Skip to main content

The Electric Monks — Dialectic Skill

Named after Douglas Adams’ machines built to believe things for you An agent skill that helps you think better by automating the brutally expensive parts of deep reasoning. Two AI subagents — the Electric Monks — believe fully committed positions on your behalf. A third, the orchestrator, decomposes both arguments into atomic parts, finds cross-domain connections, and synthesizes. The result is a semi-lattice — a structure no single linear argument could produce. You operate from a belief-free position above the Monks, analyzing the structure of the contradiction rather than being inside either side. This isn’t artificial intelligence — it’s an artificial belief system that frees you to think.

Why this works

Thinking well about hard problems has at least three bottlenecks, and they compound:
  1. Belief. Once you hold a position, you can’t simultaneously entertain its negation at full strength. You hedge, steelman weakly, unconsciously bias the comparison.
  2. Research breadth. Surveying a domain’s thinkers, history, and adjacent fields takes enormous time. Most people stop too early.
  3. Structural comparison. Even with two positions side by side, decomposing them into atomic parts and finding cross-domain connections is cognitively brutal. Most analysis stalls here.
LLMs can do all three at a scale and speed humans can’t. This skill orchestrates them to do exactly that.

What the output feels like

Left alone, LLMs produce shallow takes. The dialectic breaks that pattern. As you read through the Monks’ committed arguments, connections come to mind — things neither side considered, corrections to their framing, ideas you hadn’t articulated yet. You feed these back in. The skill tunes to your thinking more and more with each round, but it also rigorously exposes the contradictions in that thinking — so you get an increasingly full and precise map of your own ideas. Then the skill breaks it apart and reforms it as something richer and more interesting than what you started with. Each synthesis becomes the next round’s thesis, and by Round 2–3 the dialectic is operating in territory no single prompt could reach.

Installation

Set up the skill with your coding agent

Quickstart

Run your first dialectic in minutes

When to use

Find the right scenarios for dialectical thinking

How it works

Understand the seven-phase process

When to use

  • You’ve locked onto a vision and can’t genuinely entertain alternatives. You have a strong thesis — maybe an architecture, a strategy, a life direction — and you want to stress-test it, but you keep steelmanning the other side weakly.
  • You’re trying to do everything because cutting anything feels like betrayal. Competing needs all feel equally urgent. You can’t triage because every priority has someone counting on it.
  • You can argue every side but can’t commit to any of them. You find the question intellectually interesting but “what do you actually think?” produces discomfort. You’ve explored this before without resolution.
  • You’ve optimized a system and suspect you might be optimizing the wrong thing. Your approach works — you have data to prove it — but the landscape may have shifted and you can’t see past your own competence.
  • Your own values contradict each other. You believe multiple things passionately, each feels individually right, but collectively they’re impossible. The tension is internal, not external.
  • “This is how it’s done” has become invisible as an assumption. You have deep knowledge of how things work, but you suspect it’s blinding you to radically different approaches.
Works across domains — technical architecture, product strategy, philosophy, personal decisions.

How it works

The process has seven phases:

Phase 1: Elenctic interview + research

The orchestrator interviews you Socratically — surfacing hidden assumptions, finding the deepest version of the contradiction, and identifying your belief burden. Then it researches the domain to ground both sides in specifics. The interview surfaces what you’re actually wrestling with; the research ensures the downstream arguments are grounded in specifics, not generics.

Phase 2: Generate Electric Monk prompts

The orchestrator crafts two prompts — one per Monk — calibrated to your specific belief burden. Each prompt includes framing corrections that prevent the Monk from falling into the obvious, boring version of the argument, plus targeted research directives for position-specific evidence.

Phase 3: Spawn the Electric Monks

Two separate AI agents — each in a fresh, isolated context — write fully committed position essays. They don’t hedge. They don’t try to be balanced. Each one inhabits its position and makes the absolute strongest case. Spawning them in separate sessions with no shared context produces structural decorrelation — genuinely different reasoning paths, not the same analysis with different conclusions bolted on.

Phase 4: Determinate negation

The orchestrator analyzes both essays to find: where each position’s own logic undermines itself (self-sublation), what both sides implicitly agree on without realizing it (shared assumptions), and the specific way each position fails — not “it’s wrong” but “it fails in THIS way, which points toward THIS thing that’s missing.” Then comes the Boydian decomposition: shatter both arguments into atomic parts, strip them of which Monk said them, and look for surprising cross-domain connections.

