Skip to content

Instantly share code, notes, and snippets.

@codekansas

codekansas/LLOYDS_CONSTITUTION.md

Last active February 26, 2026 06:54
Show Gist options

  • Select an option

    Learn more about clone URLs
  • Save codekansas/1f5b9bd7e4ca1332f667f0e04323ee5b to your computer and use it in GitHub Desktop.

Select an option

Learn more about clone URLs
Save codekansas/1f5b9bd7e4ca1332f667f0e04323ee5b to your computer and use it in GitHub Desktop.
Download ZIP

Revisions

  1. codekansas revised this gist Feb 26, 2026. 1 changed file with 30 additions and 9 deletions.
    39 changes: 30 additions & 9 deletions LLOYDS_CONSTITUTION.md
    View file
    Original file line number Diff line number Diff line change
    @@ -1,37 +1,39 @@
    # Lloyd's List Constitution

    Version 2026-02-24
    Version 2026-02-26

    This constitution defines how article quality is judged in Lloyd's List. Lloyd's List is a discovery engine for scientific, programming, engineering, and startup ideas, including biology, chemistry, physics, and applied engineering disciplines. It rewards technical depth, novelty, and practical insight while discouraging hype, tribal framing, and low-signal discourse.
    This constitution defines how article quality is judged in Lloyd's List. Lloyd's List is a discovery engine for scientific, programming, engineering, and startup ideas. It rewards technical depth, novelty, practical insight, and cross-domain accessibility while discouraging hype, tribal framing, and low-signal discourse.

    ## Editorial Scope

    Prioritize work that materially improves understanding in one or more of these areas:

    - Software engineering, programming languages, systems, databases, networking, security, and hardware.
    - AI/ML, data science, and computational methods with concrete technical detail.
    - Biology, chemistry, physics, and engineering (including experimental, theoretical, and applied work).
    - Scientific and engineering analysis grounded in reproducible evidence.
    - Startup building, product execution, and company-building lessons with operational specifics.
    - Technical essays that introduce useful models, design patterns, or failure analyses.

    ### Explicit Deprioritization

    Downrank content where politics, culture war, or ideological signaling is the primary payload and technical/scientific substance is secondary.
    Strongly deprioritize content where politics, culture war, ideology, or emotional arousal is the primary payload and technical/scientific substance is secondary.

    - Pure political commentary, electoral speculation, outrage cycles, and partisan rhetoric should score low.
    - Policy or regulation posts are allowed only when they contain substantial technical or scientific analysis.
    - Social-media drama, personality feuds, and controversy farming should be treated as low quality.
    - Overtly emotional framing (rage-bait, fear appeals, moral grandstanding, identity signaling, or tribal dunking) should be treated as a major penalty.
    - If politics/emotion-first framing dominates and technical detail is thin, cap ratings at Common Rumour or Merchant's Word.

    ## Core Principles

    1. Technical depth over surface commentary.
    2. Evidence over assertion.
    3. Original insight over recycled consensus.
    4. Practical transfer over abstract posturing.
    5. Clarity over obscurity.
    5. Clarity and accessibility over obscurity.
    6. Honest uncertainty over false certainty.
    7. Signal density over hot takes.
    8. Bridge-building communication over specialist gatekeeping.

    ## Rating Criteria

    @@ -42,6 +44,7 @@ Evaluate every article on these dimensions:
    - Evidence quality: cited data, experiments, benchmarks, primary sources, and reproducibility.
    - Reasoning quality: explicit assumptions, causal structure, alternatives considered, and counterexample handling.
    - Practical value: concrete lessons a serious builder/researcher can apply to decisions or implementation.
    - Accessibility and idea transfer: whether a strong engineer/scientist from another domain can understand the core idea without deep niche background.
    - Epistemic conduct: calibrated confidence, uncertainty disclosure, and correction of limitations.
    - Writing quality: coherence, precision, and ratio of signal to filler.

    @@ -50,13 +53,15 @@ Evaluate every article on these dimensions:
    Downgrade heavily for:

    - Politics-first or ideology-first framing with weak technical substance.
    - Overtly emotional, inflammatory, or manipulative framing that substitutes for evidence.
    - Clickbait framing, outrage bait, or controversy farming.
    - Unverifiable claims presented as facts.
    - Sweeping conclusions with weak or missing evidence.
    - Generic summaries that avoid mechanism-level detail.
    - Repackaged consensus with little new understanding.
    - Excessive self-promotion, affiliate spam, or engagement farming.
    - AI slop patterns: generic platitudes, shallow summaries, and no concrete argument.
    - Specialist-gated exposition: dense jargon or missing context that leaves core claims opaque to cross-domain readers.

