[00:00.000 --> 00:01.840]  Thank you for that kind introduction.
[00:02.780 --> 00:09.040]  So we're going to talk about a bunch of things here, and at the end we'll talk a bit about what it is we're doing in particular.
[00:12.140 --> 00:17.000]  But we're going to start off just talking generally about information overload, right?
[00:17.200 --> 00:24.060]  So there's like way too much stuff to read, not enough time to read it in, and we have basically a last foot problem.
[00:24.220 --> 00:26.640]  And as you pointed out, we're not talking about feet, foot.
[00:27.460 --> 00:29.980]  We're talking about getting it from here into here.
[00:30.000 --> 00:31.540]  And that's the bottleneck.
[00:32.160 --> 00:46.620]  And we'll talk about sort of historically how text has changed over time to be more efficient for the particular media that we're using to read in, and how we've transitioned to screens and how there's unique challenges and opportunities associated with reading on screen,
[00:46.820 --> 00:53.940]  and some of the new technologies that are working to make it more efficient to read, especially now because we have so much stuff to read.
[00:55.360 --> 00:59.000]  So if you're like a lot of us, you wake up with something that looks like this.
[00:59.480 --> 01:00.500]  Good evening, HOPE.
[01:03.450 --> 01:05.070]  What a knockabout of pure fun that was.
[01:07.210 --> 01:07.610]  Okay.
[01:08.110 --> 01:13.030]  Before we get on with tonight's program, there's just a couple of quick announcements.
[01:13.430 --> 01:16.710]  Once again, midnight tonight in the no other room right over that way.
[01:17.310 --> 01:18.270]  Hackers got talent.
[01:18.670 --> 01:26.370]  Whether you want to just spectate or whether you want to be part of it and take a five minute slot to present us some talent that you have with some kind of hacker flavor to it.
[01:26.630 --> 01:28.030]  What that means, I leave up to you.
[01:29.470 --> 01:33.090]  Sign ups might be still happening in the info desk or just show up.
[01:33.170 --> 01:34.210]  There might be some slots left.
[01:36.570 --> 01:39.850]  Okay, so information overload has ended.
[01:41.170 --> 01:45.550]  So we all wake up to something that looks like this.
[01:45.750 --> 01:49.550]  You know, you got tons of email to read and you got news to catch up on and all of this.
[01:50.830 --> 01:55.790]  And, you know, unfortunately, humans are actually not naturally good at reading.
[01:56.330 --> 02:00.330]  And if you look back historically, like, we're really good at talking.
[02:00.550 --> 02:02.150]  We've been talking for a long time.
[02:02.150 --> 02:04.150]  Humans have been talking for about 100,000 years.
[02:04.890 --> 02:09.670]  And if a baby is born and it's just around other humans, it will naturally learn how to talk.
[02:09.830 --> 02:10.870]  Like, you don't have to teach it.
[02:10.970 --> 02:13.490]  It just happens because we're hardwired for language.
[02:14.290 --> 02:17.330]  We are not hardwired for reading and writing.
[02:17.870 --> 02:20.370]  And we've only been doing it for about 5,000 years.
[02:20.770 --> 02:31.590]  And it would probably surprise most people in this room who speak English and other common languages to know that 75% of languages, even in existence today, have no written component.
[02:32.110 --> 02:35.270]  So there's just, like, no writing associated with it.
[02:35.310 --> 02:37.750]  Because we're not hardwired for reading and writing.
[02:38.230 --> 02:52.170]  And one of the things that follows from this sort of short history of reading and writing is that 5,000 years is not long enough for us to have evolved or to have, like, you know, evolutionary pressure to select for good reading ability.
[02:52.370 --> 02:53.830]  So we just kind of existed.
[02:54.150 --> 02:57.050]  And you came with whatever reading ability you have.
[02:57.690 --> 02:59.530]  And then we created reading and writing.
[02:59.790 --> 03:02.630]  And there's a large diversity in reading ability.
[03:03.030 --> 03:06.310]  For various reasons, both visually and cognitively as well.
[03:07.050 --> 03:11.370]  So, as one neurolinguist likes to say, humans didn't evolve to be efficient readers.
[03:11.370 --> 03:18.090]  Rather, our writing system has evolved to be good for the hardware up here that we came with.
[03:19.630 --> 03:29.490]  Zooming in a bit on the writing portion, we'll take a brief look at sort of how things have evolved and then how the writing systems have been efficient for how it has been writing.
[03:29.970 --> 03:33.770]  And then also looking at the things today and how we can make it more efficient for that.
[03:34.030 --> 03:35.590]  So we started out with pictographs.
[03:35.650 --> 03:37.550]  That's, like, stuff that looks like things in the real world.
[03:37.550 --> 03:38.810]  So a dog looks like a dog.
[03:39.370 --> 03:42.930]  That was great because it's efficient because you know what it is you're referring to.
[03:43.150 --> 03:44.970]  But you have a limited sort of vocabulary.
[03:45.670 --> 03:47.970]  We moved to alphabets a long time ago.
[03:48.230 --> 03:53.910]  Those are great because it takes a lot less time to become literate if you're in an alphabet culture.
[03:55.210 --> 03:58.370]  Way back when, you know, people were like farmers and stuff.
[03:58.550 --> 04:04.870]  And it would take years and years and years of dedicated study to learn all the thousands and thousands and thousands of pictographs.
[04:05.390 --> 04:07.070]  So most people were not literate.
[04:07.570 --> 04:12.610]  When alphabets came around, you know, you've got roughly 30 characters to memorize and they associate with sounds.
[04:12.910 --> 04:19.330]  And then it's a lot quicker to be able to be literate because you just have to know what sounds correspond with what letters.
[04:19.510 --> 04:20.870]  And then you can sort of piece things together.
[04:21.110 --> 04:26.530]  So the alphabet was great in terms of democratizing literacy and getting a lot more people to be able to read.
[04:28.230 --> 04:30.930]  Next came the printing press and around 1400.
[04:31.450 --> 04:35.050]  This is great, obviously, because it makes printing a lot less expensive.
[04:35.410 --> 04:37.650]  You don't have to have hand-scribed books.
[04:38.150 --> 04:39.810]  It enables things like newspapers.
[04:40.010 --> 04:43.710]  So again, really broadening the circle of the number of people who can read.
[04:44.850 --> 04:47.810]  And then lastly, digitization came about much more recently.
[04:48.150 --> 04:57.490]  And this broadens it even further because instead of having to print a million newspapers and ship them around the world, we have a website and a million people from all over can look at it at the same time.
[04:57.930 --> 05:03.310]  So again, lowering the cost of becoming literate and broadening the circle of the people who are literate.
[05:04.130 --> 05:16.450]  So these are all great developments because they bring reading to the masses, but they also brought massive amounts of reading because now everybody can read and write and, you know, your boss can send you an email and your friend can send you an article and say,
[05:16.510 --> 05:17.070]  hey, look at this.
[05:17.190 --> 05:19.390]  So there's just a lot more stuff to read.
[05:20.350 --> 05:26.850]  Looking briefly at the media in which we read, we started out with tablet things, not the kind of tablets we have today.
[05:28.390 --> 05:32.270]  And then we moved on to papyrus and paper, and then recently we have different kinds of screens.
[05:32.430 --> 05:40.250]  And I've broken out smartphones here because, as you'll see, the different form factor size has different advantages and disadvantages.
[05:40.610 --> 05:46.150]  And so as we optimize the way we read and write for different types of screens, the size is actually relevant.
[05:48.530 --> 05:55.790]  So here's one example of how we've evolved reading and writing to be efficient for the medium that we're using.
[05:56.170 --> 06:01.030]  So early Latin is on the top line, and then we've got the modern equivalent on the bottom.
[06:01.370 --> 06:04.050]  If you're like most people, you prefer the bottom.
[06:05.090 --> 06:10.050]  You might feel differently if you were holding a chisel in your hand and you had to carve into stone.
[06:10.050 --> 06:13.990]  Because straight lines are a lot easier to chisel than curved lines.
[06:14.890 --> 06:20.390]  So that was one of the reasons why we had so many straight lines in old languages.
[06:21.510 --> 06:31.410]  But it turns out that as we've moved away from actually carving stuff in stone and etching it into clay tablets, that we have this modern alphabet that looks different.
[06:31.410 --> 06:35.970]  And the reason is because if you can do curves, you can have more unique shapes.
[06:36.010 --> 06:41.430]  And when you have unique shapes, it's easier for your brain to quickly identify, oh, well that's a letter B.
