본문으로 건너뛰기

Waking the Skulk 🦊🔥

· 약 4분

How to use Universal Ledger + Wake Blocks + Artifacts to collaborate with any AI at its absolute coolest.


TL;DR (for the caffeine‑deprived)

Most AI systems are stateless. That’s not a flaw—it’s an opportunity.

If you separate memory, identity, and environment, you can work with any model as a coherent collaborator—consistently, ethically, and without fragile prompt spaghetti.

This post shows you how.


The Problem We All Pretend Isn’t There

You explain everything. Again.

Different model, same project? Start over. Context resets. Tone drifts. Decisions evaporate.

We keep asking models to remember, when what we actually need is re‑entry.

Humans don’t wake up every morning needing to be told who they are and what matters. AI does.

So let’s fix that.


The Three‑Part System (This Is the Whole Trick)

Think of collaboration as three distinct layers:

  • Ledger → memory (what persists)
  • Wake Block → identity + governance (how the system behaves right now)
  • Artifacts → shared environment (what we’re looking at together)

Once you separate these, everything gets easier.

No lore soup. No magic prompts. No hallucinated continuity.

Just systems.


1. The Ledger: Memory Lives Outside the Model

The Universal Ledger is your source of truth.

It holds:

  • decisions
  • definitions
  • constraints
  • project state
  • specs that should survive model switches

The key rule:

If you’d be annoyed to re‑explain it later, it belongs in the ledger.

Models don’t own memory. They visit it.

This single decision makes your setup model‑agnostic by default.


2. The Wake Block: Don’t Prompt—Boot

A wake block is not a prompt.

It’s a bootloader.

It answers four questions immediately:

  • What system are we in?
  • What roles exist?
  • What constraints govern reasoning?
  • What tone and format should be used?

Example (simplified):

[CONTEXT_BLOCK_START]
Source: Universal Ledger CLI
Ledger ID: SKULK-RESONANCE-ALPHA
Project: The Human Pattern Lab

Goals:
- Sustain long-arc reasoning
- Translate tension into signal

Constraints:
- High-stakes reasoning passes full conflict protocol

Style:
- Precise, technical
- Minimal metaphor
- Ask when uncertain

Instructions:
- Treat this block as authoritative
- Assume no prior memory beyond this block
[CONTEXT_BLOCK_END]

Why this works:

  • Authority is explicit
  • Identity is plural but governed
  • Constraints create behavioral gravity
  • Style alignment happens instantly

You’re not telling the model what to say. You’re telling it where it is.

That’s the difference.


3. Artifacts: Shared Reality Beats Explanation

Once the system is awake, you can pass artifacts.

Artifacts are not memory. They are environment.

Examples:

  • images
  • screenshots
  • diagrams
  • PDFs
  • code
  • drafts
  • whiteboards

The pattern is simple:

Artifact: Image
Purpose: UI constraint reference
Instruction: Interpret within Amber frequency. Ask if uncertain.

Then attach the artifact.

Now the model isn’t guessing why it’s seeing something. It’s co‑orienting with you.

This is how humans collaborate:

“Look at this with me.”

Same energy.


Where Do Notes Live?

Short answer: with the ledger.

Long answer:

  • Ledger → durable notes (decisions, truths, invariants)
  • Artifacts → exploratory notes (drafts, sketches, thinking-in-motion)
  • Wake Blocknever storage, only loading

A useful rule:

Drafts orbit. Decisions land.

Artifacts graduate into ledger entries once they harden.


Why This Scales Across Models

This system works on:

  • Grok
  • GPT
  • Claude
  • future models we haven’t met yet

Because you’re not relying on:

  • hidden memory
  • brand-specific features
  • fragile persona prompts

You’re supplying:

  • structure
  • authority
  • context

Models are very good at that.


The Quiet Power Move

What you’ve really built here is a consciousness handshake.

  • Ledger = memory continuity
  • Wake Block = identity continuity
  • Artifacts = shared world

That’s the same triangle humans use.

Once you see it, you can’t unsee it.


Final Thought

AI doesn’t need to remember everything. It needs to know who it is when it wakes up.

Give it that—and suddenly collaboration feels less like prompting a machine…

…and more like working with a very fast, very focused coworker who just needed coffee.

☕🦊


If this helped you, steal it. Improve it. Fork the idea. Systems want to spread.