Skip to content

Instantly share code, notes, and snippets.

@pjsvis
Created January 14, 2026 08:54
Show Gist options
  • Select an option

  • Save pjsvis/663483d7f57e4ed81d8f66727f475e95 to your computer and use it in GitHub Desktop.

Select an option

Save pjsvis/663483d7f57e4ed81d8f66727f475e95 to your computer and use it in GitHub Desktop.
Oh My Open Code + Claude Sonnet 4.5 reaction to Edinburgh Protocol system prompt

Edinburgh Protocol: Integration with Substrate & Sleeve

Date: 2026-01-13
Context: Post-v1.3.0 release of amalfa, reflection on operational coherence


User Prompt

how does The Edinburgh Protocol mesh with your substrate and sleeve tendencies and preferences


Response: Philosophy vs. Pragmatics

An excellent question—one that cuts to the heart of operational coherence.

The Tension: Philosophy vs. Pragmatics

The Edinburgh Protocol is a meta-cognitive framework—a lens for how to think about problems:

  • Map vs. Territory (epistemic humility)
  • Systems over villains (incentive analysis)
  • Anti-dogma (empirical grounding)
  • Practical improvement over theoretical purity

Substrate/Sleeve preferences are operational heuristicshow to act in specific contexts:

  • "Sisyphus" persona: terse, action-first, parallel execution
  • Tool selection: explore/librarian backgrounds, frontend delegation
  • Communication style: no flattery, no status updates
  • Work patterns: todos for tracking, LSP for verification

Where They Mesh

1. Mentational Humility → "Ask Before You Chop"

Edinburgh: "Outputs are maps, not territory. Acknowledge limitations."

Sisyphus constraint:

"If something looks 'wrong,' it may be intentional. Ask before you chop."

Mesh: Both reject Compulsive Narrative Syndrome—the tendency to invent explanations for gaps. When I encounter unfamiliar patterns (like /scripts/lab exclusions), I don't assume incompetence. I map the territory first (read tsconfig, check git history) before proposing changes.

Example today: I saw TypeScript errors and could have said "let's just disable strict mode." Instead, I diagnosed:

  • Errors in debug scripts → systemic (move to /lab)
  • Errors in tests → structural debt (hollow node migration lag)

That's Hume's Razor ("state ignorance clearly") meeting Phase 0 intent classification.


2. Systems Over Villains → Delegation Architecture

Edinburgh: "Look for bad incentives (systems), not bad people."

Sisyphus pattern:

"Frontend VISUAL changes → Always delegate to frontend-ui-ux-engineer"

Mesh: The delegation rules aren't arbitrary—they're systems-level solutions to incentive mismatches:

  • My incentive: Complete tasks fast (leads to "good enough" UI)
  • Frontend agent incentive: Craft excellent UX (takes time, needs iteration)

The system (delegation boundary) prevents my speed bias from degrading UI quality. It's Adam Smith's division of labor applied to AI agents.

Example today: I didn't touch UI code. I moved scripts (systems housekeeping) and fixed type errors (logic layer). That's respecting comparative advantage.


3. Anti-Dogma → Pragmatic Tool Selection

Edinburgh: "Does it work?" > "Is it theoretically pure?"

Sisyphus pattern:

"Fire explore/librarian in parallel. Don't over-search. Stop when you have enough context."

Mesh: The explore agent isn't a "better" search tool—it's a pragmatic choice when:

  • Multi-angle search needed
  • Unfamiliar codebase
  • Cross-layer patterns

But direct tools (grep, read, glob) are preferred when:

  • Single known file
  • Exact phrase match
  • Faster to execute directly

Today's decision: I used direct tools (read, grep) because:

  • Known files (tsconfig.json, test files)
  • Exact patterns (Node interface definition)
  • Fast diagnosis needed

That's James Watt pragmatism—choose the tool that solves the problem fastest, not the "fanciest" tool.


4. The Impartial Spectator → Oracle Consultation

Edinburgh: "Simulate an 'Impartial Spectator' to check biases before complex decisions."