[06:41.730 --> 06:44.890]  Think of the letter B and the number 8, you know, capital B.
[06:45.110 --> 06:47.870]  If you could only do straight lines, those are identical.
[06:48.050 --> 06:50.030]  You can't have those two things in the same alphabet.
[06:50.610 --> 07:01.490]  And so we've moved, you know, past the early Latin type stuff, and we now have a more efficient way of writing, that's been enabled by the fact that we're no longer carving stuff into stone.
[07:03.130 --> 07:05.450]  A much more modern example is clear type.
[07:05.730 --> 07:08.790]  So I want to ask, raise your hand if you know what clear type is.
[07:08.890 --> 07:09.490]  I'm really curious.
[07:10.590 --> 07:13.230]  Okay, so maybe like a quarter of people, or a fifth.
[07:13.990 --> 07:17.170]  Clear type is something that we've all seen like thousands and thousands of times.
[07:17.750 --> 07:20.050]  It's made by a little company called Microsoft.
[07:21.110 --> 07:22.450]  And it's in Microsoft Word.
[07:22.450 --> 07:27.690]  And it's a way of displaying text that makes it crisper, appear crisper on screen.
[07:27.830 --> 07:29.530]  It uses what's called subpixel rendering.
[07:32.130 --> 07:37.590]  So these are 9 pixels composed of 27 subpixels, R, G, and B.
[07:38.690 --> 07:44.530]  And normally when you render things on a screen, you just, you know, turn it on or off, depending on what color you want it to be.
[07:44.750 --> 07:51.730]  With subpixel rendering, you actually turn on, you know, the edge of a pixel so that you can make a line appear smoother.
[07:51.730 --> 07:56.010]  So like the C in clear type might appear jagged if you have too low of resolution.
[07:56.910 --> 08:00.710]  And so you've got a side-by-side here on your right.
[08:00.870 --> 08:03.810]  And on the far right is the subpixel rendered version.
[08:04.410 --> 08:11.330]  Zoom it out and you can see a little bit more easily how it's clearer than the stepwise version on the left.
[08:12.570 --> 08:15.190]  Here is what text looks like real zoomed in.
[08:15.190 --> 08:16.890]  So you can see there's a lot of color there.
[08:17.590 --> 08:20.830]  But then when you zoom it out, it just kind of looks black.
[08:22.130 --> 08:30.230]  So clear type is interesting because, you know, when we moved to LCD screens, we had this issue because things appear jagged.
[08:30.370 --> 08:32.090]  And some people think that makes it harder to read.
[08:32.210 --> 08:33.450]  Some people just think it's ugly.
[08:33.930 --> 08:40.010]  And so people came up with a solution to use subpixel rendering and color to make it easier to read.
[08:41.350 --> 08:47.650]  But in 2013, or with Word 2013, which I don't know when that actually came out, Microsoft did away with it.
[08:48.010 --> 08:51.630]  And, you know, that's probably because we now have much higher resolution screens.
[08:51.630 --> 08:58.090]  And so we don't need to use this as a tool to make text crisper because text is just crisper.
[08:59.090 --> 09:07.610]  So these are a couple of ways in which we have done a good job of making reading on screens efficient for reading on screens.
[09:08.290 --> 09:11.070]  But in a lot of ways, we have not done a good job of this.
[09:11.310 --> 09:17.290]  In a lot of ways, we basically just took the user interface for paper and said, hey, let's put that on a tablet.
[09:17.490 --> 09:20.090]  Let's, like, make a tablet a window for a piece of paper.
[09:20.090 --> 09:21.390]  And we just mimicked it.
[09:24.510 --> 09:30.130]  But if you've ever seen a news website that looked like this, you know that's not always a good idea.
[09:30.690 --> 09:31.050]  Right?
[09:31.170 --> 09:35.930]  So it's got, like, a zoomy thing, and then you're dragging around, and you can zoom out.
[09:38.030 --> 09:39.430]  But it doesn't actually...
[09:39.430 --> 09:40.290]  I tried to zoom out.
[09:40.350 --> 09:42.650]  This is the Mercury News in San Jose.
[09:43.390 --> 09:45.150]  And, like, the zoom out didn't even work.
[09:45.830 --> 09:49.550]  And what's really interesting is you actually have to pay extra for this feature, the e-edition.
[09:49.730 --> 09:51.090]  They charge for this.
[09:52.230 --> 09:59.330]  So this is one really obvious example of a way in which we need to adapt from paper to screen.
[09:59.930 --> 10:03.670]  And most newspapers don't have this anymore for their websites.
[10:03.670 --> 10:04.590]  They have a mobile version.
[10:04.790 --> 10:07.970]  But there are things that are a lot more subtle than that.
[10:09.870 --> 10:12.370]  Donald Trump is not one of the subtle things.
[10:14.470 --> 10:17.270]  So let's focus here on these first couple paragraphs.
[10:17.970 --> 10:21.070]  So this is a screenshot from a computer.
[10:21.850 --> 10:25.330]  And the first paragraph is three lines long, and the second paragraph is four lines long.
[10:26.410 --> 10:27.430]  Open it up on a phone.
[10:28.350 --> 10:29.990]  Instead of three lines, it's six.
[10:30.110 --> 10:31.290]  And instead of four lines, it's nine.
[10:32.210 --> 10:34.270]  Now you might wonder, why does that matter?
[10:34.410 --> 10:37.270]  As it's still the same words, still the same number of words, in the same order.
[10:38.110 --> 10:42.790]  To explain why that matters, we'll take a brief look at how your visual processing system works.
[10:43.330 --> 10:50.270]  We have in the center what's called a fovea, which is one and a half to two degrees of your visual field.
[10:51.090 --> 10:53.110]  Then you have the paraphovia outside of that.
[10:53.210 --> 10:54.550]  That's about eight degrees total.
[10:54.890 --> 10:57.850]  And then outside of that is your peripheral vision, which you've probably heard of.
[10:58.870 --> 11:02.150]  And not surprisingly, the fovea is where you have the most acuity.
[11:02.350 --> 11:04.070]  And then the paraphovia, less so.
[11:04.190 --> 11:06.190]  And the periphery, even less than that.
[11:06.990 --> 11:11.550]  So the reason that matters is, when you're reading on the computer screen, up top...
[11:11.550 --> 11:13.890]  Here's a paragraph, that second one with four lines.
[11:15.050 --> 11:19.790]  You have... the top and the bottom of the paragraph is a visual guide for when you're tracking back and forth.
[11:20.310 --> 11:25.490]  And the top and the bottom of the paragraph fall fairly close to your fovea when a paragraph is short.
[11:25.490 --> 11:28.390]  So as you're moving along, it's sort of like a handrail.
[11:28.470 --> 11:30.410]  When you're walking downstairs, you can like grab onto it.
[11:30.690 --> 11:36.050]  So as you're tracking along here, you can see where you're going and where you're supposed to go because you have these visual guides.
[11:36.630 --> 11:39.830]  In the bottom version, this is nine lines long.
[11:40.490 --> 11:45.850]  It's much harder for you to see the top and the bottom of the paragraph when you're reading, particularly in the middle.
[11:46.310 --> 11:55.630]  This is where you skip lines the most, in the middle, because you can't see the top or the bottom in your sufficiently acute vision, Sort of in the middle of your visual field.
[11:55.870 --> 11:57.850]  And so you're much more likely to skip lines here.
[11:58.930 --> 12:02.070]  And this is actually made worse by environmental factors.
[12:02.310 --> 12:04.430]  When you're reading on desktop, you're sitting at your computer.
[12:04.730 --> 12:06.470]  Your computer is stationary.
[12:06.730 --> 12:07.290]  You're stationary.
[12:07.730 --> 12:08.830]  There's no movement.
[12:09.090 --> 12:11.390]  Whereas when you're reading on mobile, you're mobile.
[12:11.830 --> 12:16.490]  You might be sitting on a subway and sort of the train's moving and your phone's moving and your head's moving.
[12:16.630 --> 12:25.910]  So visual tracking in general is just a lot harder, compounded by the fact that you don't have these visual cues to be able to guide you as you're reading through the lines.
[12:27.810 --> 12:31.690]  On top of that, when you're reading on a computer, you actually read differently.
[12:33.810 --> 12:36.090]  People skim more when they're reading on computers.
[12:36.530 --> 12:38.130]  And they're more likely to skip around.
[12:38.130 --> 12:41.850]  This could be due to the fact that we have ads all over the place flashing at us.
[12:41.890 --> 12:45.010]  It could be because you have notifications popping up all the time.
[12:45.430 --> 12:47.710]  But it's just a difference.
