So, the London community is arranging a Hackday where some of us will get together and code. In order to ensure we work on the awesomest idea(s) possible, we decided to ask LessWrong to add to our list of candidates. So here is the question:

What could a few developers do in a day or less worth of coding that will be awesome? Also, as a way of checking calibration, you can give your estimate for how long such a thing would take to build.

Note: While we will take ideas and voting here into account, there is no guarantee that we will actually end up choosing one or more of them.

32 comments, sorted by Click to highlight new comments since: Today at 2:29 PM
New Comment

A poll function built into the less wrong code.

I like the idea of the Serious games based on biases.

This seems to be polling well, so we might end up working on it. Would you be able to give more details on what we would actually code, or what we would need to do to reach some codeable spec?

It's probably something that requires a lot more thought... it'd probably take us the day just to come up with ideas and designs (preferably by somebody that knows how to design game-play).

I'd consider starting with something simple like the "predicting red or blue" card game focusing on calibration, mentioned here: Worse than random

Ok, I've created a discussion topic (my first!) at: Designing serious games

Rationalist clue would be fun too... but much longer to implement... almost certainly not on our timescale...

I will vote up suggestions that estimate some utility function value for the proposed projects.

E.g. Lives saved per programmer hour or 10 -9 percentage points chance increase of a positive singularity .

Your reasoning can be indirect, e.g. a simple successful project will encourage future hackdays that can potentially reduce the chance of X existential risk by 0.01%.

A bot to do the reposting of the sequences to the discussion section.

I had written about bringing back the sequences, which was well received. The best way to implement that without messing with the LW codebase is to make a scraper to extract the data from the wiki and a bot to post in the discussion area on a regular basis.

I predict this should be doable by a couple of people working for a day.

The benefit would be that it would allow new members to read the sequences in a fun way, and the more experienced of us to re-examine the sequences in light of having absorbed most of the material and the insights that came after. In the spirit of the lens that sees its flaws, I expect us to find places that the sequences could be improved, in light of the sequences themselves.

I don't like the idea of straight reposts.

  • Eliezer deserves a hell of a lot more karma for his sequences than he's got. A bot reposting them would give the bot the karma.
  • There is a lot of useful discussion on the existing posts in comments. That would be lost.
  • You'd split incoming visitors between the two different versions of the post, depending on where they came to it from. People could thus easily miss out on relevant discussion / discussions could be fragmented.
  • Less of an issue here, but not having a canonical URL for the articles bugs me. Everything else aside, it penalises page rank on the search engines if links are split between different urls.

I think this would be much better done with a code change for the site, allowing editors to pick a featured article that sits at the top of the main page. My estimate would be 1 day for someone familiar with the source, maybe 2 for someone unfamiliar. Of course those estimates should be at least doubled to 2 / 4 days.

While I'm unfamiliar with the code myself, display code for the current featured article should be simple. The other, harder parts of the task would be an interface to allow editors to choose a featured article (but not necessarily that complicated, you might be able to take a url for instance and then editors could find it with a simple search), and storing the current featured article.

It woudnt be a straight repost. It would be a summary with a link. People will be directed to the original and encouraged to upvote that. As for the discussion, I am of two minds: yes it would be good to add to the existing, but there is something too be said for having a blank canvas.

But all this is moot. Nobody with commit privileges is around, and even if the changes were made (which will need to touch the database too btw), it would be a crapshot as to whether they would actually be applied. While I agree that code modifications would be ideal, as far as I am concerned, the better is the enemy of the good.

Well, how about rather than reposting, we just add a "featured sequence post of the day" section, or something. It'll look just like "recent posts", but it'll go through the sequences one by one.

Possibly we could just move the "promoted" functionality into the sidebar... I have to admit I never actually click on anything in the header... but I click a lot on Right-hand navigation.

No reposting so no double-ups and karma stays where it is, it's just a different way of featuring existing posts.

Hmm.. if this drew data from an outside source (like the link to OB does, a feed maybe?) we could avoid going too deep into the code. This could be little more than a template edit plus a feed hosted somewhere else. Interesting.

An alternative that I've seen some long-running web comics offer is to provide a personalised RSS feed of old postings that will supply a subscriber with all of the material, at a user-selected number per day.

