The Model · v0.2.0
Why 66 years?
By Taketoki Fujita · Last reviewed
Because that is what the parameters say — not because anyone picked it. Every input lives in a versioned file, the simulation runs from a fixed seed, and anyone who clones the repository reproduces this number exactly.
0.22
healthy years gained per calendar year, today
2061–2172
80% interval for reaching escape velocity
94%
of simulations reach it at all
What has to happen
Escape velocity is not a cure. It is an arithmetic condition: remaining healthy life expectancy has to start growing by at least one year per calendar year. Below that you lose ground as you age. Above it, your expected remaining healthy life grows faster than you spend it.
Today that figure is roughly 0.22, driven almost entirely by ordinary medicine rather than by anything from the longevity field. The chart below is the model's estimate of how that number moves, and how uncertain it is.
Probability, not a date
A single date implies a precision this evidence cannot support. The honest output is a distribution.
- by 2050
- 2%
- by 2075
- 28%
- by 2100
- 58%
The regulatory gate
Most forecasts in this space model the science and stop. This one multiplies everything by a regulatory and deployment readiness term, currently scored 15/100. Aging is not an approvable indication, no aging biomarker is an accepted surrogate endpoint, and no payer reimburses prevention in healthy adults. A working therapy behind a closed door contributes zero healthy years.
This turns out to be the finding rather than a footnote. Regulatory readiness has the slowest derived growth rate of anything in the model. Moving any single research field 20% faster changes the answer by under two years. Moving the regulatory term by the same amount changes it by roughly twenty.
Growth rates are derived, not chosen
Each field is scored twice — today and in 2006 — and the logistic rate connecting the two is derived, then damped by ×0.85 because clinical translation is slower than discovery. That makes the forward projection a falsifiable claim rather than a guess: the next 20 years resemble the last 20, slowed.
The 2006 column is the most attackable set of numbers here. That is why it is published.
| Field | 2006 | 2026 | derived r | max contribution |
|---|---|---|---|---|
| Rejuvenation & Regeneration | 12 | 40 | 0.0674 | 0.35 |
| Biomarkers & Diagnostics | 10 | 44 | 0.0831 | 0.08 |
| Geroscience Drugs & Trials | 15 | 33 | 0.0436 | 0.28 |
| Gene Therapy & Delivery | 12 | 36 | 0.0602 | 0.30 |
| AI Drug Discovery | 8 | 52 | 0.1072 | 0.14 |
| Organ Replacement & Biofabrication | 15 | 38 | 0.0529 | 0.25 |
| Immune Engineering & Cancer Control | 18 | 45 | 0.0559 | 0.30 |
| Enabling Technology & Automation | 22 | 50 | 0.0538 | 0.10 |
| Regulatory & Deployment Readiness | 8 | 15 | 0.0301 | gate, not additive |
How this could be wrong
These are the assumptions that actually move the answer. If you want to argue with the countdown, argue with these rather than with the headline.
- Regulatory score of 15/100. Move it to 20 and the central estimate comes in by roughly a decade. It is the single most consequential judgement here.
- Damping of ×0.85. Set it to 1.0 and you get the naive “the past repeats” scenario. ±20% swings the answer by about 24 years.
- Total headroom. The model assumes a mature version of all eight fields could deliver about 1.8 healthy years per year. That is a guess, and it is the largest single source of uncertainty.
- The 2006 scores. Every one is a defensible judgement and not one is a measurement.
Model icd-lev-v0, parameters v0.2.0, 20,000 draws, seed 20260819, base year 2026. Parameters and code are published under CC-BY-4.0. For informational purposes only. Not medical advice, and not a prediction about any individual.