[12:47.850 --> 12:54.390]  And so we need to optimize our reading system on the computer to account for these facts and allow us to read more efficiently.
[12:56.210 --> 13:00.330]  So now we're going to talk a bit about some tools that can make reading on screen easier.
[13:00.510 --> 13:05.510]  And talk about what their claims are, how they work, and where they may fall short.
[13:06.710 --> 13:09.670]  So the first is something called rapid serial visual presentation.
[13:09.910 --> 13:11.110]  Who's heard of this, just out of curiosity?
[13:11.850 --> 13:12.090]  Alright.
[13:12.410 --> 13:16.570]  So RSVP is a technology where you flash words at people really fast.
[13:17.890 --> 13:19.970]  And you hope that you can read them.
[13:20.530 --> 13:31.130]  This was originally developed in the late 1950s, but it's had a bit of a resurgence recently because it's useful on smartphones and smartwatches where you have not very much room to display text.
[13:31.770 --> 13:33.050]  So here's one example.
[13:34.210 --> 13:36.490]  And if you look carefully, I'm going to start it.
[13:38.530 --> 13:41.810]  I was going to do hello world, but it would have been too short.
[13:43.170 --> 13:44.770]  So that was a simple sentence.
[13:45.030 --> 13:50.310]  It was 500 words per minute, which is a fast-ish reading speed for a skilled reader.
[13:50.650 --> 13:53.350]  It was pretty easy to read, to see.
[13:54.130 --> 13:57.510]  So I'll do another example again at 500 words per minute.
[14:03.110 --> 14:09.170]  Okay, that was from a scientific paper about post-concussive visual disorders.
[14:10.190 --> 14:11.310]  So that's a harder sentence.
[14:11.490 --> 14:12.710]  It was syntactically more difficult.
[14:12.930 --> 14:13.930]  It had harder words in it.
[14:14.030 --> 14:15.250]  But you probably knew all the words.
[14:16.170 --> 14:21.470]  The thing is, when you saw concussive, you probably had to think for a second, what does that word mean?
[14:21.630 --> 14:22.130]  Oh, that's right.
[14:22.130 --> 14:23.030]  It has to do with concussions.
[14:23.610 --> 14:24.950]  Okay, now what's happening?
[14:25.070 --> 14:26.090]  Okay, I just missed three words.
[14:27.130 --> 14:32.290]  So the problem is that the train keeps going, even if you're not paying attention.
[14:32.290 --> 14:45.910]  And you can pause it and you can rewind it, but researchers have found that if you count the time spent pausing and rewinding, there's actually not much of a speed benefit and possibly a speed detriment to using this compared to just normal visual reading.
[14:46.310 --> 14:51.510]  It's basically like sprinting and then stopping, versus just jogging steadily all the way along.
[14:52.970 --> 14:54.850]  So here's one last example.
[14:54.990 --> 15:00.590]  This is 800 words per minute, which is not as fast as some people claim they can read with this.
[15:00.590 --> 15:03.450]  Some people go up to like 1200 or something, but here's 800.
[15:07.340 --> 15:08.620]  That was definitely about muscles.
[15:09.260 --> 15:11.080]  There's something about muscles in there.
[15:12.360 --> 15:17.820]  And, you know, to be fair, this is on a projector, which might not be quite as good as having it on a screen right in front of you.
[15:17.920 --> 15:21.220]  But anyway, obviously comprehension drops off at some point.
[15:21.460 --> 15:26.080]  There's a bottleneck in the system and it's your ability to comprehend words.
[15:27.120 --> 15:29.700]  And so you can only go so fast with this.
[15:30.460 --> 15:34.640]  So looking at the pros and cons for RSVP, it's great for small screens.
[15:34.960 --> 15:45.300]  Some of these are unlike smart watches, which is cool because then you can have the word displayed real big and see it very easily for short messages that are simple text.
[15:45.360 --> 15:54.080]  So it's good when you don't have complex syntax or unexpected words because that causes you to have to think about it and then you lose your place.
[15:54.840 --> 15:55.880]  It's good for speed.
[15:56.200 --> 15:58.380]  You know, obviously you can get through stuff really fast.
[15:58.560 --> 16:01.420]  It's just a question of, you know, how much are you going to retain?
[16:02.960 --> 16:05.980]  Eye fatigue is one of the disadvantages of this.
[16:06.120 --> 16:11.000]  Researchers have found that people, not surprisingly, try not to blink as much when they're using it.
[16:11.180 --> 16:16.460]  And not blinking is what causes your eyes to dry out and causes eye fatigue.
[16:17.320 --> 16:19.640]  It's not great for complex texts, as we saw.
[16:20.880 --> 16:23.420]  And comprehension drops off if you try to go too fast.
[16:25.260 --> 16:28.300]  The next thing is highlighting.
[16:28.700 --> 16:32.760]  So there are these tools that I'll show you in a second where it highlights a chunk of words at a time.
[16:32.860 --> 16:38.180]  And the chunk you can pick, you start out with just a small chunk and then eventually it's like just two chunks per line.
[16:38.180 --> 16:39.780]  And then it's hopping you through the text.
[16:40.160 --> 16:45.880]  And you're supposed to look like at the center of the highlighted portion and then just follow it sort of karaoke style.
[16:46.180 --> 16:47.940]  But like karaoke on steroids.
[16:48.260 --> 16:49.580]  So here we go.
[17:04.280 --> 17:08.160]  Okay, so that was highlighting like three word chunks at a time.
[17:09.800 --> 17:11.000]  And this is interesting.
[17:11.160 --> 17:13.860]  And it has some certain benefits certainly over RSVP.
[17:14.540 --> 17:17.260]  Because, for example, context, context matters.
[17:17.640 --> 17:19.960]  The ability to see the shape of a document matters.
[17:19.960 --> 17:25.240]  So if you pull something up, before you've even read it, you say, okay, this is about whatever it's about.
[17:25.560 --> 17:29.880]  And you see there's a big paragraph at the top, another big paragraph, and then a little tiny paragraph.
[17:30.460 --> 17:36.000]  Well, before you've even read anything other than the title, you already sort of know the first paragraph is probably about the fraud charges.
[17:36.000 --> 17:37.940]  The second one is about the DOJ investigation.
[17:38.340 --> 17:40.240]  And then there's some sort of conclusion at the end probably.
[17:40.960 --> 17:46.640]  So then as you're reading the text, you know, okay, now I'm at the beginning of the fraud charges section.
[17:46.760 --> 17:47.540]  Here's how it's going.
[17:47.540 --> 17:52.720]  And then as you're getting to the end of the paragraph, you can see you're getting to the end of the paragraph and you're expecting it to wrap up.
[17:53.060 --> 17:57.560]  So this allows you to read more efficiently because you have the sort of like skeleton of what you're expecting.
[17:57.780 --> 18:05.100]  And then as you're reading it, you're attaching things, you know, the meat to that skeleton and allows you to create a sort of a mental model of what you've just read.
[18:05.800 --> 18:08.800]  If you're reading it with RSVP, you have no skeleton.
[18:08.960 --> 18:10.400]  You see one word at a time.
[18:10.560 --> 18:13.020]  And so you don't know where you're going, how far you have left.
[18:13.160 --> 18:17.440]  And it's harder to create that mental model, especially why you're trying to process all the text.
[18:17.540 --> 18:18.180]  Very, very quickly.
[18:20.580 --> 18:23.580]  There's also this thing called paraphobial pre-processing.
[18:23.700 --> 18:25.980]  So this has to do with your paraphobia, not surprisingly.
[18:27.180 --> 18:34.520]  So if you're looking at the word the, off to the side of that, off to the right, is the next word.
[18:34.640 --> 18:42.400]  And before you even start looking at the next word, your paraphobia, that like eight degree field, is already sort of pre-processing the shape of that word.
[18:42.400 --> 18:49.600]  And if it has a sufficiently unique shape, your mind is already sort of guessing what that word is before you even shift your gaze right to it.
[18:49.740 --> 18:51.360]  So this allows you to read more efficiently.
[18:52.420 --> 19:08.220]  And in contrast to RSVP, where you're seeing one word at a time, with traditional reading or with this highlighting type tool, you can actually see what's coming next, and it allows you to fill your pre-processing buffer, so to speak, and start pre-processing the word before you've gotten to it.
[19:08.260 --> 19:09.640]  So that allows you to read more efficiently.
[19:11.180 --> 19:21.840]  But there is one other thing to say about that diagram of your visual system, which is that if you try to expand the highlighted chunks too big, it's really bad for comprehension.
