Longevitea: building an open dataset for what survives the cup

Longevity claims are easy to make and hard to test. A cup of tea is a better research object: it is accessible, repeatable, chemically measurable, and small enough to study without pretending that a cell assay is a lifespan trial.

We started Longevitea as an open-source dataset and research pipeline for that question: what compounds are actually delivered by a hot-water infusion, how do they change with temperature and time, and which mixtures are worth measuring next?

The first candidate set

The registry now includes green, white, and black tea; jiaogulan; olive leaf; hibiscus; moringa; goji; lemon balm; rosemary; mint; yerba mate; tulsi; and ginger. They are not treated as equivalent. Green tea has the deepest human consumption literature. Hibiscus has direct tea-level blood-pressure trials. Olive leaf, jiaogulan, tulsi, moringa, and ginger have more extract or powder evidence than cup-level evidence. Goji is often better treated as a food because eating the rehydrated berry delivers a different matrix from drinking its water.

Large cohort studies are useful context, not proof of causality. A recent tea meta-analysis pooled 38 prospective cohort datasets and nearly two million participants, while a flavonoid-diversity cohort linked wider dietary variety with lower disease and mortality risk. Those findings justify testing diversity; they do not establish a magic dose or prove that a particular blend extends life.

Temperature is part of the ingredient

“Boil the water” is not a complete protocol. In one green-tea study, 85 °C for three minutes gave a strong extraction/sensory result; another found catechin levels and composition changed across 60–95 °C and long steep times. White tea also showed temperature- and time-dependent catechin, caffeine, theanine, and amino-acid delivery. Hibiscus has direct hot-versus-cold extraction data, lemon balm has hot-versus-room-temperature comparisons, and yerba mate shows a polyphenol/tannin trade-off between 70 and 100 °C.

Our first measurement matrix is therefore:

• green, white, and black tea: 75/85/95 °C × 3/5/10 minutes;
• hibiscus and lemon balm: 90/100 °C × 5/10 minutes;
• yerba mate: 70/85/100 °C × 5/10 minutes;
• ginger: 80/95 °C × 5/15 minutes, fresh and dried separately.

The project records actual liquid temperature, water volume, pH, hardness, particle size, cover, filtration, lot, and analytical method. A kettle setting alone is not reproducible chemistry.

How the ranking works

The current simulator is deliberately modest. It samples one-to-three ingredient compositions, rewards provisional coverage, chemical diversity, accessibility, and penalizes uncertainty, caffeine, iron-interaction risk, and safety flags. Green, white, and black tea share one camellia diversity group, so combining them no longer receives a false diversity bonus. Ginkgo remains excluded by a hard safety gate.

The mixed-group ranking currently points to green tea paired with mint, hibiscus, moringa, rosemary, lemon balm, or jiaogulan as measurement priorities. That is an experiment queue, not a consumer prescription. The priors will be replaced with measured cup concentrations before any serious optimization.

The pipeline

Longevitea is moving through discovery, identity and safety review, structured literature extraction, standardized preparation, LC-MS/HPLC and mineral analysis, compound-stability modeling, interaction graphs, Pareto-frontier optimization, and reproducible Plotly publication. Every claim carries its evidence tier: measured cup chemistry, human intervention, observational cohort, animal/in-vitro mechanism, or tradition/marketing.

The repository is open source, with the starter data, protocol, source citations, temperature matrix, tests, and simulation code available now. The next meaningful result will not be a “perfect tea.” It will be a transparent table showing what survived the cup, how uncertain that measurement is, and which combination deserves the next experiment.

This is research and decision-support tooling, not medical advice or a treatment recommendation. Human intervention evidence, medication interactions, pregnancy, contamination, and species identity all matter more than a compelling pathway diagram.