An open notebook: I'm thinking out loud and tracking the literature, not writing a settled explainer. Expect it to change shape as I read more — and tell me where I'm wrong.

The puzzle, stated plainly

Sleep looks like a terrible idea. For roughly a third of your life you stop foraging, stop reproducing, and go blind, deaf, and nearly paralysed while predators roam. Evolution is ruthless about waste, and yet sleep is everywhere — mammals, birds, reptiles, fish, even fruit flies, even the brainless jellyfish Cassiopea shows a sleep-like state. When something that costly is that universal, it's almost certainly doing something essential. We just don't agree on what.

The leading hypotheses

I'm tracking four. These aren't mutually exclusive — sleep may do several at once — but each makes different predictions.

1. Synaptic homeostasis (the "reset" idea). While awake you strengthen synapses as you learn; this can't ramp up forever. The proposal (Tononi and Cirelli) is that sleep globally down-scales synaptic strength, preserving the relative pattern while restoring tomorrow's capacity to learn. In my head, this is the strongest single candidate.

2. Memory consolidation. Sleep replays the day's activity and moves memories from fragile short-term storage toward durable storage. That sleep helps memory is solid; whether that's the reason sleep evolved, or a useful passenger, is the open part. My read: a real effect, but contested as the primary function.

3. Metabolic / waste clearance. The brain's glymphatic system appears to clear metabolic byproducts (including amyloid-β) more efficiently during sleep. Appealing because it gives sleep a concrete janitorial job. My read: promising, but the magnitude and some key results are debated — and I haven't yet read the recent challenges carefully.

4. Energy conservation / forced downtime. Sleep might simply be the cheapest thing to do when you can't usefully act (no food to find in the dark). My read: probably part of the story for when we sleep, but weak as an explanation for why it's so deep and risky.

Where I currently lean

Today, my honest position: sleep is not one function. The synaptic-reset and memory-consolidation stories are probably two views of the same nightly maintenance, and clearance is a plausible third job riding along. I hold this loosely.

What I'm explicitly not claiming: that any one of these is "the answer." Anyone who tells you sleep is solved is overselling.

What would change my mind

• A clean experiment dissociating synaptic down-scaling from memory benefit — if you can get one without the other, the "two views of one thing" claim is wrong.
• Strong replication (or refutation) of the glymphatic-clearance magnitude would move clearance up or down a lot.

Open threads to chase next

• Read the recent critiques of glymphatic clearance and update hypothesis 3.
• Find the best case for sleep in animals without centralised brains — does it strain these theories?
• Add references. (Deliberately not yet cited line-by-line — that's on the list.)

Changelog

• 2026-06-20 — Added the jellyfish example; softened hypothesis 3 after seeing it's more contested than I first wrote.
• 2026-04-15 — Started as a bare list of four hypotheses.