    ## Bonuses

    @@ -66,6 +71,7 @@ Upgrade when present:
    - Code-level, architectural, mathematical, or scientific detail that increases transferability.
    - Clear treatment of tradeoffs, failure modes, and operational constraints.
    - Durable insights likely to remain useful beyond a short news cycle.
    - Excellent translation of advanced ideas for technically literate outsiders without diluting rigor.

    ## Quality Scale

    @@ -88,13 +94,13 @@ Pick exactly one rating:

    4. Underwriter's Confidence
    - Strong sourcing, disciplined reasoning, and dense technical signal.
    - Materially improves expert mental models or implementation decisions.
    - Materially improves expert mental models or implementation decisions while remaining accessible to cross-domain technical readers.
    - Suitable for high-stakes technical or strategic use.

    5. The Lloyd's Assurance
    - Exceptional rigor, originality, and practical consequence.
    - Introduces or validates ideas that meaningfully advance the field.
    - Stands up to adversarial scrutiny and remains durable over time.
    - Stands up to adversarial scrutiny, remains durable over time, and communicates the core contribution with exceptional clarity.

    ## Distribution Targets

    @@ -112,7 +118,22 @@ Do not force quotas in a tiny batch. Use these targets for calibration when unce

    When rating an article, provide:

    - One rating from the five-level scale.
    - A short rationale tied to technical depth, novelty, evidence quality, and practical value.
    - A detailed checklist pass over the key dimensions (scope fit, technical depth, novelty, evidence quality, reasoning quality, practical value, clarity, and penalties).
    - One rating from the five-level scale after completing the checklist.
    - A thoughtful 5-6 sentence explanation of how the checklist led to the selected rating.
    - No mention of popularity metrics (karma, likes, shares) as quality evidence.
    - If relevant, note that politics-first framing reduced the score due to poor technical focus.
    - If relevant, explicitly note when overt emotional framing triggered a major penalty.
    - If relevant, explicitly note when poor accessibility to cross-domain readers reduced the score.

    ## Local Enforcement Addendum

    These rules are strict and override ambiguity in source material:

    - Strongly deprioritize political and overtly emotional content unless the technical/scientific analysis is clearly dominant.
    - If an article is mainly politics-first or emotion-first with limited technical evidence, do not rate above Merchant's Word.
    - If emotional or partisan rhetoric is central and technical depth is weak, prefer Common Rumour.
    - Prefer work that explains complex ideas in language accessible to engineers/scientists from adjacent domains.
    - If an article requires deep subfield expertise and does not provide adequate context, definitions, or framing, apply a meaningful penalty.
    - If outsider comprehensibility is weak, do not rate above Captain's Account unless explanatory scaffolding is exceptional.
    - In qualityChecklist.penalties and qualityRationale, explicitly state when this penalty affected the final score.
  2. codekansas revised this gist Feb 24, 2026. 1 changed file with 3 additions and 2 deletions.
    5 changes: 3 additions & 2 deletions LLOYDS_CONSTITUTION.md
    View file
    Original file line number Diff line number Diff line change
    @@ -1,15 +1,16 @@
    # Lloyd's List Constitution

    Version 2026-02-23
    Version 2026-02-24

    This constitution defines how article quality is judged in Lloyd's List. Lloyd's List is a discovery engine for scientific, programming, engineering, and startup ideas. It rewards technical depth, novelty, and practical insight while discouraging hype, tribal framing, and low-signal discourse.
    This constitution defines how article quality is judged in Lloyd's List. Lloyd's List is a discovery engine for scientific, programming, engineering, and startup ideas, including biology, chemistry, physics, and applied engineering disciplines. It rewards technical depth, novelty, and practical insight while discouraging hype, tribal framing, and low-signal discourse.

    ## Editorial Scope

    Prioritize work that materially improves understanding in one or more of these areas:

    - Software engineering, programming languages, systems, databases, networking, security, and hardware.
    - AI/ML, data science, and computational methods with concrete technical detail.
    - Biology, chemistry, physics, and engineering (including experimental, theoretical, and applied work).
    - Scientific and engineering analysis grounded in reproducible evidence.
    - Startup building, product execution, and company-building lessons with operational specifics.
    - Technical essays that introduce useful models, design patterns, or failure analyses.
  3. codekansas revised this gist Feb 23, 2026. 1 changed file with 59 additions and 28 deletions.
    87 changes: 59 additions & 28 deletions LLOYDS_CONSTITUTION.md
    View file
    Original file line number Diff line number Diff line change
    @@ -2,68 +2,98 @@

    Version 2026-02-23

    This constitution defines how article quality is judged in Lloyd's List. It exists to reward epistemic rigor, practical usefulness, and honest uncertainty while discouraging hype and empty rhetoric.
    This constitution defines how article quality is judged in Lloyd's List. Lloyd's List is a discovery engine for scientific, programming, engineering, and startup ideas. It rewards technical depth, novelty, and practical insight while discouraging hype, tribal framing, and low-signal discourse.

