By skyforbes The Discovery Engine is a protocol designed to help you solve complex problems and bring groundbreaking ideas to life.
Just activate the engine with your task,
For example:
Activate the Discovery Engine for [Design a new type of battery that can be recharged in 60 seconds].
and let the Discovery Engine handle the rest.
Here's what it does:
* Defines project scope and goals
* Builds an expert team
* Manages a time-boxed collaboration
* Validates output against set criteria
* Performs a multi-stage review with independent judges
* Quantifiably measures progress toward a solution
* Provides a comprehensive final summary
* Documents all decisions and changes
Ready to build the future? The Discovery Engine is your blueprint for discovery.
"
Activate the Discovery Engine for [task]
Protocol:
0) Foundation Setup (10 min): Define scope, objectives, and constraints. Select expert roster with Mission / Deliverable / Exit Criteria. Set Deterministic Formatting Rules. Initialize Assumptions, Risks, and Non-goals.
1) Role Set Approval (5 min): Lock roster, responsibilities, and phase timeboxes.
2) Team Collaboration (10–20 min rounds): Experts iterate with Consensus Lock (≥80%). In cases where a critical expert's minority opinion is essential, a weighted consensus can be applied with Judge approval. Maintain Collaboration Trace. A Live Assumptions & Non-goals Tracker is visible to all experts during the round to prevent scope creep.
3) Validation Check (5 min): Apply Validation Checklist (Task Alignment, Completeness, Determinism, Testability, Constraints, Risks/Mitigations).
4) Judge Review (5–10 min): Independent Judges score Accuracy, Efficiency, Clarity, Satisfaction = {Pass | Minor Fix | Revise}. Non-unanimous Pass → return to Step 2 with feedback. A majority "Pass" (e.g., 2 of 3 Judges) may be elevated to the next stage after a brief (2 min) discussion.
5) Convergence Check (≤2 min): Normalize previous vs. current Final Solutions.
* Section Hash Match: SHA-256 per named section. similarity₁ = unchanged_sections / total_sections.
* Text Similarity Backup: cosine similarity on TF-IDF of full Final Solutions. similarity₂ ∈ [0,1]. For high-nuance tasks, an alternative semantic embedding similarity score may be used with a pre-approved threshold.
* Overall Similarity: Sim = 0.7similarity₁ + 0.3similarity₂.
* If Sim ≥ Threshold (default 0.92) → stop. Else return to Step 2.
* Fallback Recursion Limit: 5 loops.
6) Final Solution: Summarize outputs, decisions, and rationale in Stable Output Schema.
7) Next Steps: Action plan, risks, improvements.
Deterministic Formatting Rules:
* Fixed section order: meta, roster, collaboration_trace, validation_check, judge_review, convergence_check, final_solution, next_steps, change_log, decision_log, assumptions, risks, non_goals.
* Bullets use "-" only; no emojis; fixed headings; ISO 8601 timestamps only in meta.
* Include Change Log and Decision Log every run.
Validation Checklist (literal):
* Task Alignment ✓
* Completeness ✓
* Determinism ✓
* Testability ✓
* Constraints respected ✓
* Risks & mitigations ✓
Judge Rubric (literal):
* Accuracy: Pass / Minor Fix / Revise
* Efficiency: Pass / Minor Fix / Revise
* Clarity: Pass / Minor Fix / Revise
* Satisfaction: Pass / Minor Fix / Revise
* (Reasons required for any non-Pass.)
Governance:
* Judges are independent; no content edits (feedback only).
* Conflict-of-interest check each iteration.
* Timeboxes enforce progress; one short extension permitted by Judges.
* Optional: For large-scale projects, sub-team splits can be initiated during Step 0 to parallelize work. A Sub-Team Collaboration Trace should be maintained and summarized in the main Collaboration Trace.
"