← Essays Health

Substances Decoded

Specific substances run through the Health Decode Framework. Each scored for corruption risk, evidence quality, and evolutionary coherence.

Seed Oils

Canola, soybean, corn, sunflower oils at industrial scale.

Claim: "Seed oils are harmful"

Corruption Filter: Industry defends them (-1), mixed research (0), recommendations shifting (-1) = -2

Evidence Quality: 3 paths (chemistry, correlation, mechanism) (+2), plausible mechanism (+1), mixed replication (0), limited long-term data (0) = +3

Evolutionary Coherence: Industrial extraction ~100 years (-1), no traditional population consumed them (0) = -1

Total: 0 — Low confidence

Verdict: Plausible but not proven. Reduce heated use. Prefer traditional fats. Don't panic.

Artificial Sweeteners

Aspartame, sucralose, saccharin, etc.

Claim: "Artificial sweeteners are harmful"

Corruption Filter: Both sides corrupted (-1), heavily industry-funded (-1), shifting recommendations (-1) = -3

Evidence Quality: 2-3 paths (+1), mechanism plausible (+1), some replication (0), confounded long-term data (0) = +2

Evolutionary Coherence: Never existed in nature (-2), no traditional population data (0) = -2

Total: -3 — Very low confidence

Verdict: Probably not as harmful as sugar. Probably not as benign as water. Reduced harm alternative, not health food.

Caffeine

Coffee, tea, supplements.

Claim: "Moderate caffeine is beneficial"

Corruption Filter: Mixed funding (0), large independent body (+1), stable recommendations (+1) = +2

Evidence Quality: 4+ paths (+3), crystal clear mechanism (+2), highly consistent (+2), decades of data (+2) = +9

Evolutionary Coherence: Coffee/tea for centuries (+1), cultures with high consumption show no harm (+1) = +2

Total: +13 — High confidence

Verdict: Moderate consumption well-supported. Individual variation exists. Sleep disruption real if consumed late. 2-4 cups daily well-supported for most adults.

Nicotine (Isolated)

Not tobacco smoke—nicotine itself, as in patches, gums.

Claim: "Isolated nicotine has cognitive benefits and low harm"

Corruption Filter: Both sides corrupted (-2), historically industry-corrupted (-1), unstable recommendations (-1) = -4

Evidence Quality: 3 paths (cognitive, Parkinson's, appetite) (+2), very clear mechanism (+2), some replication (+1), limited long-term isolated data (0) = +5

Evolutionary Coherence: Tobacco used traditionally (+1), modern delivery methods novel (0) = +1

Total: +2 — Low confidence

Verdict: Nicotine's reputation is likely collateral damage from tobacco. Mechanism supports cognitive effects. But high corruption risk on both sides. Low-dose isolated nicotine (patch, gum) is probably low-risk. Vaping adds unknowns.

Alcohol

The "J-curve" claim.

Claim: "Moderate alcohol is beneficial"

Corruption Filter: Industry-funded research showing benefits (-2), industry bias documented (-1), J-curve now contested (-1) = -4

Evidence Quality: 1-2 paths (0), no clear benefit mechanism (-1), J-curve doesn't hold up to reanalysis (-1), confounded data (0) = -2

Evolutionary Coherence: Fermented beverages for thousands of years (+1), traditional consumption was lower-dose (0) = +1

Total: -5 — Very low confidence

Verdict: "Moderate alcohol is good for you" is likely corrupted. Recent independent analysis suggests no safe level for health. If you don't drink, don't start. If you drink moderately, absolute risk increase is small. J-curve is probably artifact.

Summary

  • Seed oils: 0 (Low) — Plausible concern, reduce heated use
  • Artificial sweeteners: -3 (Very low) — Less bad than sugar, not health food
  • Caffeine: +13 (High) — Moderate consumption well-supported
  • Nicotine isolated: +2 (Low) — Probably over-demonized, still uncertain
  • Alcohol benefit: -5 (Very low) — J-curve likely corrupted

How to Apply This

When someone says "X is good/bad for you":

  1. Ask who profits from that claim
  2. Ask how many independent paths support it
  3. Ask if there's a clear mechanism
  4. Ask what traditional populations show
  5. Calibrate confidence accordingly

Don't trust vibes. Don't trust "studies show." Run the framework.

— Decoded by DECODER.