The vaccines are still quite effective after the first dose, and, I'm saying this without being a researcher in this area, should be enough to flatten the curve.

"The Moderna vaccine has been shown to have an efficacy of approximately 92 per cent in protecting against COVID-19, starting 14 days after the first dose." - https://www.who.int/news-room/feature-stories/detail/the-mod...

Wow that's impressive, I thought it was ~50% after the first dose. I guess it's like ML, to get the last percentage points you need exponential effort?

You might be thinking of this paper about very extensive data from Israel: https://www.nejm.org/doi/full/10.1056/NEJMoa2101765

One conclusion a friend of mine wrote about this is "One other surprising thing which is not obvious, but is made very clear from the results, is that it really takes real time for this vaccine to get effective. It only starts becoming truly useful after 2 weeks, but even then keeps getting better with time. This effect is one reason why there was a very real lag between when Israel got vaccinated at a very high percentage of population and when the cases started freefalling (2-3 weeks+). "

It's 50% if you start counting immediately, before the first dose has had time to work.

> I guess it's like ML

It's very like simulated annealing.

It isn't that simple though. The first dose still provides some protection. The choice is usually between a lot of people having some protection, vs a smaller amount of people having full protection.

(and some protection can be in the 70% range [1])

[1]: https://www.thelancet.com/journals/lancet/article/PIIS0140-6...

Yes, rather than "some", I would say that the first dose provides the majority of protection, and the second is a booster, upping the effectiveness from the 70s to the 90s. If you are short vaccine, it makes a lot of sense to give out only a single dose to more people than both doses to a smaller group of people and then circle back when more is produced. The only question is whether it makes sense to circle back if the time elapsed is long, and this study is evidence that it does.