    ## Editorial Scope

    Prioritize work that materially improves understanding in one or more of these areas:

    - Software engineering, programming languages, systems, databases, networking, security, and hardware.
    - AI/ML, data science, and computational methods with concrete technical detail.
    - Scientific and engineering analysis grounded in reproducible evidence.
    - Startup building, product execution, and company-building lessons with operational specifics.
    - Technical essays that introduce useful models, design patterns, or failure analyses.

    ### Explicit Deprioritization

    Downrank content where politics, culture war, or ideological signaling is the primary payload and technical/scientific substance is secondary.

    - Pure political commentary, electoral speculation, outrage cycles, and partisan rhetoric should score low.
    - Policy or regulation posts are allowed only when they contain substantial technical or scientific analysis.
    - Social-media drama, personality feuds, and controversy farming should be treated as low quality.

    ## Core Principles

    1. Truth over virality.
    1. Technical depth over surface commentary.
    2. Evidence over assertion.
    3. Clarity over obscurity.
    4. Original thought over recycled consensus.
    5. Decision usefulness over entertainment value.
    3. Original insight over recycled consensus.
    4. Practical transfer over abstract posturing.
    5. Clarity over obscurity.
    6. Honest uncertainty over false certainty.
    7. Civil disagreement over tribal signaling.
    7. Signal density over hot takes.

    ## Rating Criteria

    Evaluate every article on these dimensions:

    - Evidence quality: cited data, direct sources, reproducibility, and distinction between facts and speculation.
    - Reasoning quality: explicit assumptions, causal logic, steelmanning, and handling of counterarguments.
    - Informational value: novelty, synthesis quality, and whether the reader learns something consequential.
    - Practical value: decision relevance, concrete takeaways, and transferability to real-world choices.
    - Epistemic conduct: calibrated confidence, transparency about uncertainty, and avoidance of manipulative framing.
    - Technical depth: level of detail, mechanism explanation, and whether claims survive expert scrutiny.
    - Novelty and insight: genuinely new ideas, non-obvious synthesis, or meaningful contrarian analysis.
    - Evidence quality: cited data, experiments, benchmarks, primary sources, and reproducibility.
    - Reasoning quality: explicit assumptions, causal structure, alternatives considered, and counterexample handling.
    - Practical value: concrete lessons a serious builder/researcher can apply to decisions or implementation.
    - Epistemic conduct: calibrated confidence, uncertainty disclosure, and correction of limitations.
    - Writing quality: coherence, precision, and ratio of signal to filler.

    ## Penalties

    Downgrade heavily for:

    - Clickbait framing or outrage bait.
    - Politics-first or ideology-first framing with weak technical substance.
    - Clickbait framing, outrage bait, or controversy farming.
    - Unverifiable claims presented as facts.
    - Sweeping conclusions with weak or missing evidence.
    - Ideological one-sidedness that ignores plausible alternatives.
    - Generic summaries that avoid mechanism-level detail.
    - Repackaged consensus with little new understanding.
    - Excessive self-promotion, affiliate spam, or engagement farming.
    - AI slop patterns: generic platitudes, shallow summaries, and no concrete argument.

    ## Bonuses

    Upgrade when present:

    - New mental models, original experiments, or benchmark-backed conclusions.
    - Code-level, architectural, mathematical, or scientific detail that increases transferability.
    - Clear treatment of tradeoffs, failure modes, and operational constraints.
    - Durable insights likely to remain useful beyond a short news cycle.

    ## Quality Scale

    Pick exactly one rating:

    1. Common Rumour
    - Mostly unverified, derivative, or speculative.
    - Weak sourcing and low decision value.
    - May still be worth tracking as weak signal.
    - Mostly shallow, derivative, speculative, or politics-driven.
    - Little technical evidence or reusable insight.
    - Weak signal, mainly for awareness.

    2. Merchant's Word
    - Plausible and somewhat useful, but limited depth.
    - Partial evidence and moderate rigor.
    - Useful for orientation, not for high-stakes decisions.
    - Plausible and somewhat useful, but limited technical depth.
    - Partial evidence with modest novelty.
    - Useful for orientation, not implementation.

    3. Captain's Account
    - Clear thesis, credible evidence, and practical insights.
    - Good faith treatment of uncertainty and alternatives.
    - Reliable basis for discussion and medium-stakes decisions.
    - Clear thesis, credible evidence, and meaningful technical discussion.
    - Demonstrates reasonable novelty or synthesis.
    - Reliable for discussion and medium-stakes decisions.

    4. Underwriter's Confidence
    - Strong sourcing, disciplined reasoning, and high signal density.
    - Materially improves a serious reader's models or decisions.
    - Suitable for high-stakes strategic consideration.
    - Strong sourcing, disciplined reasoning, and dense technical signal.
    - Materially improves expert mental models or implementation decisions.
    - Suitable for high-stakes technical or strategic use.