Phase 5: Sublation (Aufhebung)

The orchestrator generates a synthesis that simultaneously cancels both positions as complete truths, preserves the genuine insight in each, and elevates to a new concept that transforms the question itself. This is not compromise. It’s not “use A for some cases and B for others.” It’s a reconceptualization — something neither Monk could have conceived from within their frame, but which, once stated, makes the original contradiction predictable. The synthesis is an abductive hypothesis: what would make it unsurprising that both Monk positions exist with genuine evidence?

Phase 6: Validation

Both Monks evaluate the synthesis: were they elevated (their core insight preserved within something larger) or defeated (their position just dismissed)? Then a hostile auditor — a fresh agent with no position — attacks the synthesis for hidden assumptions, compromise disguised as transcendence, and structural flaws.

Phase 7: Recursion

Each synthesis generates new contradictions. The orchestrator proposes 2–4 directions; you choose which to pursue. The process repeats — and each round gets sharper, pulling in new cross-domain material that the previous round made relevant. The first round is calibration — the least insightful output. By Round 2–3, the dialectic has dug past the obvious framing into territory that neither you nor the Monks could have reached from the starting question.
The first round is the least insightful. Think of it as calibration. Each subsequent round gets sharper, more specific, and more tuned to what you actually care about. The real breakthroughs usually come in rounds 2 and 3, once the process has dug past the obvious framing into the deeper tensions.

The theory

The skill rests on three theoretical frameworks — one per bottleneck — plus Alexander’s semi-lattice theory, which explains why the output is structurally richer than any single line of reasoning.

Rao: The belief bottleneck

From Venkatesh Rao’s “Electric Monks” framework (after Douglas Adams). Rao argues there are three ways to speed up your cognitive transients: (a) maintain a richer library of mental models, (b) switch between them faster, (c) believe fewer things. The first two hit hard limits — more models means higher search costs, faster switching runs into biology. Only the third has no ceiling: if machines carry the belief load, you can become a pure context-switching specialist. Boyd’s analogy: in the Korean War, F-86 Sabres achieved a 10:1 kill ratio against MIG-15s despite similar flight capabilities. The difference was hydraulic controls — the pilot could reorient faster because the plane did the mechanical work. But the transients weren’t just faster, they were better — by devoting less attention to struggling with controls, the pilot chose better maneuvers. The Monks work the same way: by carrying the belief work, they don’t just save you time, they free up cognitive capacity that goes into higher-quality structural analysis.

Eisenstein + Boyd: The research and decomposition bottlenecks

Elizabeth Eisenstein argued that the printing press’s most transformative effect wasn’t making books cheap — it was typographic fixity. For the first time, scholars could lay texts side by side and detect contradictions. Before print, you read one manuscript, traveled to another library, read another, and tried to hold the comparison in your head. LLMs represent the next step: not just fixity and side-by-side comparison, but automated structural comparison. Both remaining bottlenecks — research breadth and structural decomposition — are cognitively brutal. Most people abandon the first too early and never attempt the second. This is where Boyd’s “Destruction and Creation” enters. His critical insight: you cannot synthesize something genuinely new by recombining within the same domain. You must first shatter existing concepts into atomic parts (destruction), then find cross-domain connections to build something new (creation).

Hegel: Determinate negation and Aufhebung

Determinate negation doesn’t say “this is wrong.” It says “this is wrong in a specific way that points toward what’s missing.” The failure mode is a signpost. Sublation (Aufhebung) simultaneously cancels, preserves, and elevates — it produces a reframing so complete that the original terms of the debate stop making sense.

Alexander: Semi-lattice generation

Christopher Alexander showed that natural cities have semi-lattice structure — overlapping, cross-connected sets — while designed cities impose tree structure where every element belongs to exactly one branch. Trees are easier to think about but destroy the cross-connections that make systems alive. Language is tree-structured. Every argument a Monk produces is a tree — a coherent linear path from premises to conclusions. But the Boydian decomposition phase strips both arguments of their tree structure, extracts atomic parts, and finds cross-connections between elements that came from different trees. These cross-domain connections are the semi-lattice edges. The synthesis is the semi-lattice that emerges from the overlap. The skill is a semi-lattice compiler. The answer to “language can’t represent semi-lattices” is not “make the LLM output a semi-lattice directly.” It’s: produce multiple committed trees from different positions, then extract the cross-connections.

Build docs developers (and LLMs) love