That's pretty cool, and I have wanted to build a generic system like that for this exact reason (the sequences). I stopped having such an intense desire after finding , but I think a big part of the issue here is taking the whole group through a specific pace, so that there is a social experience involved to make the whole re-reading more fun.

Might be a silly question, but why is a bot doing this better than just creating a new account and someone manually cutting and pasting the posts: surely that would only take a few minutes work per article at most?

Nothing more than saving us the manual work, and perhaps offering a little more continuity since humans (especially me) are not particularly consistent about routine tasks. The alternative is indeed to do it like the rationality quotes threads, make it a community effort. But why make people do a computer's work if they don't have to?

Time to crunch some numbers: ~250 sequences posts at 5 minutes each makes 20 hours of work in reposting the sequences manually.

Since you reckon this could be done by two people in a day, I guess that's less than 20 hours' work: added to the fact that writing the bot is almost certainly more fun than copy-pasting the sequence, it looks like the bot gets my vote.

Just to pick a nit, I am under the impression that there are much more than 250 posts in the sequences given that Eliezer posted daily for about two years. (Of course we could stop speculating and count, but I can't be bothered) That would move the estimation much more in favour of a bot, but if its only feature is to save 40-50 hours, maybe there is something else we could do that simply isn't doable by humanpower.

Not only that, but 20 hours in one stretch is much easier than 20 hours stretched over however many months... making sure of a consistency that invariant with mood, sleeping-in and natural disasters... bots are much better at the latter.

This probably requires specific knowledge that might make it unsuitable for a group project, but I really like the idea of someone making an Effiicient Charity smart phone app - push a button and some (user defined) amount of change goes directly to your chosen optimal charity (probably with Givewell's top-ranked charity as the default).

Part of the gimmick would be that when you get that little pang of guilt from not giving to someone collecting for charity in the street/beggars/sponsoring your friend to run the marathon, you can easily alleviate it by giving to a useful charity instead.

Unfortunately, I doubt this is the sort of thing on which a small group could produce something awesome in one day, and it probably also requires iOS experience if it's going to get to as wide an audience as possible.

Ideally the process would be slightly longer than pressing a button, to increase the feeling of having done something.

Nice idea. Possibly for the future. We would also need a Mac to develop it on, and a cert.

Would it run foul of Apples 30 percent cut rule in the app store?

What I found out about charity apple wasn't very encouraging.

A predictive compression engine. Fun and educational to work on, testcases are easy to write (compression followed by uncompression is the identity), lots of existing papers and code, and if you write the center part well, you can accept code drops of specialist agents, taking advantage of their expertise in some situations without allowing them to degrade your aggregate performance in other situations.

Applications: a shell that predicts what the user will type, finding patterns in logfiles, roshambo and similar games, predicting wiggly graphs, Schmidhuber's artificial curiosity, probably other things too.

As discussed at the meetup: attacking some of the low-hanging bugs in the LessWrong code, e.g. fixing the karma circle. There should be a list. If there isn't, compiling one beforehand would be good. (I would if I were a coder, which I'm so not.)

What's the karma circle?

The little green circle at top-right showing your score in the LessWrong video game! The one Eliezer has overflowed so much it just shows "1023".

That's more of a graphic design problem than a coding problem. But nevertheless, just about every project has low-hanging fruit amongst the bugs and feature requests.

hmmm - won't format nice but still:

main.css line 424

#side-status div.userinfo span.score { -moz-border-radius: 12px; -webkit-border-radius: 12px; -khtml-border-radius: 12px; border-radius: 12px; height: 22px; line-height: 22px; margin: 0 3px 0 0; font-weight: bold; text-align: center; min-width: 22px; background:#538d4d; color: white; }

also - you can remove the other ".score" CSS rule too.. now superfluous.

PS - won't work in IE... but do we care?

The circle under a username (top right) showing the karma, it isn't readable for high Karma values, like Eliezer's.

[-][anonymous]11y 0

Very minor issue, but the sidebar is also a minute or two behind regarding post/comment/user scores.

Adding agree/disagree buttons to comments (with some kind of summary of the total) in addition to vote up/ vote down.

I don't know how long it would take, but I really, really want a way to make Cyberduck (open-source FTP client; I use the Mac one) respond to global keyboard commands, so I could upload things to the current directory while I had them open or selected in Finder without clicking and dragging. I really want that a lot. Anyone who can make that happen for me wins my undying gratitude.