Share the Graph, Not the Deck

February 2026

"A 5-minute conversation now produces a full presentation. When that's true, you stop thinking about decks and start thinking about the dependency graph of what you know."

The 5-Minute Conversation That Produces a Deck

It's Not About Instant Generation

The key insight isn't that you can generate a deck instantly — it's that a 5-minute conversation with an agent is now enough to produce a full presentation. You talk through the idea, riff on structure, and the document just falls out of the conversation.

The artifact is a byproduct. The thinking is the work, and 5 minutes of thinking is often enough. This can happen often — you can have these conversations daily.

The Old Way: Linear Decks

Intro Concept A Concept B Demo Summary

Hours of work per deck. Locked into one sequence. Can't reuse pieces.

Move Up the Stack

When a conversation produces a presentation, you stop thinking about making decks and start thinking about the structure above them. Just like when coding got fast, you started thinking about architecture.

The structure above presentations is the dependency graph of the knowledge they contain.

The New Way: Knowledge Graphs

When you can have a 5-minute conversation and get a presentation, you accumulate them fast. The question becomes: how do these relate? Share the graph of related presentations — with explicit dependencies between them.

Full Workflow: "Build an App"
↑     ↑
Composing Tools Agent Patterns
↑     ↑     ↑
CLI Basics Git Markdown

Each node is a standalone presentation. The graph shows how they compose.

Why This Matters

Every node can be taught by itself and reused in other flows. The "CLI Basics" node serves your curriculum but also someone else's completely different one.

When you share, you make this clear by sharing the graph — so people see both the piece and where it fits.

A Taxonomy of Knowledge Nodes

Four Node Types (for Coding Knowledge)

Not prescriptive — this is what emerged in my own graph. Yours will differ.

Foundational

The primitives. CLI tools, languages, protocols, data formats. Things that exist independent of any workflow.

e.g. "Git fundamentals", "JSON & YAML", "HTTP basics"

Orchestration

How you use the foundational pieces together. Patterns for combining tools, configuring systems, building pipelines.

e.g. "Composing CLI tools", "Agent + extension patterns"

Workflow

End-to-end processes. The full path from intent to outcome, drawing on foundational + orchestration nodes.

e.g. "Ship a feature with Pi", "Deploy to production"

Authoring

Meta-knowledge about creating the knowledge artifacts themselves. How to write, present, structure, share.

e.g. "Vibe-coding presentations", "Writing good READMEs"

Example: A Coding Curriculum Graph

CLI Tools Git Markdown Compose Tools Agent Patterns Ship a Feature Full Project Lifecycle Presentations foundational orchestration workflow

Key Property: Reusability

The "Git" node appears in your coding curriculum, but also in someone else's "Open Source Contribution" curriculum, and in another person's "Technical Writing" flow.

The node doesn't belong to one graph. It lives at the intersection of many. That's the whole point.

Graph Overlays: A New Communication Protocol

Why This Structure Helps

People always say I'm a bit scattered — multiple ideas running in parallel, connecting things across domains. That's not a bug, it's a graph. The problem was never the thinking — it was that linear formats (decks, docs, emails) can't represent parallel threads with dependencies.

A knowledge graph makes the structure explicit. Instead of flattening 5 concurrent ideas into a single sequence, you show them as what they are: nodes with edges. People can enter anywhere, follow dependencies, and build their own path.

This site — p10q.com — is itself a live DAG. About is the foundation node at the bottom, Learning Resources sits on top, and everything else branches upward. Click any node on the home page and preview it without leaving the graph. The structure is the interface.

Who Benefits

Humans

Multi-threaded thinkers can finally share how they actually think. Receivers can explore at their own pace.

LLM Agents

Agents can consume a graph's feed.json to understand what you know, what's new, and what depends on what. No ambiguity.

Teams

A team's knowledge becomes a shared, browsable graph — not a wiki that nobody reads.

The Graph Overlay Handshake

When you pass knowledge to someone, it's a "graph overlay" — your graph laid on top of theirs. The overlap is where you already agree. The new edges are the knowledge transfer.

your_graph ∪ my_graph → shared_understanding
Δ(new_nodes) → the actual knowledge transferred

Browsing Common Overlaps

In the future, knowledge sharing is mostly about browsing overlapping graphs.

Your graph
A B C D
Their graph
A E C F

Shared nodes: A and C — the overlap is where collaboration happens

Could This Become a Norm?

Imagine if sharing knowledge meant sending a graph overlay. Not "here's my 40-slide deck" but "here's my knowledge graph — it shares 6 nodes with yours, here are the 3 new ones."

This works for teams (onboarding = "here's the team graph, you already know these nodes"), courses (a syllabus is a DAG, not a list), and any knowledge transfer.

The format is simple: a JSON file with nodes and edges. The rendering is up to the reader. The structure is the protocol. Subscribe to someone's graph and receive only the delta — new nodes, updated nodes, new edges. Respect people's attention.

Subscribe to Graphs, Receive Deltas

Versioned Knowledge

Every graph publishes a feed.json with a version number and per-node update timestamps. When something changes, the version increments.

{ "version": 3, "lastUpdated": "2026-02-19",
  "nodes": [
    { "id": "pi_is_all_you_need",
      "updated": "2026-02-01" },
    { "id": "graph_overlays", ★ new
      "updated": "2026-02-19" }
  ] }

The Delta Email

A subscriber gets a daily (or weekly) diff showing:

NEW New nodes added to the graph
UPD Updated nodes with what changed
EDGE New connections between existing nodes

Not "here's my whole blog again" — just the Δ. Respect people's attention.

The Stack Keeps Moving Up

Subscribing to graph deltas next
Browsing overlapping graphs soon
Generating dependency graphs now
Vibe-coding presentations easy
Manually making slides old

Each layer becomes trivial, so you think about the layer above it.

Implementation

It's simple enough to build today:

1. A presentations.json defines nodes + edges

2. A build script generates the graph view + feed.json

3. A Lambda diffs versions daily, emails subscribers the Δ

4. Subscribers see new/updated nodes with context

This site (p10q.com) already does steps 1–2. The subscription layer is next — a Lambda that diffs feed.json versions and sends subscribers just the Δ: what's new, what changed, what got connected.

This presentation is itself a node in the graph you're browsing. On p10q.com, you can tap around the graph and preview any node without leaving. The graph is the interface — not a list of links, but a structure that shows how ideas connect. Maybe this becomes how people share what they know.