I had to open things in a new tab (in chrome) to see them (odd, but it worked?)

Edit: Seems to be working now anyways :).

For anyone else running into the same problem, I just copied my HipChat.app in /Applications, and ran the second hipchat next to the first hipchat (you can also change the second hipchat's icon for easy distinguishment). It's not perfect (upon logout/restart it still remembers the preferences for the other instance), but it solved 90% of my hassle.

This has been driving me crazy for months, I had one version of the old buggy Adobe Air app running and one of the new Native app. Somehow just creating a copy of the new native app just never occurred to me. You just made my day substantially better.

You can also do in the terminal:

  hipchat --detach &

Or for even any app:

    open -n /path/to/application.app

From the open man page:

    -n  Open a new instance of the application(s) even if one is already running.

That's smart. But still, it's crazy that something like multiple organizations would require you to do this on your end. There are better options out there :)

I've found using hipchat with coworkers to be much nicer than using XMPP with coworkers mostly for it being closer to IRC: there are rooms where most discussion takes place, so you can get to know people by lurking there, and when you need to 1:1 message someone, you probably have some context about who they are, which makes a world of difference to an introverted new hire. The realtime team-wide chat just makes everything feel much more friendly, imho.

I'm sure it's possible to get that setup with XMPP, and it's definitely possible to get an equivalent setup with IRC (which I have an external server running ZNC for), but it's not out of the box, and there is something to be said for having a good configuration set up and working for everyone out of the box.

nice!

similarly,

    var oldPlaceObject = map.placeObject;
    var newPlaceObject = function(x, y, type) {
    	if (type !== 'block') {
        	map.placeObject = oldPlaceObject;
        	map.placeObject(x, y, type);
            map.placeObject = newPlaceObject;
        }
    };
    map.placeObject = newPlaceObject;

was my solution to multiple levels ;).

I also had to overwrite the map.verifyXitems (or whatever it was) to return true regardless.

First, I'm very sorry about that!

Second, wow, a scale factor of 256! Off the top of my head it might be numerical error building up. If that's the case, I'm very sorry. (It's also sometimes possible to get infinitesimally close to the target without technically reaching it mathematically, in which case the interface unfortunately looks like the answer is correct, when it's not.)

All of the puzzles are solvable with dilations of scales only 2 or 1/2.

I thought these were really great, and the UI was nice too, once I figured it out. Some minimal instructions might be helpful for someone thrown directly into the puzzles without having gone through whatever led up to them.

I ran into a situation where it seemed like there was no viable solution which kept the figure under manipulation within the visible bounds at all times, so it was made more difficult by having to visualize where it went offscreen in order to keep track of what I had to do next. It's entirely possible that I could refactor the operations into something that kept it onscreen the whole time, but I couldn't find it, so I really wanted a way to zoom out or pan around, or at least some encouragement like "this should be possible without the shape leaving the visible area." Maybe giving the user the ability to zoom out but also drawing a bounding box that shows the space you're expected to do the transformation in would help.

Thanks for the feedback!

The undo button thing would be a huge improvement; I'll fix that. (Also, if you do two transformations that cancel each other out, it will notice and remove both rather than adding two. So you can actually undo a reflection by clicking the reflector again.)

As some other comments have pointed out, you might have seen the text based dilation, which is used for one of the problems for a specific reason :). The dragging dilation tool is far from perfect, but is a bit closer to what you described.

Maybe more importantly, they would know whether their message was classified as spam (or a pretty good heuristic thereof). A naive bayes classifier for spam relies on the spammer not knowing whether the message was marked as spam or not. Such a classifier isn't difficult to fool if you can test how a given message is classified. (Presumably gmail's spam filter is more advanced these days, but the idea that it is easier to fool if you can tell the result still seems reasonable.)

Here's an alternative idea: what if clients would only honor rel='unsubscribe' links with an HTTPS URL scheme, and only finish the TLS handshake for those requests if the host sends the client a valid Extended-Validation certificate?

Every spammer who wanted to "trick" the auto-unsub mechanism would basically have to first dox themselves for all the world to see. And any certificate that turned out to not be a valid means of contacting the spammer would be quickly revoked.

Well can't they do that anyways, by sending emails to an address they own and seeing what gets classified as spam?

Like most tests, my personal opinion is that it tests both how well you know the material as well as how familiar you are with the format and content of the test. There is value in the former, and hopefully more access to preparation materials can alleviate some of the variance in the latter.

The College Board is releasing previously-unreleased SAT questions on Khan Academy, with answers and explanations [1]. The College Board announcement emphasized the importance of practice, as well as transparency about preparation materials such as practice tests.

While we don't have an ideal test-like practice environment yet, I hope that we will be able to provide one in the future.

[1]: https://www.khanacademy.org/test-prep/sat

I should have emphasized that full-length tests is what I think is valuable here.

College Board has been offering daily SAT questions by email for awhile now (and there's no lack of individual test questions online either), but what's useful is a full test, with questions of varying difficulty and a time limit enforced. Taking it on paper is important too; your score when you take the test in front a screen versus when you have paper, pencil, and a calculator won't be the same. There needs to be dozens of these tests -- I found myself getting better each time, but re-taking tests isn't helpful because I simply remembered my mistakes rather than improving my capacity to anticipate mistakes.

This is not at all to downplay the amazing work Khan Academy has been doing! I think this will be a big deal for a majority of high schoolers.

Thanks! Yes, I agree that full-length tests are important. We're not there yet, but I think we'll improve on that with time (there has been some talk of different styles of how/which/how many questions appear in an exercise for other reasons, and I think this just gives another reason to focus on that in the future).

I don't have any practical experience with angular (but I'd like to, and definitely think its testability is a step in the right direction).

That said, I've found react to be very declarative, especially over the raw jQuery and Backbone I'd used before. Most of the components I write have no state, and render() declares how to render them from any state, without having to worry about how to change things on state changes.