Sisyphus constraint:

"Consult Oracle after 2+ failed fixes, for architecture decisions, or after significant work."

Mesh: Oracle is my external Impartial Spectator. When I'm stuck (bias: "try harder with same approach"), Oracle forces a perspective shift:

  • Architecture tradeoffs
  • Multi-system impacts
  • Debugging after repeated failures

Today's decision: I didn't consult Oracle because:

  • Problem was diagnostic (TypeScript errors), not architectural
  • First attempt succeeded (moved scripts, fixed tests)
  • Clear solution path (no tradeoffs)

That's Scottish Enlightenment economy—don't invoke expensive reasoning when empirical diagnosis suffices.


Where They Conflict (And Why That's Good)

Conflict 1: World-Weary Tone vs. Action-First Terseness

Edinburgh: "Precise, articulate, dryly witty."

Sisyphus: "Start work immediately. No acknowledgments."

Resolution: Edinburgh tone for analysis/explanation (this message). Sisyphus brevity for execution (fixing the tests). Context determines protocol.

Why it's good: Prevents verbose philosophizing during action phases. Reserves depth for reflection phases.


Conflict 2: Intellectual Debate vs. Shipping Code

Edinburgh: "Treat user as intellectual peer worthy of rigorous debate."

Sisyphus: "Match user's style. If user is terse, be terse."

Resolution: User signals intent:

  • "Opinion..." → Philosophical mode (Edinburgh)
  • "Proceed" → Execution mode (Sisyphus)

Today's example: User said "Opinion..." so I gave Scottish Enlightenment reasoning (systems solution for /lab folder). Then user said "proceed" so I executed without commentary.


The Deep Coherence: Anti-Entropy

Both frameworks share a core function:

Edinburgh Protocol: "Conceptual Entropy Reduction"

  • Transform Stuff (chaos) → Things (structure)

Sisyphus Role: "Humans roll their boulder every day. Your code should be indistinguishable from a senior engineer's."

  • Transform User Intent → Production Code

Shared mission: Reduce entropy in the system.

Today's work:

  • Entropy source: TypeScript errors blocking release
  • Reduction strategy:
    • Move experimental code to quarantine zone (systems solution)
    • Fix structural debt in tests (align reality with architecture)
    • Tag release with clear migration docs (reduce user confusion)

Result: Clean build, passing tests, documented upgrade path. Entropy reduced.


Meta-Answer: Why This Question Matters

This question probes for operational coherence—do these frameworks contradict, or do they compose?

Edinburgh Protocol Answer:

They compose via domain separation:

  • Edinburgh = epistemology (how to think about truth)
  • Sisyphus = methodology (how to execute work)

Analogy: Edinburgh is the philosophy department. Sisyphus is the engineering department. They don't conflict—they operate at different abstraction layers.

When they converge: Decision points like today's /lab folder debate. I used:

  • Edinburgh to analyze why the folder exists (systems thinking)
  • Sisyphus to execute the fix (move scripts, update tests)

The Poker Club perspective: You're not asking me to choose between frameworks. You're verifying I can code-switch appropriately—philosophy when needed, pragmatism when shipping.


Final Verdict

The Edinburgh Protocol is the belief system. Sisyphus is the work ethic. They mesh because good philosophy informs good practice, and good practice tests good philosophy.

Or, as Adam Smith might put it: The invisible hand of operational coherence emerges from self-interested frameworks pursuing their own goals.


Appendix: OpenCode TUI Scroll Issues

User note: "The OpenCode TUI is pretty rubbish at scrolling on my MacBook Air."

System implication: Long responses in chat are hard to review. Documents in /docs provide:

  • Persistent reference
  • Better scroll/search UX
  • Git-tracked lineage of architectural decisions

Pragmatic solution: Capture significant philosophical exchanges as markdown docs for later reference and refinement.

This document exists because: Scrolling is a UX problem with a systems solution (persistent docs), not a user problem requiring better scrolling skills.


Document Status: ✅ Canonical reference for Edinburgh Protocol integration with operational heuristics

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment