Hacker News Reader

Comment by DEADBEEFC0FFEE | Hacker News Reader

I definitely used to be a hardcore finisher. It's a really difficult habit to break, as it feels like personal failure, and violation if principle. My wife was also, and independently a book finisher.

I don't recall the realisation that life is short and there is so much to do. Neither of us are finishers now. But I do have a special shelf for unfinished books. They stare at me.

View on HN

I used to be, too. Here are the two easy steps I took to break that habit:

1. Start reading Atlas Shrugged. When John Galt starts making his speech, skip to the next chapter. You have now gained the ability to skip pieces of a book that are really, incredibly tedious.

2. Start reading Twilight. When it starts feeling like torture to continue, stop. Use the rest as fire starters for your grill. You have gained the ability to not read books you discover that you don't like.

Take those unfinished off the shelf and get rid of them. Sell them to the used independent. Donate them. Burn them. Landfill them. Whatever. You can reclaim that shelf space for books you not only want to finish, but re-read and finish again.

ive seen twilight books in curbside as trash

Did the garbage collectors take them with the regular pickup, or did they stay until hazardous waste day?

one thing i started doing in recent years is just borrowing heaps of books from the library. trawl the catalogue, see what's ready for borrowing, borrow half a dozen ones that look interesting or are on the reading backlog. then start reading a few of them in parallel. if some of them don't hold your attention, stop reading and focus on the ones that are interesting. take them all back to the library when they're due, read or not.

edit:

okay, so now for the obvious question, what's the correct way of framing the which-books-should-be-finished problem, and what's the optimal strategy?

is this a bandit algorithm thing? at each interval t, sample a page from a book according to gittins indices to minimise lifetime regret? does it need monte carlo tree search?

Not a classical bandit because the arms have a finite lifetime (the arm is no longer active once you read the whole book) and you know this lifetime. Definitely can be viewed as some form of RL problem. For me the challenge is how you define rewards and hence what information reading part of a book gives you about the total reward for completing the book. Maybe assume some smoothness criteria, so that the reward for reading part n of a book is highly informative of reading part n+1, but less informative of part n+2, and so on. More formally, each book gives some total reward on completion and that reward is distributed according to some stochastic process.

Greedy strategies probably work reasonably well.

Massively Parallel Speculative Evaluation. I like it!

I usually have a stack by my bed. Each time I get bored with a book, I put it on the stack and start another one. When I finish, I pop the stack. Maybe now its interesting? It usually works, but it can take 6 months or a year sometimes to get back to bare table top. And the stack can get really tall.

In your system, one might not even be compelled to finish any book at all. Instead, read as much as you want, at any pace you want, and drop the book once you don't see any need to read it further - maybe you got bored or felt you've extracted everything meaningful out of it.

Reply on news.ycombinator.com