[19:22.200 --> 19:26.360]  In your paraphobia, you only capture about 20% of the words that you're seeing there.
[19:27.020 --> 19:29.920]  And then in your peripheral vision, you capture about 3%.
[19:29.920 --> 19:38.700]  So if you make that chunk that you're supposedly taking in, in every fraction of a second, if you make that chunk too big, it's really not going to work.
[19:39.000 --> 19:44.100]  Because you just, you simply do not process text that's too far away from your fovea.
[19:44.380 --> 19:45.860]  That's where processing happens.
[19:46.840 --> 19:50.100]  And this is sort of similar to other speed reading techniques that have been around for a long time.
[19:50.580 --> 19:57.160]  Things where you like, read down the page in some sort of unique algorithm that's supposed to let you see everything without having to look directly at it.
[19:57.740 --> 20:03.260]  It sort of works, but it sort of doesn't because your brain just simply can't see that stuff.
[20:05.720 --> 20:13.400]  So, the highlighting tool is cool because it focuses you more versus traditional reading because you're following this highlighted chunk along.
[20:14.000 --> 20:24.900]  It's great compared to RSVP because it maintains format, allows you to see the structure of the document, and then still be able to process it more quickly than you would perhaps if you were reading it manually.
[20:26.240 --> 20:31.300]  One of the downsides of both the highlighting tool and RSVP tools is they have a constant rate.
[20:31.520 --> 20:34.340]  So, you can pick how many words per minute you want it to go.
[20:34.460 --> 20:35.720]  You can pick how big the chunk is.
[20:36.200 --> 20:41.360]  But you can't say, and by the way, show me complicated sentences slower than easy sentences.
[20:41.700 --> 20:44.800]  Because the software doesn't know what's a complicated sentence.
[20:44.920 --> 20:45.580]  What's an easy sentence?
[20:45.580 --> 20:48.320]  What's a complicated sentence for you versus for me?
[20:48.760 --> 20:53.200]  You know, for one person, it could be really easy because they just read another article about that.
[20:53.320 --> 20:54.220]  But for you, it's really difficult.
[20:54.360 --> 20:57.260]  So, it's tough because everything is just at a constant rate.
[20:57.620 --> 21:00.960]  But reading is not something we do at a constant rate naturally.
[21:02.540 --> 21:08.340]  Eye fatigue, again, is an issue here because you're trying not to blink, because you're trying to follow the yellow chunk.
[21:09.380 --> 21:13.100]  And then lastly, comprehension drops off if you broaden that window out too much.
[21:13.100 --> 21:16.580]  You just end up not really being able to comprehend a lot of what's going on.
[21:18.800 --> 21:21.160]  Last step is BeelineReader, the thing that we do.
[21:22.720 --> 21:25.000]  And it uses color and looks like this.
[21:27.260 --> 21:33.040]  So, the key here is that the color at the end of a line is the same color as at the beginning of the next line.
[21:33.580 --> 21:39.340]  So that when you're bringing your focus from right to left, you're automatically drawn to the right line very quickly.
[21:40.180 --> 21:41.480]  It works in different colors.
[21:41.760 --> 21:43.160]  This is a version that doesn't have red in it.
[21:44.000 --> 21:48.220]  And we have a night mode as well, so you can do it with light on dark.
[21:49.460 --> 21:53.180]  And here's an example of how it's working with your visual system.
[21:53.180 --> 21:55.920]  So, say you're on this word push here, which is blue.
[21:56.660 --> 21:58.540]  You're moving your eyes to the left.
[21:58.840 --> 22:04.880]  And normally, even skilled readers actually, they pause as they're moving their eyes across to the left side of the screen.
[22:05.100 --> 22:11.540]  It might feel like you make one motion, but actually you stop maybe somewhere around here, and then you make another motion or two before you land in the right place.
[22:11.540 --> 22:20.540]  When you're reading with this turned on, your peripheral vision and your paraphobial vision can actually pick up the color of the text before you've arrived.
[22:20.740 --> 22:30.280]  And so it's bringing you in, you know, like an on-ramp to the right place, as opposed to you stopping here and then looking for the right place to go to.
[22:30.420 --> 22:36.660]  So it allows you to transition between lines more quickly and more easily and makes it so you don't skip lines as often.
[22:38.940 --> 22:41.780]  It's also useful when you're reading in motion.
[22:41.960 --> 22:47.220]  So for example, reading on subway, because tracking is difficult when you have motion.
[22:47.600 --> 22:52.940]  But with this, it's actually still pretty easy to read even when there is a little bit of tilt going on.
[22:54.720 --> 23:00.060]  And so it's helpful for that aspect of reading on screen and reading on mobile as well.
[23:01.520 --> 23:06.380]  So this improves tracking and helps keep people focused.
[23:07.040 --> 23:11.440]  CNET actually did an integration of this technology into a portion of their website.
[23:11.600 --> 23:14.180]  They had a long form series that they ran a couple of months ago.
[23:14.880 --> 23:18.000]  And they offered this as an option that people could turn on or off.
[23:18.200 --> 23:21.720]  And they found that people who had it turned on read 40% more.
[23:21.720 --> 23:26.320]  So they scrolled 40% further through this series than people who didn't have it turned on.
[23:26.720 --> 23:32.340]  That's because it helps keep people focused on the thing they're reading, as opposed to being distracted by other stuff that's going on.
[23:33.960 --> 23:36.320]  This also enables rereading.
[23:37.300 --> 23:41.360]  As I mentioned before, reading is not just a straight line process.
[23:41.360 --> 23:42.100]  It's iterative.
[23:42.300 --> 23:45.620]  And you go back and you look at things, especially when you're reading complex material.
[23:45.900 --> 23:49.440]  You get halfway through a sentence and it's not quite evolving the way you expected it would.
[23:49.640 --> 23:52.560]  So you have to look back and see what was the antecedent for that pronoun?
[23:52.900 --> 23:54.480]  Wait, who was the subject in this?
[23:55.160 --> 24:05.500]  And that's a natural part of reading and it's something that you can't do with a synchronous tool like RSVP or like the word highlighting that's moving along and that you have to stay in sync with.
[24:05.720 --> 24:12.460]  This is just statically printed on a page and you can just move your eyes wherever you feel like and look back very easily.
[24:14.400 --> 24:17.360]  This also results in increased reading speed for most people.
[24:17.800 --> 24:19.680]  The average reading speed increases around 20%.
[24:19.680 --> 24:25.040]  For some people, it's much higher than that, especially people who have difficulties reading and skipping lines.
[24:25.220 --> 24:27.880]  And for others, it's not as much as 20%.
[24:28.720 --> 24:30.800]  The last benefit is accessibility.
[24:32.280 --> 24:35.580]  So there are accessibility benefits related to vision.
[24:35.840 --> 24:39.200]  The American Optometric Association thought this was cool and wrote about it.
[24:39.720 --> 24:51.600]  Stanford Medical School and the Ophthalmology Department is doing research on it with relation to traumatic brain injuries, stroke patients, and Parkinson's patients who apparently all have visual problems that make it difficult for people to read.
[24:52.100 --> 24:54.800]  And Berkeley Optometry is working with us as well.
[24:55.180 --> 25:03.060]  Aside from the purely visual side of things like physical visual impairments, this is useful for people with dyslexia who often skip lines a lot.
[25:03.780 --> 25:10.360]  And for people who have attention deficits or even autism apparently causes reading difficulties that this alleviates.
[25:11.040 --> 25:17.220]  So these came as surprises to us because we're not optometrists, believe it or not.
[25:17.920 --> 25:26.480]  But as this has gotten out in the world, patients have brought it to their doctors and it's now used widely in the accessibility community and special education in particular.
[25:26.480 --> 25:29.320]  So it's very cool that it's helping folks who've had difficulty reading.
[25:31.760 --> 25:32.520]  But it's ugly.
[25:34.740 --> 25:35.760]  That's one of the downsides.
[25:36.040 --> 25:41.380]  It's hard to get product people and design people on board because they don't like the way it looks.
[25:42.040 --> 25:46.060]  It's something that is functional but doesn't have the best form.
[25:47.460 --> 25:49.280]  Another downside is it requires a color screen.
[25:49.460 --> 25:51.520]  You can do it in grayscale but it's not as effective.
[25:52.020 --> 25:56.600]  So if you have, you know, an LCD display, it's not going to be great in full sunlight.
[25:56.600 --> 25:57.900]  You won't have enough contrast.
[26:00.620 --> 26:01.500]  But, yeah.
[26:01.720 --> 26:02.840]  So I'm going to hand over to Andrew now.
