Ask HN: Review my side project: Sqrt(2) | Hacker News Reader

Ask HN: Review my side project: Sqrt(2)

10 pointsdrewcrawford18 comments

Hey there, HN! In our spare time, we've been working on this side experiment: what would happen if a social news site required people to demonstrate intelligence in order to submit / vote on news stories?

Thus is born Sqrt(2), here's 100 invites:

http://1.414214.com/login.html?m=newsyc

We're really concerned about scaling, so we'll increase the ceiling on the invite code as quickly as we can confirm our server's not going to explode. If you can't get in, try again in a few hours.

Heading off a couple of things:

Yes, our presentation sucks. How can we make it better without annoying everyone with a long-winded description every time they log in? We've agonized over trying to make the WTF page clear without wasting anyone's time, but you guys are the test of whether it actually makes sense. Remember, the audience is smart, cool people such as yourselves, not grandma.

"But I don't have time to solve crappy puzzles!" I'm busy too. But I enjoy solving little puzzles as a break from coding. If you don't "have time" to do X, it's because you don't make time to do X, because you view X as inferior to whatever else it is you're doing. Which is fine. But I want news from people who occasionally make time for logic/critical thinking, so if you don't, I don't particularly care for your news. No offense. Complaints about the specific nature of the specific puzzles are, of course, perfectly valid. I've coded some things that interest me, but I would love suggestions on what interests you.

Is this going to work? Has something similar been done before? We have no idea! It just seemed like a neat experiment and it was painless enough to code. What do you think?

Nice to see that I'm not alone in trying to build a better social news site as a side project. And by the way, when you say "we", we all know there's only one of you. And forget about scaling. Learned that long ago myself...

My side project is at http://mushpot.net, and I think the goals are similar to sqrt(2):

- multiple votes

- make it harder to game the system

- give more productive members more influence over the content

But instead of having users solve puzzles, on Mush Pot, the amount you get to vote depends on how many votes your submissions have received. And since I'm already started, here are some of the other ideas behind it:

- long posts, not just links (posts can have hyperlinks)

- there's no difference between top-level posts and comments..anything can appear on the front page

- make users provide a reason for downvoting (or pile onto a reason given by another user)

- tags, ability to weight your front page using tags

And it also lets you login with Facebook and post your submissions back to your Facebook feed.

Wow, I'm impressing myself with the length of this feature list, although I have been building the site for quite a while. Now, everyone, at the count of three...click...HERE!: http://mushpot.net

Actually, there are three of us: http://www.experimenthouse.com/bio (Come on, the link was in the profile... seriously?)

Not to start a flamewar, but did you really look? We've got markdown support (which means images, links, long posts, emphasis, lists). We also hand out currency based on received votes (as well as puzzle solves).

The "anything can hit the front page" is an interesting idea though. I thought about that, but I wondered how well it would work for discussions, which often don't make sense out-of-context. I poked around, but I didn't see any comments hit the mushpot FP. How well does that work in practice?

Instead, we moved a few "best" comments for each story into the RSS feed. Hopefully that will give good comments high visibility while also providing some context for what's going on.

I registered just to see what the puzzles were. Some thoughts:

I found that you can answer the same puzzle multiple times, which you probably want to change.

For three of the puzzles, if you submit a blank form the website breaks.

You should make it more obvious how many sickles a user has. And why are they sickles anyway?

I don't know if this breaks your idea, but maybe add trivia questions as well as puzzles. That might be helpful when you're struggling to find new puzzles, and if a theme built up it might help build a community.

Good find on the blank submission. :-) Just pushed up a new build

Answering the same puzzle multiple times is by design: most of them randomly generate (except meaningoflife). And in theory our algorithm scales your reward based on the real-world difficulty of the problem. I'm sure it will take some tweaking, but it looks like meaningoflife is handing out 5-10 sickles at the moment, spiralling downward with each solve, while the resistor lattice is in the hundreds of sickles.

Sickle balance is on the page footer. Did you miss it? Should we move it to the top somewhere?

What kind of trivia questions do you want to see?

I didn't realise most puzzles change, I just answered the Meaning of Life three times and assumed I was cheating.

As for the sickle balance, I did find it, but think having username and balance on the top right works well on HN. Somewhere near the top would be an improvement on having it in the footer.

I didn't have any trivia questions in mind, I just figured the kind of people who like puzzles are probably mostly the same as those who like trivia. Maybe that's too easy to game though.

Your first puzzle sucks: I hate registering to see things. Let us try it right now!

You really need a not-logged-in version.

I agree, you can use OpenID, it's a lot faster to register.

Yes, for the .1% of people online that use OpenID. OpenID is a failure.

I agree. We're paranoid about scaling right now (loading the frontpage is a massive task, even with memcache), so that's why it's login-only for right now.

Generally speaking, it's better for new web apps to be paranoid that no one will care about your app, vs paranoid too many will be.

I've thought about this for a long time. (I've been an online moderator since 1992, on a variety of networks.) I would prefer a system of testing people on specific knowledge of specific subjects, and giving them visible badges to show other users they know about, for example, economics or education policy or computer industry issues. Then use the people's knowledge badges to weight karma scoring (and introduction to moderation privileges) behind the scenes. If you program it, I and my friends will visit the site.

An argument could be made that the resistor lattice is a "specific knowledge" test of electricity, and that mastermind is a specific knowledge test in combinatorial logic.

What type of knowledge are you looking to assess specifically? Economics and education policy are particularly troubling to assess objectively (I'm a fan of the Austrian school, but there are a few smart Keynesians, for instance). How do you assess economics knowledge? What sort of questions would you ask?

You have to let users see the site first to get them interested. Logging in should be for posting privileges, not just for viewing.

See my other comment in this thread about why I prefer specific knowledge tests to puzzles. Puzzles neither screen out the people who most need screening out nor draw in the people who can add the most value to the site if drawn in.

is it a bug or feature that I can keep just answering one puzzle over and over and build my "sickles"?

Also it seems pretty that this could be hi-jacked by a cabal of users sharing answers to puzzles and "block voting" on stories. I like the general idea though and I am trying to think of a suggestion for how puzzles could not easily be manipulated like that.

Maybe having a much larger pool of generated puzzles which when answered correctly are taken out of circulation permanently? At least this would make people looking for answers have to learn the general theories behind the puzzles and thus in a round-about way rehabilite the cheaters. haha!?

If you're talking about "meaningoflife", that's mostly something we're using to benchmark our difficulty algorithm. The difficulty is approaching zero pretty quickly, meaning it won't be handing out sickles for very long. Sorting out problem difficulty algorithmically instead of arbitrarily makes it harder to game the system--if someone writes a bot, difficulty will go to 0 pretty quickly without us having to do anything.

Most of the other ("realer") puzzles randomly generate after you solve them the first time. This keeps the difficulty (and sickle handout) up.

I have been sporadically working on a site quite like this one, but with a very different puzzle type. Congratulations on bringing out a working first!

Interested in your project. Care to share any details? E-mail in profile

That's incredibly hilarious.

Reply on news.ycombinator.com