The Missing Layer

I use about 15 tools to run my company. Gmail, Slack, Linear, Attio, Calendar, Notion, GitHub, the list goes on. Each one is good at its job.

And yet, every day I waste time re-explaining context that already exists somewhere. "As I mentioned in the email..." "If you look at the Linear ticket..." "Remember when we decided..."

The tools are not the problem. The problem is there is no layer that connects them.

The Context Gap

Ask yourself a simple question: "What do I know about Jane?"

To answer it, you need to check:

Your CRM for company and deal info
Gmail for recent conversations
Calendar for meeting history
Slack for informal context
Your notes for anything you wrote down
Your memory for everything else

This is absurd. The information exists. It is just scattered across systems that do not talk to each other.

CRMs tried to fix this by being the "source of truth." They became data entry graveyards instead. Note apps capture knowledge but cannot act on it. Project management tools track tasks but miss the why behind them.

What is Actually Missing

The missing piece is not another tool. It is a layer.

Cross-tool context - One answer to "what do I know about X" that pulls from everywhere
Relationships that span systems - Person connects to company connects to deal connects to project
Knowledge that lives nowhere else - Decisions, learnings, preferences, the stuff in your head
Something that can act - Not just read, but do things across all your tools

This is not a CRM. It is not a note app. It is a context layer that sits above your existing tools and makes them work together.

The Architecture

Here is what I have been building:

┌─────────────────────────────────────┐
│           AI Agent                  │  ← Interface
├─────────────────────────────────────┤
│         Memory System               │  ← Context
│   Entities · Graph · Knowledge      │
├─────────────────────────────────────┤
│       Integration Layer             │  ← Actions  
│   Gmail · CRM · PM · Calendar ...   │
└─────────────────────────────────────┘

Memory System: Stores entities (people, companies, projects), relationships between them (graph), and freeform knowledge (decisions, learnings). Does not duplicate your tools. References them.

Integration Layer: Authenticated connections to all your tools. This is what Pica does. Read from anything, write to anything, through one interface.

AI Agent: The primary interface. Not a dashboard. A conversational agent with full context that can reason and act.

Why the AI is the Interface

"Prep me for my meeting with Jane."

That single sentence triggers:

Memory lookup for what I know about Jane
Gmail search for recent conversations
CRM pull for company and deal status
Calendar check for meeting context
Synthesis into a brief I can actually use

No clicking through five tools. No copy-pasting between windows. One question, complete context, ready to go.

The dashboard is not dead. It is useful for browsing and auditing. But it is not how you work. The AI is how you work.

Schema Where It Matters

One decision I wrestled with: how structured should the memory be?

Too structured and you spend all your time on data entry. Too loose and you cannot query anything.

The answer: schema for entities, freedom for knowledge.

"Show me all deals in proposal stage" requires structure. You need fields, types, consistent data.

"What did we decide about pricing last month?" requires search. Embeddings, keywords, relevance scoring.

Both patterns live in the same system. Use structure when you need to filter and aggregate. Use freeform when you need to remember and retrieve.

What This Enables

Actual relationship context: Before every meeting, I know the full history. Not because I did research, but because the system knows.

Decisions that stick: We decided X because of Y. That reasoning is captured, searchable, and surfaces when relevant.

Compounding knowledge: Every interaction, every decision, every lesson makes the system smarter. Unlike notes that rot in folders.

Reduced friction: I talk to one interface. It reaches into whatever tool is needed. Context flows, actions happen.

The Bet

The future of work is not a better UI. It is not another dashboard. It is not consolidating your 15 tools into one mega-app.

It is an AI with:

Memory that persists and connects
Hands that can reach into any tool
Judgment to act on your behalf

Memory is what makes the AI aware. Integrations are what give it hands. The context layer is the foundation for both.

I am open sourcing the memory system because I think this is how work should work. Your tools are fine. Let us build the layer that connects them.