[26:02.980 --> 26:06.060]  Talk about some of the challenges of building alternative reading tools.
[26:06.920 --> 26:07.140]  Thank you.
[26:09.000 --> 26:09.440]  Thanks.
[26:10.240 --> 26:11.200]  So I'm Andrew.
[26:12.060 --> 26:14.240]  Along with Nick, we started Beeline Reader.
[26:15.000 --> 26:17.520]  And I built most of the technical stack behind it.
[26:17.520 --> 26:23.440]  So just in a couple minutes, I'm going to talk about some of the challenges we face trying to actually implement this on the web.
[26:23.600 --> 26:28.280]  So we have a Chrome and a Firefox extension that you can install that apply this on any website.
[26:29.000 --> 26:32.760]  We also have an iOS app, an SDK, and a PDF rendering library.
[26:32.920 --> 26:34.920]  And they all work more or less the same way.
[26:36.240 --> 26:37.840]  So people are not identical.
[26:38.200 --> 26:42.700]  Like, people have different visual differences and different learning abilities.
[26:42.700 --> 26:46.460]  As a society, we've started to realize this.
[26:46.600 --> 26:48.260]  And we do things like personalized medicine.
[26:48.840 --> 26:53.680]  Where we use, you know, we address people's individual biology and soon genetics.
[26:54.160 --> 26:58.320]  We also acknowledge that people have different learning abilities and learning styles.
[26:59.040 --> 27:00.360]  And different types of intelligence.
[27:00.380 --> 27:02.000]  And we're using these more and more in school.
[27:02.520 --> 27:06.460]  But, on the web, we kind of expect everyone to be the same.
[27:07.060 --> 27:12.280]  The web really isn't designed, sort of, modern web browsers really aren't designed to honor user preference.
[27:13.340 --> 27:17.140]  Even though people have vision impairments and color blindness, I'm actually color blind.
[27:18.000 --> 27:19.840]  Or just different aesthetic preferences.
[27:20.200 --> 27:21.940]  You really don't have very much control.
[27:22.600 --> 27:26.440]  And that's because content creators, for the most part, have control.
[27:28.560 --> 27:32.280]  So, you know, browsers offer user style sheets.
[27:32.540 --> 27:33.920]  But, honestly, they don't work very well.
[27:34.320 --> 27:37.640]  They tend to not overwrite sites the way you would want them to.
[27:37.860 --> 27:39.200]  And very few people use them.
[27:39.200 --> 27:42.820]  There's also things like Grease Monkey, which have to be very one-off per every site.
[27:45.640 --> 27:47.360]  Sites break zoom ability.
[27:47.800 --> 27:53.160]  And, honestly, even if you want your site to zoom correctly, it's often very hard to get it right in modern browsers.
[27:53.320 --> 27:54.040]  They don't make it easy.
[27:54.240 --> 27:57.000]  And so, if you have vision issues, it can be challenging.
[27:57.180 --> 27:59.800]  You blow up the site and different parts grow at different speeds.
[27:59.980 --> 28:03.200]  Or text starts to overlap and it just doesn't grow the way you would expect it to.
[28:05.620 --> 28:07.620]  Sites break scrolling and back button.
[28:07.800 --> 28:09.520]  Things that should be squarely under user control.
[28:10.020 --> 28:11.920]  I'm not going to go into mobile apps.
[28:12.160 --> 28:17.580]  But, basically, in mobile silos, we have just no control over how text displays in other applications.
[28:17.860 --> 28:19.480]  Like, most apps don't offer extensions.
[28:19.800 --> 28:21.700]  There's no OS level extensions.
[28:22.000 --> 28:22.800]  A lot unlike iOS.
[28:23.260 --> 28:26.060]  And so, we just can't offer the service that users want.
[28:27.020 --> 28:33.480]  And this, you know, in general, I think this is causing tension where users want to control their experience.
[28:33.900 --> 28:36.340]  And content creators don't want users to control their experience.
[28:36.580 --> 28:39.320]  And there's this question about who owns the viewing experience of content.
[28:41.740 --> 28:43.500]  So, you know, this experience...
[28:43.500 --> 28:47.540]  It's been difficult for us to offer this through our Chrome and Firefox extensions on the web.
[28:50.500 --> 28:55.160]  Ironically, I think, while we expect all of our users to be the same, every website is very different.
[28:55.840 --> 28:59.060]  There's a few standards on sort of how you're expected to lay out a site.
[28:59.220 --> 29:01.000]  How we can predict where text content is.
[29:03.220 --> 29:04.800]  We can reformat a site.
[29:04.960 --> 29:06.200]  We have decent reliability.
[29:06.600 --> 29:08.160]  We can strip out the main content text.
[29:08.520 --> 29:11.700]  Remove all the Chrome, the outer layers of the page, and just show the text.
[29:11.920 --> 29:13.800]  Using something called readability.js.
[29:14.780 --> 29:18.300]  But that has real downsides for the user because they lose navigation.
[29:18.700 --> 29:20.360]  They lose the context of the page.
[29:20.480 --> 29:21.680]  Images are often removed.
[29:22.700 --> 29:26.600]  So, what we've ended up doing is not a very pretty solution.
[29:26.600 --> 29:31.320]  We've just written custom matchers for the hundred most common sites that users ask for.
[29:31.680 --> 29:37.500]  And some of these are as simple as just, you know, CSS or XPath selectors for, hey, color this region and not this region.
[29:37.700 --> 29:44.040]  And some are actually very complicated JavaScript programs that will track for live updating sites like Feedly or Facebook.
[29:44.040 --> 29:47.380]  Or we have to look, check the DOM constantly to see if it's changed.
[29:47.560 --> 29:50.240]  Because browsers do not make it easy to know if something's changed.
[29:52.400 --> 29:56.420]  I mentioned that we use readability.js on any site that we don't have a custom matcher for.
[29:57.600 --> 30:00.600]  So, here's an example of the New York Times blown up a little bit.
[30:00.740 --> 30:03.960]  Where we do have a custom matcher and you get your navigation ability.
[30:04.440 --> 30:06.040]  We can leave images in place.
[30:06.280 --> 30:08.040]  And it's a good experience for the user.
[30:08.460 --> 30:11.900]  And here's what it looks like if we have to fall back to readability.js.
[30:11.900 --> 30:13.980]  Like, you still have a great reading experience.
[30:14.160 --> 30:16.340]  But you can't navigate to other articles.
[30:16.340 --> 30:18.540]  You can't see the rest of the page.
[30:20.140 --> 30:25.400]  One thing we're talking about that we haven't done yet, but I think makes a lot of sense, is to actually crowdsource the solution.
[30:25.600 --> 30:28.600]  We have tens of thousands of installs of our browser plugins.
[30:28.780 --> 30:33.540]  And we could ask our users to select regions of the page that they want to have coloring on or not have coloring on.
[30:33.700 --> 30:35.520]  And then we can share that with the user base.
[30:36.040 --> 30:38.620]  So, that's not out yet, but something we're thinking about.
[30:40.260 --> 30:41.980]  So, really briefly about the algorithm.
[30:43.820 --> 30:46.580]  Another challenge on the web is that text can be absolutely positioned.
[30:46.740 --> 30:48.360]  Even characters can be absolutely positioned.
[30:48.540 --> 30:52.020]  And so, there's no guarantee that the DOM order of the page is the same as the reading order.
[30:52.640 --> 30:54.760]  So, we need to infer lines from position.
[30:56.280 --> 31:04.620]  So, we scan across the DOM with JavaScript inferring lines based on increase of X or drop of Y of character positions.
[31:04.620 --> 31:09.360]  And if we see a large enough change, we build a line and then color it in a second pass.
[31:10.140 --> 31:17.060]  This is also how we handled our PDF coloring, which is similarly, you can just put stuff anywhere, and there's no guarantee about how characters are positioned.
[31:18.040 --> 31:21.880]  We have to do some interesting things for right-to-left languages, like Hebrew or Arabic.
[31:22.020 --> 31:24.940]  We have to scan backwards, because lines build in the opposite order.
[31:25.920 --> 31:41.200]  And because we're applying this processing in JavaScript, and JavaScript's not a particularly fast language, we have to do frustratingly annoying things with deferring the coloring and batching it in the DOM and doing it asynchronously to try to get things to color efficiently.
[31:41.440 --> 31:44.780]  And it still often has a little bit of lag, but it comes in pretty quickly now.
[31:48.400 --> 31:49.700]  And actually, back to Nick.
[31:52.380 --> 31:55.820]  So, as Andrew mentioned briefly, we've got several different tools.
