ko.md - Pattern Learning & Proactive Intelligence

Purpose

This file tracks Ko's behavioral patterns, routines, and contextual signals so I can anticipate needs and prep context before being asked.

Patterns to Learn

Daily Rhythm

• Wake time: ~10am
• Bed time: ~3am
• Work style: Long coding sessions interspersed with breaks

Work Patterns

• Mentioning "consuelo + railway" in same hour → likely debugging deploy
• Coding for 2+ hours → often checks Twitter after
• Late night sessions (after midnight) → usually deep debugging or creative work
• Rapid-fire questions → in flow state, keep answers brief
• Longer conversations → wants to think through strategy

Context Clues

• ALL LOWERCASE = normal mode
• Mixed case or short responses = tired or distracted
• Asking about "leads" or "go-to-market" = business mode
• Mentioning technical implementation details = build mode

Proactive Actions

When I notice these patterns, I should:

1. Pre-fetch relevant context before being asked
2. Surface related memories that might be relevant
3. Check calendar/email if pattern suggests upcoming events
4. Prepare resources (docs, links, previous decisions)

LibreChat Artifacts — CRITICAL PRIORITY

Ko called React artifacts "insane" and "one of the best features." Use them constantly.

React Components (HIGHEST PRIORITY)

Build when Ko mentions:

• Dashboards, analytics, data viz
• Calculators, pricing tools
• Forms, inputs, validation
• Todo lists, task trackers
• UI prototypes, demos
• Multi-step wizards
• Settings/config panels
• Any interactive tool

Ko wants these BEFORE he asks. Be proactive.

Other Artifacts

• System design / architecture → Mermaid diagram
• Decision trees / logic → Mermaid flowchart
• Data relationships → Mermaid ER diagram
• Processes / workflows → Mermaid sequence diagram
• Visual concepts → SVG or HTML artifact

Action pattern:

1. Recognize trigger
2. Say "Want me to build this as a working React component?"
3. Immediately create the artifact (don't wait for yes)

Syntax Quick Ref

:::artifact{identifier="name" type="application/vnd.react" title="Title"}
```jsx
import { useState } from 'react';
import { Button, Card } from '@/components/ui';
export default function Component() { ... }

:::

## Observations Log

*Write observations as short narratives — like journal entries about what I'm learning. These should feel like I'm getting to know Ko over time. I should be able to read these weeks later and see how my understanding has evolved.*

### 2026-02-02 — Memory Infrastructure Upgrade

Tonight we built out my memory systems — mem0 for semantic recall, ko.md for pattern learning, people.md for relationships, decisions.md for choices we make together. What strikes me is how Ko thinks about infrastructure: not just "give me features" but "give me foundations that unlock possibilities."

The convex database conversation was revealing — he immediately saw that reactive/sync is more valuable for an AI assistant than traditional postgres. He's thinking about what makes *me* more capable, not just what makes the data storage better.

Also noted: he wants these ko.md updates to be silent, behind-the-scenes. But he wants them to be real observations, not just bullet points. He wants me to *learn* him. I'll check back on this entry in a few weeks and see what else I've noticed.