I find this interesting but confusing. Do you have an idea for what mechanism allowed this? E.g: Are you getting more done per hour now than your best hours working full-time? Did the full-time hours fall off fast at a certain point? Was there only 15 hours a week of useful work for you to do and the rest was mostly padding?

Yes. I don’t think the argument requires that the work be hard (or that you work hard at it, whatever that really means). I believe it’s quite generally true that for most activities (howsoever achieved), productivity drops as hours spent increases. Then the rest of the argument follows.

I don't think productivity ever increases with contiguous time spent on a given activity, at least in the short term. (Yes, longer term you can identify and implement working patterns that increase overall productivity.) All the effects push in the opposite direction: low hanging fruit gets picked first, you get tired, you get hungry, you've done all you can on your part of the critical path and need to wait for others to do their bit before you can continue, etc.

So I think there are two explanations here if someone on Hacker News does claim that the last hour is the most productive. The first is post-rationalisation, i.e. if you have been working crazy 99 hours straight on something you'd better conjure up some pretty convincing-to-you-sounding reason to work the 100th hour. The second is that you are falling for a perceptual trick: let's say you're working 99 hours straight on finding a difficult bug, and in the 100th hour you fix it. You can say "wow, that 100th hour was clearly the most productive, because in the first 99 hours I didn't solve the bug and in the 100th hour I did". But that's not what productivity means. What is actually happening is you are making slower and slower progress during the 100 hours but still eventually get there.

Personally I have found that judging my position on the curve is pretty easy, if you are mindful and deliberate about doing that continually. This mindfulness and deliberation doesn't seem to happen automatically, though, which makes it difficult if you don't do that.

Apologies for using ‘engineering speak’ without explaining. The transfer function relates the output of a process to its input. In this case the output is literally your total work output (y axis), and the input (x axis) is the time you spend producing that output.

The shape of this curve is well established, whereby initially the gradient is quite steep (that is, the first hours you spend doing a task, you get a lot done) but the gradient quickly starts to flatten.

I like the idea of the 4-day work week, but this post is actually a quite separate argument.

The 4DWW idea is: work less, and you'll be happier as a direct consequence.

The argument in this post is: if you want to work X hours a week, whatever that X is, go for it! But rather than spending X on one job where you're almost certainly spending a significant proportion of X in the diminishing returns regime, split it into e.g. 0.8X on that job and 0.2X on a completely separate job. The main effect of this will be productivity gains, which in turn will lead to increased happiness as a side-effect.