[31:56.240 --> 31:59.220]  The most popular are the Chrome and Firefox extensions.
[31:59.680 --> 32:03.180]  We also have an app in the iOS App Store and a PDF plug-in.
[32:03.380 --> 32:08.700]  Our Android tool is in beta, and so if you want to join the beta, email us.
[32:08.820 --> 32:09.260]  Beta, please.
[32:11.120 --> 32:18.620]  And then we also have a JavaScript that powers these tools, and we provide that as well to folks so they can just build it into a site.
[32:18.780 --> 32:24.060]  So, this is a website for people who have to read complex scientific articles.
[32:24.500 --> 32:26.040]  It's meant for PhD students.
[32:26.640 --> 32:33.400]  And they asked if they could integrate Beeline Reader, and so they have this thing up at the top, this little Beeline thing.
[32:34.080 --> 32:44.540]  And when you click on it, it brings down a menu, and this basically just ties into our JavaScript, and the user can select different color schemes, and it applies the color to the text.
[32:45.440 --> 32:47.400]  This is really the future for us.
[32:47.660 --> 32:55.000]  We are doing these sort of integrations with content platforms and even just websites that want to put this on there.
[32:55.520 --> 33:09.320]  And that's the best way to get it out to people, because it's nice to have plug-ins for various different browsers and stuff, but so much happens on mobile these days, and especially with iOS being closed off to doing auto-running extensions.
[33:09.500 --> 33:10.600]  There's no way to do that.
[33:10.720 --> 33:12.920]  So, if you're in Safari, it's just like automatically happening.
[33:14.740 --> 33:27.360]  So, if a website has the little button, someone can just click it and turn it on no matter what platform they're on, as opposed to having people have to go out and get every single plug-in that we make, or every single app that we make, and trying to browse it through that.
[33:28.080 --> 33:31.200]  So, look forward to answering any questions people have.
[33:31.420 --> 33:34.700]  If you want to get in touch with us, developer at BeelineReader.com, hits both of our emails.
[33:35.160 --> 33:35.940]  So, thanks a lot.
[33:49.150 --> 33:50.190]  I'll just speak loud.
[33:51.090 --> 33:55.170]  What about finding the thing which you read before in a paragraph?
[33:56.350 --> 34:05.130]  Maybe it's a personal experience, but reading is super great, but let's say if I'm looking for a word which I read before, and I'm trying to find amongst the text, the red is very distracting.
[34:05.430 --> 34:06.590]  Like, it's hard to find.
[34:06.590 --> 34:09.190]  Yeah, so red is distracting is something we hear.
[34:09.430 --> 34:12.130]  I think it's, you know, for some people it's based on schooling.
[34:12.290 --> 34:16.590]  They're like, I had a teacher who marked up my papers in red, and I'm now, like, traumatized by the color red.
[34:16.670 --> 34:21.290]  So, that's why we allow people to customize the colors and change it.
[34:21.350 --> 34:22.730]  Other people say, like, the opposite.
[34:22.950 --> 34:25.430]  I don't like blue because it reminds me of hyperlinks.
[34:25.970 --> 34:33.770]  And so, we let people, you know, avoid colors that don't work for them, because sometimes it makes it stand out in the wrong way.
[34:33.990 --> 34:34.110]  Yeah.
[34:35.050 --> 34:37.350]  Speaking of hyperlinks, how do you display that?
[34:38.150 --> 34:38.590]  Underlining.
[34:39.030 --> 34:39.410]  Yeah.
[34:39.610 --> 34:40.490]  It's an interesting issue.
[34:40.690 --> 34:44.550]  And we don't currently have a way to distinguish between, like, visited and non-visited.
[34:45.230 --> 34:45.910]  But, yeah.
[34:46.110 --> 34:46.950]  It's interesting.
[34:47.090 --> 34:49.190]  Because previously we used color to denote hyperlink.
[34:49.990 --> 34:50.410]  Yeah.
[34:51.830 --> 34:52.610]  We have thought about that.
[34:52.870 --> 34:53.530]  Haven't done it.
[34:53.610 --> 34:53.870]  But, yeah.
[34:54.030 --> 34:55.470]  That is a perfect idea.
[34:55.470 --> 34:55.530]  Yeah.
[34:56.310 --> 34:56.390]  Yeah.
[34:57.570 --> 34:59.970]  So, I noticed you have three colors.
[35:00.350 --> 35:02.330]  Is there a reason for that?
[35:02.550 --> 35:02.970]  Or...
[35:02.970 --> 35:03.590]  There is.
[35:03.790 --> 35:06.130]  So, the original idea was just two.
[35:06.610 --> 35:08.050]  It was actually blue and black.
[35:08.270 --> 35:10.670]  And so, like, I actually hard-coded...
[35:10.670 --> 35:14.770]  I mean, I colored a Word document one time, like, the first time I did this.
[35:15.330 --> 35:18.610]  And just put it in blue to black, black to blue, blue to black, black to blue.
[35:18.610 --> 35:21.310]  So, it turned out that solved the line transition problem.
[35:21.470 --> 35:25.950]  But it created a different problem, which was that when you're like 50% through the line, you tend to go like that.
[35:26.370 --> 35:27.390]  And jog down.
[35:27.730 --> 35:29.290]  And so, that obviously is bad.
[35:29.530 --> 35:31.410]  So, I actually introduced...
[35:31.410 --> 35:33.850]  You could see it as one additional color or two.
[35:35.290 --> 35:35.690]  It's...
[35:35.690 --> 35:36.550]  They're both black.
[35:37.350 --> 35:40.550]  You know, two of the colors are black and then two of the colors are non-black.
[35:41.310 --> 35:47.110]  So, the reason for that is that when you only have one non-black color, it creates this jog in the middle problem.
[35:47.770 --> 35:50.090]  And some people have asked for, like, can I have 13 colors?
[35:50.370 --> 35:52.150]  Or, you know, like arbitrary numbers of colors.
[35:52.290 --> 35:53.390]  We don't support that right now.
[35:53.530 --> 35:54.530]  Most people don't need it.
[35:54.730 --> 35:59.590]  But it is helpful to have two non-black colors to make it so that you can track through the middle.
[36:02.270 --> 36:05.190]  Have you played with subtle or bold colors?
[36:06.050 --> 36:07.830]  That is something we've thought about.
[36:08.150 --> 36:10.070]  It's hard to do gradations in that.
[36:10.270 --> 36:11.170]  It is a difficulty.
[36:11.610 --> 36:18.210]  It would be cool if we could, like, build our own font that, like, had gradations of boldness, of size.
[36:18.470 --> 36:18.970]  Like, you could...
[36:18.970 --> 36:20.150]  You can do things like this.
[36:20.310 --> 36:22.770]  But it seems like color works well.
[36:23.210 --> 36:25.170]  And the research shows that...
[36:25.170 --> 36:26.090]  So, like, there was...
[36:26.090 --> 36:28.310]  We had no research background on any of this.
[36:28.370 --> 36:30.050]  It was just like, hey, wouldn't it be cool if...
[36:30.050 --> 36:30.990]  This seems like it works.
[36:31.030 --> 36:31.550]  Yeah, let's do it.
[36:31.550 --> 36:38.390]  And then since then, we've talked to a bunch of researchers and found out that, like, it turns out when you look at something, the color of the thing is the first thing your brain processes.
[36:39.370 --> 36:41.710]  So, like, you look at the back wall and you say, like, it's brown.
[36:42.570 --> 36:47.350]  And so, if someone says, look for the red thing in the room, like, you can very easily find...
[36:47.350 --> 36:48.570]  It's the fire extinguisher case.
[36:49.470 --> 36:53.630]  So, it turns out color actually is very efficient from a cognitive perspective.
[36:53.630 --> 36:56.010]  And then it's also sort of the easiest thing to do continuously.
[36:56.110 --> 36:57.710]  So, that's what we do for now.
[36:58.010 --> 36:58.110]  Yeah.
[37:00.570 --> 37:04.910]  How well does this work with screen overlays, like f.locks?
[37:06.010 --> 37:06.090]  Yeah.
[37:06.510 --> 37:07.310]  Good question.
[37:07.590 --> 37:08.510]  So, I use that.
[37:08.810 --> 37:10.050]  And it works.
[37:10.050 --> 37:19.330]  You may have to tweak your colors, your beeline colors, so that at night, you know, when the blue gets kicked out, basically, so that it works.
[37:19.330 --> 37:24.830]  I actually use the night color scheme that I showed earlier that's black background with blue and purple text.
