Science & health, explained

What the evidence consistently supports

Across decades of research and many populations, four things repeatedly predict better long-term health outcomes. They are unglamorous, which is why so much is sold alongside them.

• Regular movement. The largest gains appear between doing nothing and doing something modest. Going from sedentary to a daily walk moves the needle more than going from fit to very fit.
• Adequate sleep. Consistently short sleep is associated with worse outcomes across nearly every system studied, and it degrades judgement before people notice it has.
• A diet built mainly on whole foods. The specific framework matters far less than the pattern. Most diets that work share the same mechanism: they reduce heavily processed food and are followed for long enough to matter.
• Maintained social connection. Isolation shows up in long-term health data at a magnitude that surprises people unfamiliar with the literature.

Nothing marketed as a shortcut has matched these. Anything specific to your own condition, medication or family history belongs with a clinician who can see your history — not with an article, and not with a chatbot.

Reading a study without a science degree

You do not need statistics to filter out most bad reporting. Five questions do nearly all the work.

1. Who was studied, and how many? Twenty undergraduates is a pilot. Findings in mice are findings in mice; the headline usually omits this.
2. Was it observational or a trial? Observational studies show association — the two things move together. Only a controlled trial supports a causal claim. Most nutrition headlines quietly cross this line.
3. Has it replicated? A single paper is a data point. The finding is the replication. Fields have learned this expensively.
4. What are the absolute numbers? "40% higher risk" is meaningless alone. If risk rises from 1 in 10,000 to 1.4 in 10,000, the relative figure is accurate and the impression it creates is not.
5. Who funded it, and what was pre-registered? Funding does not invalidate a result, but a study whose outcome measure changed after the data came in deserves scepticism.

Weather, climate, and the difference

Weather is what happens this week. Climate is the statistical distribution those weeks are drawn from.

A single cold winter no more contradicts a warming climate than one tall person disproves an average height. Climate claims are statements about how a distribution is shifting — its mean, and increasingly its tails, which is where the damage concentrates. A two-degree shift in an average sounds trivial precisely because it is an average; what it does is make the previously rare extreme substantially more common.

This is also why "the climate has always changed" is true and beside the point. The scientific question has never been whether it changes, but how fast, in which direction, and whether the infrastructure and agriculture built around the old distribution can adapt at that speed.

Why launch costs collapsed

Reusable first stages changed the economics of reaching orbit more than any advance in propulsion. The physics of getting to orbit did not improve; the accounting did.

In an expendable rocket, the most expensive hardware is destroyed on every flight. When that same hardware flies repeatedly, the cost per launch falls by roughly an order of magnitude, and the fixed cost of building it amortises across flights instead of being written off once.

That single change explains why several unrelated things appeared at nearly the same time: large satellite constellations, cheaper science missions, commercial Earth imaging, and private launch providers. None of them became technically possible in that decade. They became affordable, which for space has almost always been the binding constraint.