Skip to main content
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.

​
Why recursion is where the value lives

A genuine sublation is fertile, not final. It doesn’t close the question — it transforms it into new questions at a deeper level.
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.

​
Identify new contradictions

Based on the validated synthesis, identify 2-4 new contradictions that emerged:
1

What the synthesis assumes but didn't justify

What unstated assumptions does the synthesis rest on?
2

What it optimizes that might be the wrong target

The synthesis optimizes for X. Is X actually the right thing to optimize?
3

Internal tensions within the synthesis

Are there parts of the synthesis that pull in different directions?
4

What it makes possible that creates new problems

The synthesis enables Y. What new problems does Y create?

​
Present directions to the user

Present 2-4 directions as options, not a mandate. The user chooses based on:
  • What they’re most curious about
  • Where they feel the most uncertainty
  • What would have the highest practical impact on their decision
Direction 1: The coordination mechanism assumptionThe synthesis assumes that [MECHANISM] can coordinate without central authority. But what if that assumption breaks at scale? This would pit [NEW TENSION A] against [NEW TENSION B].Direction 2: The optimization targetWe’ve been optimizing for [X], but the synthesis reveals that [X] might be a proxy for the real goal [Y]. Should we optimize for [Y] directly? This creates a dialectic between [NEW TENSION C] and [NEW TENSION D].Direction 3: The temporal frameThe synthesis works in [TIMEFRAME], but over [LONGER TIMEFRAME], [NEW FACTOR] enters. This creates tension between [NEW TENSION E] and [NEW TENSION F].Direction 4: The adjacent domainThe synthesis is about [DOMAIN A], but it has implications for [DOMAIN B]. Exploring that boundary creates a dialectic between [NEW TENSION G] and [NEW TENSION H].

​
The dialectic queue

Create or update dialectic_queue.md — a persistent file that tracks:
  • Explored contradictions: What you’ve already run through the dialectic
  • Unexplored contradictions: Directions you haven’t pursued yet
  • Synthesis history: The chain of syntheses from Round 1 to current
The dialectic queue is your orientation library. You can come back to it in a future session and pick up where you left off. It’s a map of the conceptual territory you’ve explored and the frontiers you haven’t.

​
When to do new research for recursive rounds

Recursive rounds may or may not need new research depending on whether the new contradiction opens new conceptual domains.
No new research neededThe synthesis stays within the same conceptual domain but goes deeper. The context briefing from Round 1 still applies.Example: React/Vue Round 1 → “corporate lab vs auteur.” Round 2 → “How does legacy burden causally shape innovation character?” Same domain (OSS frameworks), deeper question.

​
Repeat from Phase 2 (or Phase 1 if new research needed)

If no new research needed

Go directly to Phase 2 — generate new monk prompts based on the chosen contradiction. The monks read the updated dialectic queue and the previous synthesis.

If new research needed

Go back to Phase 1d — do targeted research on the new domain, update the context briefing, then proceed to Phase 2.

​
How many rounds?

Default: at least 3 rounds.
  • Round 1: Calibration — surfaces the obvious framing
  • Round 2: Refinement — digs past the obvious into deeper structure
  • Round 3+: Insight — operates in territory no single round could reach
In test runs, the highest-value insights consistently appeared in Rounds 2-4. Round 1 is setting the stage. Don’t judge the process by Round 1.

​
When to stop

Stop when:
  • The user feels they have the understanding they came for
  • The contradictions are getting more abstract without adding practical value
  • You’ve reached a stable attractor — the synthesis doesn’t generate fertile new contradictions
  • Time/token budget is exhausted
Don’t stop just because Round 1 produced something that sounds reasonable. The process is designed for depth — give it at least 2-3 rounds to dig past the surface.

​
What you’ve built

At the end of a multi-round dialectic, you have:
  • A dialectic queue — a map of explored and unexplored conceptual territory
  • A synthesis chain — each round’s synthesis, showing the evolution of thinking
  • A belief-free analysis — structural understanding that no single committed position could produce
  • An orientation library — a set of contradictions you can return to when new information arrives
The dialectic queue persists. You can come back months later, see what you explored, and pick up a thread you didn’t pursue the first time.

​
Example: 7-cycle institutional identity dialectic

From test runs:
  • Round 1: “Tradition vs innovation” (obvious framing)
  • Round 2: “Continuity of identity vs adaptation to environment”
  • Round 3: Pulls in Gödel’s incompleteness theorem — institutions can’t be both complete and consistent
  • Round 4: Pulls in Coasean transaction costs — why institutions exist at all
  • Round 5: Connects to jurisprudential concepts of precedent vs principle
  • Round 6: Synthesis: “Institutions are Bayesian updaters with identity-preserving priors”
  • Round 7: New contradiction: “What happens when the environment shifts faster than Bayesian updating can track?”
None of this was accessible from the Round 1 framing. The recursion built conceptual scaffolding that made each subsequent round possible.

​
Next steps

Advanced: Belief burden patterns

Learn to calibrate monks for different cognitive styles

Advanced: Domain adaptation

Adapt the process for technical, personal, or mixed domains

Theory: Why this works

Understand the theoretical foundations

Tips and best practices

Operational guidance for running better dialectics

Build docs developers (and LLMs) love

Get started for freeTalk to us