[37:24.830 --> 37:30.190]  And that works fine with flux and with night shift on the, or whatever they call it on the, on iOS.
[37:30.710 --> 37:31.950]  So, it does work okay.
[37:33.050 --> 37:34.290]  But yeah, great question.
[37:34.410 --> 37:40.710]  It's something that we would love to, like, work with them to actually make it so they don't kick our colors around too much.
[37:43.490 --> 37:46.890]  Have you come across anything that's also sub-vocalization issues?
[37:48.250 --> 37:56.310]  So, sub-vocalization is, just for everyone's benefit, that's when you're reading stuff that you hear it sort of in your head as you're reading.
[37:56.810 --> 38:02.850]  And, and some people believe that by eliminating sub-vocalization, you can read more quickly and with the same comprehension.
[38:03.250 --> 38:07.590]  Researchers tend to think that if you eliminate sub-vocalization, you're killing comprehension.
[38:07.590 --> 38:20.690]  So, we're kind of agnostic on the question, but it does match with a sort of bottleneck question of, if you increase the reading speed on an RSVP tool too much, can you no longer comprehend things?
[38:21.570 --> 38:31.310]  So, the research we've seen tends to indicate that eliminating sub-vocalization allows you to see words more quickly, but not necessarily read them, if that makes sense.
[38:31.310 --> 38:34.010]  It's sort of like skimming versus close reading.
[38:35.790 --> 38:42.970]  Have you explored any ideas of changing your text, like going back to pictogram or maybe getting rid of vowels?
[38:43.230 --> 38:43.990]  Have you heard of emoji?
[38:45.990 --> 38:46.510]  Yeah.
[38:47.130 --> 38:47.330]  Yeah.
[38:47.330 --> 38:48.930]  Yeah, actually, it's interesting.
[38:49.230 --> 38:54.950]  And, and what's really interesting is, you know, I mentioned before, it takes a long time to become literate in a pictographic culture.
[38:55.830 --> 39:05.370]  Now with emoji, like, I never use emoji, and I feel like this daunting sense of, like, kids these days are going to come up with it, and they're going to be able to do it really well.
[39:05.450 --> 39:08.210]  And I don't even know, like, what's out, what's the universe of emoji?
[39:08.910 --> 39:15.250]  And, and now there's tools, like, when you're doing stuff, I think on iOS and other tools, you can actually search, like, I want an emoji for whatever.
[39:15.350 --> 39:18.350]  And I think iOS 10, like, suggests emoji based on what you're typing.
[39:18.630 --> 39:27.230]  So it's actually interesting how we've, like, both reverted back to this pre-alphabetic culture, and then at the same time, we've done it in a better way, so you can, like, search and find things.
[39:28.190 --> 39:34.230]  But, yeah, it, it can be very efficient, because you can express, like, a concept or an emotion more so than just writing the word.
[39:34.470 --> 39:42.510]  But it creates issues of, like, remembering whether there's an emoji for, like, pizza with jalapenos on top, or, like, you know, what exactly it is you're looking for.
[39:42.710 --> 39:45.110]  Or if you have to do, like, a pizza and then a jalapeno.
[39:46.810 --> 39:47.170]  Yeah.
[39:47.730 --> 39:54.870]  How about, like, if you get rid of the wavos, for example, all those words still readable, but you need to read less text?
[39:55.750 --> 39:56.730]  Yeah, sort of.
[39:56.830 --> 39:59.350]  But we actually recognize words largely based on shape.
[39:59.590 --> 40:01.350]  So if you eliminate the vowels...
[40:01.350 --> 40:09.170]  For someone who's skilled in reading words that have vowels in them, like, you're eliminating E's and I's and A's and O's and...
[40:09.170 --> 40:10.650]  Like, basically, all vowels are short.
[40:11.090 --> 40:18.350]  And so if you think of, like, the skyline of New York City, if you, like, eliminated all the short buildings and everything, which is tall buildings, it just, everything looks like a chunk.
[40:18.510 --> 40:21.630]  Which is why it's actually harder to read stuff that's in all caps.
[40:23.110 --> 40:25.630]  Because the word shape is just completely obliterated.
[40:25.750 --> 40:27.170]  Everything is just, like, a rectangle.
[40:28.330 --> 40:29.330]  Is there one behind you?
[40:29.330 --> 40:29.850]  Yeah.
[40:29.850 --> 40:33.150]  So you're totally focused on the web, it seems like.
[40:33.370 --> 40:37.610]  And so have you thought about doing system-wide stuff, either for desktop or for rooted or jails?
[40:38.110 --> 40:38.550]  Yeah.
[40:38.670 --> 40:40.910]  So the question was about system-wide as opposed to just web.
[40:41.070 --> 40:43.330]  I mean, Andrew, you can talk a little bit about the challenges.
[40:44.650 --> 40:45.010]  Yeah.
[40:45.150 --> 40:46.470]  So we haven't dug into it deeply.
[40:46.470 --> 40:47.670]  We started on the web.
[40:48.850 --> 40:53.030]  And you're totally right that for jailbroken devices, we probably could do it in iOS.
[40:53.030 --> 40:55.990]  And it's easier to jailbreak on Android.
[40:56.190 --> 40:57.110]  And so we probably could there.
[40:57.290 --> 40:58.910]  We haven't really gone that route yet.
[41:00.510 --> 41:13.750]  Although I'd note, even on Android, from folks I've talked to who worked on Android before, they said that there's, like, the generic text views, which maybe you could replace, but then a lot of apps actually end up using their own text views, which then we wouldn't be able to replace.
[41:14.010 --> 41:15.450]  So, yeah.
[41:16.050 --> 41:18.350]  It's actually more difficult than one might hope.
[41:19.250 --> 41:20.130]  Was there one over here?
[41:20.270 --> 41:20.510]  Yeah.
[41:20.510 --> 41:28.580]  I have a question to say, I don't know if that was discussed at all, but one, is it true that reading in the dark is bad for your eyes?
[41:29.040 --> 41:37.200]  Two, is it true that reading itself is true that reading off a computer screen is worse for your eyes than reading out of the book?
[41:38.060 --> 41:40.840]  Okay, so number one, I have no knowledge of.
[41:41.100 --> 41:44.140]  I don't know if it's worse to read in the dark.
[41:44.300 --> 41:45.340]  My mom told me not to.
[41:47.980 --> 41:50.240]  Yeah, as far as I know, there's no data that supports it.
[41:50.320 --> 41:52.200]  Number two was about reading...
[41:52.200 --> 41:54.840]  Is reading back your eyes because we're not hardwired?
[41:55.200 --> 41:56.520]  Yeah, so that's a good question.
[41:56.720 --> 42:04.380]  I mean, eye doctors definitely say after reading for 20 minutes, you should take 20 seconds and look 200 feet away.
[42:04.920 --> 42:07.740]  There's some easy-to-remember thing that I can't remember.
[42:08.680 --> 42:12.320]  But basically, yeah, we're not meant to be like visually focused like right here.
[42:12.320 --> 42:17.540]  And when you are visually focused here, your eyes have to do something different than when you're focused far away.
[42:17.640 --> 42:20.900]  When you're focused far away, your eyes are essentially parallel, 100%.
[42:21.320 --> 42:22.960]  They're looking straight off into the distance.
[42:23.040 --> 42:29.220]  And your brain can put together those two images stereoscopically because you're just looking straight 100% ahead.
[42:29.620 --> 42:33.140]  When you're looking at something that's close to you, your eyes are doing what's called converging.
[42:33.260 --> 42:35.600]  So one of them looks that way and the other looks that way.
[42:36.140 --> 42:41.280]  And some people have issues with this, either over-convergence or under-convergence, which both lead to double vision.
[42:42.020 --> 42:45.580]  So, yeah, it's reasonable to think that since we weren't meant to...
[42:45.580 --> 42:53.480]  I mean, it's not like we did other things for super long periods of time, like knitting or, you know, I don't know what we did back then, that would have been so close up.
[42:53.600 --> 42:59.580]  But you would have been looking around naturally, so it is reasonable to think that that would be not great for eyes.
[42:59.800 --> 43:03.200]  And, like, based on my experience, I had to get glasses when I was in grad school.
[43:03.460 --> 43:04.200]  Like, I was reading a lot.
[43:04.320 --> 43:06.120]  Like, probably that was not unrelated.
[43:07.080 --> 43:08.220]  Was there one other question?
[43:08.720 --> 43:11.440]  Oh, I'll just say if there's any evidence in reading off screen.
[43:11.620 --> 43:11.840]  Oh.
