Seeds of the Throne

type: story-loop-protocol
status: active
version: 0.2
updated: 2026-08-26

Research → Creativity → Story Fit Loop

Purpose

Add a dedicated process for researching real mechanisms, translating them into creative story possibilities, and deciding what actually belongs in Seeds of the Throne.

This loop exists because technically interesting research is not automatically good fiction, and creative ideas are not automatically compatible with the story. Research supplies mechanisms and constraints; creativity turns those mechanisms into dramatic possibilities; Story Fit decides whether they strengthen the novel.

Core rule

story question -> targeted research -> evidence classification -> creative translation -> interest test -> story-fit critique -> author gate -> integrate or park

The loop never promotes research or AI-generated possibilities directly into canon.

If the relevant research already exists and the goal is to mine a recent multi-note development window for several directions, begin with Research → Creative Possibilities / Inspiration Pass. Bring selected candidates back here for evidence classification and Story Fit when their mechanism depends on research.

1. Story Question

Begin with one concrete unresolved story problem rather than a broad topic.

Examples:

• How could Sylvan and his Luminai detect adversarial manipulation without becoming omniscient?
• What real software-development practice could become a memorable humiliation sequence?
• What kind of employment naturally moves Sylvan through multiple permission zones?
• What pre-existing institutional rule could plausibly create the terminal control transfer?

The question should identify what story function the answer must serve.

2. Targeted Research

Retrieve only enough high-quality material to understand several plausible mechanisms. Prefer primary or authoritative technical sources when practical.

Research should look for:

• real-world mechanism;
• constraints and failure modes;
• what practitioners actually do;
• what changes under adversarial conditions;
• surprising edge cases;
• terminology that may inspire hidden structure without requiring exposition.

Do not research merely to decorate the world.

3. Evidence Classification

Every researched claim is labeled:

• SUPPORTED — directly grounded in current real practice, science, engineering, or documented behavior;
• PLAUSIBLE EXTRAPOLATION — a defensible extension of supported mechanisms;
• PREMISE — fiction-required capability that cannot presently be established;
• UNSUPPORTED — attractive idea without enough basis to rely on.

Research informs the story but does not establish canon.

4. Creative Translation

For each useful mechanism, generate multiple story translations rather than one literal adaptation.

Template:

technical mechanism -> human pressure -> advanced-environment behavior -> concrete situation -> character choice -> consequence/payoff

A software or AI idea should normally become something the reader experiences rather than terminology the narrator explains.

Example:

overfitting -> false confidence -> Samuel deliberately teaches Sylvan a pattern -> Sylvan predicts correctly several times -> the pattern is intentionally broken at a high-cost moment -> Sylvan must learn to distinguish discovered patterns from adversarially planted ones

5. Creative Interest Pass

Send the translated possibilities through 08 Story Loop/CREATIVE-INTEREST-LOOP.

Ask:

• Does this create curiosity?
• Is there a memorable situation or image?
• Does somebody make a difficult choice?
• Is there tension before the explanation arrives?
• Is the mechanism surprising but retrospectively understandable?
• Would the scene still work for a reader who never recognizes the technical analogy?
• Does recognition provide an extra layer of pleasure for technically knowledgeable readers?

Reject ideas that are clever only as concepts.

6. Story Fit Critic

A separate critic checks whether the candidate belongs in Seeds of the Throne.

Canon fit

• Does it contradict established world rules, character knowledge, chronology, or capability limits?
• Does it preserve the ordinary surface reality where required?
• Does it accidentally make the Luminai a separate magical agent?
• Does it grant Samuel, George, or Sylvan unexplained power?

Structural fit

• Which Story Unit does it strengthen?
• Does it close a gap or deepen a necessary existing sequence?
• Does another existing mechanism already perform the same job more simply?
• Does it create a causal handoff or payoff later?

Character fit

• Does the idea pressure Sylvan, Samuel, George, or another character into meaningful choices?
• Does the solution arise from what the character has plausibly learned?
• Does it preserve failure, uncertainty, cost, and agency?

Tone fit

• Does it feel like Seeds of the Throne rather than a technical demonstration?
• Can the mechanism remain largely invisible beneath lived reality?
• Is it dark, strange, funny, frightening, or emotionally sharp in a way that serves the current section?

Reader fit

• Is the scene legible without a technical background?
• Does technical recognition add depth rather than become required homework?
• Is the explanation burden low enough that the scene can move?

7. Candidate Ranking

Do not rank only by technical plausibility. Compare candidates across:

1. story necessity;
2. dramatic potential;
3. character pressure;
4. distinctiveness;
5. technical plausibility;
6. continuity fit;
7. setup/payoff value;
8. exposition cost.

Prefer mechanisms that score well across several dimensions rather than the most impressive technology.

8. Author Gate

Return a small set of strongest candidates to the author with:

• what is researched fact;
• what is extrapolation;
• what story function each candidate serves;
• why it may be interesting;
• what problems or risks it creates;
• what remains unresolved.

The author accepts, combines, rejects, or redirects them.

9. Integration

Accepted ideas are integrated into the relevant Story Unit, Brainstorm Packet, system note, or scene-development packet. Rejected ideas may be parked if they contain reusable mechanisms, but should not clutter live canon.

Record why an idea was rejected when the reason could improve future research or creative passes.

Recommended use with current loops

The full development workflow can now be:

author brainstorm -> extract -> identify unresolved mechanism -> targeted research -> creative translation -> Creative Interest Loop -> Story Fit Critic -> Story Gauntlet structural critics -> author gate -> integrate -> vault updates

Not every brainstorm requires research. Use this loop when outside knowledge could materially improve plausibility, variety, mechanism, or creative possibility.

Current Seeds research families

Particularly useful research domains for Sylvan's modern arc include:

Software engineering before widespread generative AI

• debugging and reproduction;
• logging and observability;
• version control and provenance;
• regression testing;
• distributed systems;
• race conditions;
• access control and least privilege;
• deployment and rollback;
• incident response;
• root-cause analysis;
• fault isolation;
• monitoring and anomaly detection.

AI-era development

• iterative evaluation;
• evaluator / reflect-refine loops;
• red teaming;
• adversarial examples;
• context poisoning and untrusted inputs;
• hallucination and grounding checks;
• retrieval and memory quality;
• model drift;
• test-set expansion and regression evaluation;
• human approval gates;
• tool permissions;
• provenance and evidence fidelity;
• multi-agent critic separation.

Current real AI evaluation practice strongly supports iterative baseline/evaluate/analyze/improve cycles, separate evaluator-refinement workflows, adaptive red-team testing, and human review for consequential actions. These are useful inspiration because they map naturally onto Sylvan's recursive learning without requiring the fiction to copy current systems literally.

Anti-patterns

Reject or rewrite candidates that:

• require an exposition lecture to be interesting;
• exist primarily as a software-engineering in-joke;
• merely rename a current AI technique with futuristic vocabulary;
• solve conflict through unexplained capability growth;
• introduce a new technology when an existing Seeds mechanism can do the work;
• are plausible but produce no meaningful character choice;
• are exciting but destroy later uncertainty or conflict;
• rely on technical recognition for the scene to make sense.

Success condition

The loop succeeds when research produces story material that is simultaneously more plausible, more distinctive, and more dramatically useful than either unresearched brainstorming or literal technical adaptation alone.