Live·Open questions in longevity research
Method

How we think

The usual way to handle a hard question is to think of a few explanations, argue about them, pick the best one quickly and then defend it. Everything on this site is made the other way round. We look for the places where what is known about ageing runs out, write every explanation that could account for each one, and keep all of them in play until an experiment removes one.

5 open questions38 published rival explanations10 experiments proposed
01

What we are and are not claiming

Nothing on this site has been tested by us. A hypothesis here is a candidate explanation with a stated way of being wrong, not a finding. An experiment here is a written proposal, not a study that has been run. Publication on this site means an editor judged the page readable and honest. It does not mean the science is right.

Where a claim rests on a source, we show the source, and where a figure in a hypothesis has no source behind it, an automatic audit says so on the page rather than quietly letting it pass. Support in one experimental system, a mouse or a cell line, does not establish an effect on human lifespan, and we do not write as though it does.

02

From a goal to an experiment

Eight steps, each one owned by a different specialist. The numbering is the engine's own.

Step 1 · Initiator

Restate the goal as a question that can be worked on

A goal like "radically extend human lifespan" cannot be attacked directly. The first step rewrites it as a precise master question with a baseline, a horizon and a measure of success. This step is also where a goal can quietly become a different goal, so every run is audited afterwards against what the operator actually typed.

Step 2 · Immortalist Architect

Ask what would have to not fail

Rather than listing things that would be nice to have, this step works backwards from failure: what are the ways the goal is defeated, and what would have to hold for each of those not to happen. The result is a small set of goal pillars, each one named after a failure mode.

Step 3 · Requirements Engineer

Break each pillar into requirements that can be checked

Five to nine atomic requirements per pillar. Atomic means a requirement can be satisfied or not without argument about what it meant.

Step 4 · Domain Mapper and Domain Specialist

Find the fields that own each requirement

Seven to twelve domains, chosen so they do not overlap and do not leave holes, and then fifteen to twenty-five scientific pillars inside each domain: the mechanisms the field actually has, how ready each one is, where it is fragile, and what could be imported from a neighbouring field.

Step 6 · Strategic Science Officer

Find where the knowledge runs out

This is where a question is born. The step descends to the points a literature search cannot answer, and types each one: a void where nothing is known, a fragile claim, a proxy that may not track what matters, or a clash between two established results.

Step 7 · Instantiation Gatekeeper

Write the whole field of explanations, not the favourite

Four to seven rival mechanisms for every gap, drawn from different domains. The set is required to contain a heretic that contradicts the mainstream account and a transfer from an unrelated field. Nothing is dropped for being unpopular, and nothing is ranked at this stage.

Step 8 · Lead Investigative Officer

Turn the contest into questions an experiment can answer

Tactical questions, at least half of which have to discriminate: an explanation that predicts exactly what its rivals predict has nothing to contest and earns no experiment.

Step 9 · Lead Tactical Engineer

Specify the experiment, and what a null result would mean

Each experiment names a system, an intervention, a meter and a threshold. It also has to say what we would learn if the result comes out null — which is why every experiment on this site carries two readings rather than one.

03

The five ways knowledge runs out

Every question on this site is typed as one of these, because the five need different work.

Void

Nobody has looked. There is no result to argue with, only an absence.

Fragile

Something is claimed, but the evidence under it is thin enough that the claim could simply be wrong.

Proxy

What is being measured stands in for what matters, and it may not track it.

Clash

Two well-run studies disagree, and the field has not resolved which is right.

Adversarial

The question was written specifically to break the answer everyone expects.

04

Why we keep the losing explanations

An explanation earns its place by stating the single observation that would knock it out. Every pair of rivals under a question is bound to one observable that would come out differently under each, and that is how an experiment gets chosen: it is the one that separates the most rivals for the least money.

A rival that loses is not deleted. It may carry a weak mechanism and an excellent experiment, or a strong idea spoiled by one doubtful assumption. Both are worth keeping and worth reading, which is why the full set stays on the question page rather than only the front-runner.

Two parts of this design are not built yet, and we would rather say so than imply otherwise: rivals do not yet carry ratings from head-to-head comparison, and no new generation is bred from the parts of losing explanations. Every rival you read is a first generation.

05

Where to start reading