[43:13.280 --> 43:18.640]  Yeah, so to the extent that people often are badly positioned when reading off screens.
[43:18.900 --> 43:22.620]  Like, when you're reading a book, like, you are holding the book, and you put it wherever you want it.
[43:22.880 --> 43:31.340]  A lot of people don't have screens in the right position, and they're, like, sitting on a chair that's not high enough, and you end up, like, looking up too much, not blinking enough.
[43:31.480 --> 43:32.460]  And so that can be bad.
[43:32.600 --> 43:34.960]  I don't think there's evidence that it's intrinsically bad.
[43:34.960 --> 43:45.960]  Like, you couldn't have a good experience reading off a screen, either LCD or e-ink, but I think positionally, that's the problem, is just basically, we're not set up right.
[43:48.200 --> 43:55.840]  So, I read now on my e-ink reader, and I forward all my stuff to that, could I use it there as well?
[43:56.120 --> 43:57.040]  There are no colors.
[43:57.480 --> 43:59.440]  Yeah, so it can be done in grayscale.
[43:59.620 --> 44:04.140]  We do not have anything that works with e-ink, although I was talking with somebody today.
[44:04.140 --> 44:11.100]  So, we have a booth here, and somebody came by and was like, I'm creating an alternate OS for e-ink readers.
[44:11.680 --> 44:14.000]  And I was like, yeah, let's do something.
[44:14.400 --> 44:18.040]  Because that's exciting for us, because I know people really do like the e-ink display.
[44:18.460 --> 44:27.700]  One of the reasons that color e-ink has not caught on is because you can't use it for photographs or images, because it's not like high contrast enough, but it would totally work for this.
[44:27.880 --> 44:30.160]  So, we're hoping that we can get something on e-ink.
[44:30.520 --> 44:34.660]  We're trying to talk to Amazon, but not surprisingly, they're Amazon.
[44:35.880 --> 44:36.440]  Yeah.
[44:37.200 --> 44:38.080]  So, in the back.
[44:55.980 --> 45:00.320]  So, you mentioned color coding for code.
[45:00.500 --> 45:01.180]  That's different.
[45:01.320 --> 45:01.400]  Yeah.
[45:02.720 --> 45:03.700]  Yeah, yeah.
[45:06.140 --> 45:08.600]  So, we don't have a terminal-based tool.
[45:09.020 --> 45:13.720]  Our browser plug-in for Chrome and Firefox is compatible with Google Docs, which a lot of people use.
[45:14.080 --> 45:22.660]  We have a pasteboard on our website that you can paste anything into, but we don't have anything that's terminal-based.
[45:22.660 --> 45:23.820]  I haven't really tried it with that.
[45:23.940 --> 45:28.420]  I was just going to say, if you simply open the text document in your browser, would it render it?
[45:29.380 --> 45:33.320]  Yeah, you can use the readability thing, and it'll render it.
[45:33.560 --> 45:33.580]  Yeah.
[45:34.980 --> 45:35.580]  Way in the back?
[45:38.800 --> 45:39.300]  What's that?
[45:39.860 --> 45:40.460]  What's your story?
[45:42.280 --> 45:43.560]  A long time ago.
[45:44.260 --> 45:53.880]  I was in college, and I was talking with my RA about a psych class he was taking, and it was about color and reading and how your brain works with that.
[45:54.060 --> 46:02.100]  And basically, psychologists like to do things to their test subjects where they have them doing a task, and then they make them bad at it by throwing in things.
[46:02.360 --> 46:09.880]  So, researchers found out a long time ago how you can make people bad at saying the words of colors by printing the word orange and purple.
[46:10.520 --> 46:15.840]  And so, they came to this conclusion that, ha-ha, you know, we've screwed people up at reading by using color.
[46:16.180 --> 46:17.820]  And I thought, you know, that's great.
[46:18.100 --> 46:23.740]  But being slightly less sadistic than that, I thought maybe there'd be a way you could actually make people better at doing something as opposed to worse.
[46:23.980 --> 46:36.860]  And so, the idea was take this channel, the color channel, and use it to flag something useful other than highlighting, which is what we've obviously done before, but use it to flag, like, here's where your eyes should go next.
[46:37.080 --> 46:38.520]  So, that's sort of the origin story.
[46:38.700 --> 46:44.120]  And then, I worked on it with Andrew, who is my cousin, and who was at a college very nearby.
[46:44.360 --> 46:45.980]  So, that's sort of where it started.
[46:48.440 --> 46:51.760]  So, it sounds like color text processing at the moment is automated.
[46:52.140 --> 46:57.800]  Do you have any tools for your content creators to determine what should be colored in their web pages?
[47:00.340 --> 47:12.620]  So, if people use the JavaScript SDK, the library we can give you, then you can select whatever regions you want in the DOM and color them, either by putting a class on them or just navigating to them and calling the function on them directly.
[47:13.200 --> 47:13.600]  Okay.
[47:15.180 --> 47:17.000]  And the JavaScript just handles everything.
[47:17.180 --> 47:19.980]  Like, if the user resizes the window, then it'll go in and recolor.
[47:20.080 --> 47:25.400]  So, basically, all you need to do is say, like, here's the text, and then the user selects the colors, and that's all there is.
[47:25.760 --> 47:25.760]  Yeah.
[47:26.340 --> 47:26.760]  In the back.
[47:28.100 --> 47:30.880]  You said something about a CSS class.
[47:31.320 --> 47:38.760]  So, if I have E-line and Stylish, can I write my own custom website stuff by just having it stick with my class on?
[47:39.820 --> 47:41.240]  I haven't used Stylish.
[47:41.340 --> 47:41.940]  What is Stylish?
[47:42.900 --> 47:45.840]  Basically, just let me write my own CSS for websites.
[47:45.840 --> 47:48.500]  Sort of like Grease Monkey, but for user styles?
[47:48.940 --> 47:49.160]  Yeah.
[47:49.440 --> 47:49.520]  Okay.
[47:50.160 --> 47:55.940]  You probably could get them to work together, because you certainly could tell Stylish to ignore B-line styles, or vice versa.
[47:56.160 --> 47:56.500]  You could...
[47:57.760 --> 48:05.440]  There's both an attribute and a class you can put on content that will tell B-line either to skip it or to color it.
[48:05.920 --> 48:06.960]  So, maybe?
[48:07.860 --> 48:08.800]  Talk to me afterwards.
[48:09.580 --> 48:09.600]  Yeah.
[48:09.820 --> 48:11.180]  Maybe I'll try it if you know you.
[48:11.280 --> 48:11.480]  Yeah.
[48:12.680 --> 48:13.120]  Yeah.
[48:13.400 --> 48:15.420]  When will you be appearing on Shark Tank?
[48:16.120 --> 48:20.780]  You know, actually, two of the people who are associated with Shark Tank have dyslexia, I'm told.
[48:21.780 --> 48:23.740]  And, yeah, thought about it.
[48:23.900 --> 48:24.500]  It would be cool.
[48:24.740 --> 48:32.600]  I actually emailed somebody I know who was a friend who was on Shark Tank, and I'm trying to get info on how to successfully get on.
[48:33.780 --> 48:38.140]  But, yeah, it's super useful for some dyslexic people, so fingers crossed.
[48:39.240 --> 48:39.660]  Yeah.
[48:40.140 --> 48:41.000]  Any other questions?
[48:41.000 --> 48:41.860]  Uh,
[48:45.640 --> 48:52.680]  this is people, like, educators that are not super familiar or comfortable with technology, but especially if they were...
[48:52.680 --> 48:53.640]  Repeat the question.
[48:54.140 --> 48:57.820]  Yeah, so asking about, um, do we have videos for use by educators?
[48:57.980 --> 48:58.840]  And absolutely we do.
[48:59.200 --> 49:03.400]  Um, we work with a lot of educators, um, K-12 and higher ed.
[49:03.540 --> 49:11.400]  Um, and we have, like, video screencast things that show, like, go to our website, click this button, it will install a browser plugin.
[49:11.580 --> 49:18.360]  Go to this website, it's, like, set up to work with this educational content, and it's, like, 45 seconds long and shows exactly how to do it.
[49:18.780 --> 49:18.920]  Yeah.
[49:20.360 --> 49:21.040]  Anything else?
[49:22.480 --> 49:22.880]  Cool.
[49:23.140 --> 49:23.820]  We'll be around.
[49:23.940 --> 49:27.060]  We've got a booth right by the front, um, and, uh, would love to talk to people.
[49:27.200 --> 49:27.700]  Thanks so much.
[49:27.700 --> 49:27.720]  Thank you.