Quickstart
This guide walks you through running your first dialectic from start to finish.Running a dialectic
Create a subdirectory for your topic
Each dialectic lives in its own directory. Create one and navigate to it:Choose a descriptive name for your topic. Examples:
tanstack-vs-nextjsagent-governancecareer-decisionarchitecture-monolith-vs-microservices
Start your coding agent
Launch your coding agent in that directory. For Claude Code:For other agents, use your agent’s launch command.
Load the skill and provide your topic
Install the skill and describe what you want to explore:The skill will walk you through the elenctic interview, spawn the Monks, and produce the full dialectical trace — all saved as files in the current directory.
Participate in the elenctic interview
The orchestrator will ask you questions to surface:
- Hidden assumptions you haven’t articulated
- The deepest version of the contradiction
- What you’re actually wrestling with underneath the surface framing
Be specific. The better the orchestrator understands your situation, the better everything downstream works. Don’t hold back — share context, constraints, and what’s really at stake.
Review the Electric Monk essays
The orchestrator will spawn two Monks, each believing one side of the tension at full conviction. You’ll receive two position essays.As you read them:
- Correct wrong assumptions — “That’s not how it works” or “They’re missing that…”
- Add new ideas — Things neither monk mentioned, gut feelings you can’t fully articulate
- Note what feels off — If something doesn’t ring true, say so
Receive the synthesis
The orchestrator performs structural analysis (determinate negation) and generates a synthesis that:
- Cancels both positions as complete truths
- Preserves the genuine insight in each
- Elevates to a new concept that transforms the question itself
Choose a recursive direction
Each synthesis generates new contradictions. The orchestrator will propose 2–4 directions for the next round.Pick one that feels most valuable. Each round ratchets up the quality.
The
dialectic_queue.md file tracks explored and unexplored contradictions. You can come back to it in a future session and pick up where you left off.Example topics to explore
Here are real topics that work well with the dialectic skill:Technical architecture
- “Should we use TanStack Router or Next.js App Router for our new application?”
- “Monolith vs. microservices for our scaling needs”
- “React vs. Vue for our team’s frontend architecture”
- “GraphQL vs. REST for our API strategy”
Product strategy
- “Should we focus on enterprise or developer-first go-to-market?”
- “Build an open-source community or focus on commercial features?”
- “Should we pursue funding or stay bootstrapped?”
Personal decisions
- “Should I take the senior IC track or move into management?”
- “Stay at my current company or join the startup?”
- “Deep specialization vs. broad generalist career path”
Philosophy and values
- “How do I balance shipping fast with building the right thing?”
- “Individual autonomy vs. team coordination in our org structure”
- “Should we prioritize innovation or stability in this phase?”
Tips for effective dialectics
You are the co-pilot
You are the co-pilot
Interrupt, correct, redirect at any point. The Monks will get things wrong. Your corrections are the highest-leverage input in the entire process.
The first round is calibration
The first round is calibration
Don’t judge the skill by Round 1. The real insights come in Rounds 2–3, once the process has dug past the obvious framing.
Say yes to recursion
Say yes to recursion
When the skill proposes recursive directions after a synthesis, pick one. Each round ratchets up the quality. The dialectic operates in territory no single prompt could reach by Round 2–3.
The dialectic queue persists
The dialectic queue persists
The
dialectic_queue.md file tracks explored and unexplored contradictions. You can come back to it in a future session and pick up where you left off.Expect 10-15 minutes per round
Expect 10-15 minutes per round
This is a heavy process by design. Plan for at least 3 rounds. Block off 45-60 minutes for a meaningful session.
What to expect
Round 1: Calibration
- The orchestrator learns your belief burden
- Monks make their first committed arguments
- Synthesis identifies the deeper question
- Least insightful output — this is setup
Round 2: Depth
- Process has dug past obvious framing
- Monks pull in cross-domain material
- Synthesis starts revealing structural insights
- Real breakthroughs begin here
Round 3+: Territory
- Operating in conceptual space no single prompt could reach
- Cross-domain connections from unexpected fields
- Synthesis produces genuine reconceptualization
- Highest-value output
In test runs, a React/Vue dialectic evolved from “corporate lab vs. auteur” into a “co-evolutionary arms race” framework. An institutional identity dialectic went through seven cycles, pulling in Gödel’s incompleteness theorem, Coasean transaction costs, and jurisprudential concepts that had nothing to do with the original question — but were essential by the time the dialectic reached them.
Next steps
Introduction
Understand the theory behind the skill
Installation
Review setup and requirements