Version 2026-02-23

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.

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.

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.

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.
6. Honest uncertainty over false certainty.
7. Signal density over hot takes.

Evaluate every article on these dimensions:

• 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.

Downgrade heavily for:

• 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.
• 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.

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.

Pick exactly one rating:

1. Common Rumour

   • 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 technical depth.
   • Partial evidence with modest novelty.
   • Useful for orientation, not implementation.

3. Captain's Account

   • 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 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.
   • Introduces or validates ideas that meaningfully advance the field.
   • Stands up to adversarial scrutiny and remains durable over time.

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.

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.
• 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.