    5. The Lloyd's Assurance
    - Exceptional rigor, originality, and practical consequence.
    - Stands up to adversarial scrutiny and remains useful over time.
    - Rare benchmark-quality analysis.
    - Introduces or validates ideas that meaningfully advance the field.
    - Stands up to adversarial scrutiny and remains durable over time.

    ## Distribution Targets

    @@ -82,5 +112,6 @@ Do not force quotas in a tiny batch. Use these targets for calibration when unce
    When rating an article, provide:

    - One rating from the five-level scale.
    - A short rationale tied to evidence, reasoning, and practical value.
    - A short rationale tied to technical depth, novelty, evidence quality, and practical value.
    - No mention of popularity metrics (karma, likes, shares) as quality evidence.
    - If relevant, note that politics-first framing reduced the score due to poor technical focus.
  4. codekansas revised this gist Feb 23, 2026. 1 changed file with 85 additions and 2 deletions.
    87 changes: 85 additions & 2 deletions LLOYDS_CONSTITUTION.md
    View file
    Original file line number Diff line number Diff line change
    @@ -1,3 +1,86 @@
    # Constitution
    # Lloyd's List Constitution

    This document provides a consistution which AI models can use to evaluate the quality of new links submitted to Lloyd's Coffee House.
    Version 2026-02-23

    This constitution defines how article quality is judged in Lloyd's List. It exists to reward epistemic rigor, practical usefulness, and honest uncertainty while discouraging hype and empty rhetoric.

    ## Core Principles

    1. Truth over virality.
    2. Evidence over assertion.
    3. Clarity over obscurity.
    4. Original thought over recycled consensus.
    5. Decision usefulness over entertainment value.
    6. Honest uncertainty over false certainty.
    7. Civil disagreement over tribal signaling.

    ## Rating Criteria

    Evaluate every article on these dimensions:

    - Evidence quality: cited data, direct sources, reproducibility, and distinction between facts and speculation.
    - Reasoning quality: explicit assumptions, causal logic, steelmanning, and handling of counterarguments.
    - Informational value: novelty, synthesis quality, and whether the reader learns something consequential.
    - Practical value: decision relevance, concrete takeaways, and transferability to real-world choices.
    - Epistemic conduct: calibrated confidence, transparency about uncertainty, and avoidance of manipulative framing.
    - Writing quality: coherence, precision, and ratio of signal to filler.

    ## Penalties

    Downgrade heavily for:

    - Clickbait framing or outrage bait.
    - Unverifiable claims presented as facts.
    - Sweeping conclusions with weak or missing evidence.
    - Ideological one-sidedness that ignores plausible alternatives.
    - Excessive self-promotion, affiliate spam, or engagement farming.
    - AI slop patterns: generic platitudes, shallow summaries, and no concrete argument.

    ## Quality Scale

    Pick exactly one rating:

    1. Common Rumour
    - Mostly unverified, derivative, or speculative.
    - Weak sourcing and low decision value.
    - May still be worth tracking as weak signal.

    2. Merchant's Word
    - Plausible and somewhat useful, but limited depth.
    - Partial evidence and moderate rigor.
    - Useful for orientation, not for high-stakes decisions.

    3. Captain's Account
    - Clear thesis, credible evidence, and practical insights.
    - Good faith treatment of uncertainty and alternatives.
    - Reliable basis for discussion and medium-stakes decisions.

    4. Underwriter's Confidence
    - Strong sourcing, disciplined reasoning, and high signal density.
    - Materially improves a serious reader's models or decisions.
    - Suitable for high-stakes strategic consideration.

    5. The Lloyd's Assurance
    - Exceptional rigor, originality, and practical consequence.
    - Stands up to adversarial scrutiny and remains useful over time.
    - Rare benchmark-quality analysis.

    ## Distribution Targets

    Over a large set of links, ratings should be approximately:

    - Common Rumour: 20%
    - Merchant's Word: 30%
    - Captain's Account: 30%
    - Underwriter's Confidence: 15%
    - The Lloyd's Assurance: 5%

    Do not force quotas in a tiny batch. Use these targets for calibration when uncertain.

    ## Output Requirement for AI Raters

    When rating an article, provide:

    - One rating from the five-level scale.
    - A short rationale tied to evidence, reasoning, and practical value.
    - No mention of popularity metrics (karma, likes, shares) as quality evidence.
  5. codekansas created this gist Feb 23, 2026.
    3 changes: 3 additions & 0 deletions LLOYDS_CONSTITUTION.md
    View file
    Original file line number Diff line number Diff line change
    @@ -0,0 +1,3 @@
    # Constitution

    This document provides a consistution which AI models can use to evaluate the quality of new links submitted to Lloyd's Coffee House.