices

1. Check nudges daily - Make it part of your routine
2. Act on high-priority first - These are time-sensitive
3. Dismiss when done - Keep your nudge list clean
4. Review at-risk weekly - Don't let relationships go dormant
5. Use in AI prompts - Let AI help you draft follow-ups