Subscribe
Share
Share
Embed
Each week Adam & Ste focus on a specific feature, idea or prototype in Hardcover and iterate on it together or with guests.
adam (00:01.08)
hey hey how's it going
ste (00:03.909)
it's good we got to be here on her cover video number six
adam (00:13.1)
how has a how's your last week been
ste (00:16.669)
it's been pretty good it's been quite cold in london but it was okay inside feels good how about you
adam (00:27.48)
pretty good we had about three inches of snow last night so it's always it's always a lot of fun to just wake up and see like all the mountains covered in snow like as soon as as soon as that time of the year ends i always kind of miss it a little bit even if even if it does mean i'm able to actually go outside to the mountains and not worry about slipping every time
ste (00:37.649)
uh huh
ste (00:48.149)
that sounds very cool i mean i wish there was more snow here but yeah i mean
adam (00:55.undefined)
m hm
yeah we we i think we're like going out to walk around a dog park which is like one of the few areas that like you know i feel pretty safe walking around with my knee and there's there's a dog park here which they have to like they send in an alert people have to have their dogs on leash there now because a whole herd of elk moved into the dog park
ste (01:25.529)
what's okay so it's another part now wow
adam (01:28.42)
yeah it's an elk park it's like it's kind of in a ravine right next to the inner state so it's like down in a valley and so there's like trees and like foliage and that whole area was like the dog park with like um like fences along the sides on the top of the ravine so it was it was always a fun place to bring dogs but now yeah like a dozen or so elk have moved in
ste (01:37.209)
okay
ste (01:54.569)
wow that sounds pretty amazing and yeah a bit you know i'm guessing you'd have to like mind the elk now are they like did they get angry
adam (02:04.52)
yeah
adam (02:08.22)
i think they try to stay as far away from the dogs as possible
ste (02:11.269)
okay okay any theories as to what brought them there or is there like a treats someone left around or just like random ness
adam (02:21.1)
yeah probably just like the mountains at this time of year are just like covering in snow there's no real foliage you know we're getting towards the end of winter the foliage is kind of winding down that they've been eating through all winter so they were probably just trying to find new new green stuff to eat
ste (02:40.409)
well that's amazing i mean elk invasion in the dog park sounds like really interesting event happening yeah i'd love to see the all send me some picks of the elk i mean if you can get like close enough
adam (02:44.2)
uh uh
adam (02:57.68)
yeah if we if we spot some from a safe distance i'll make sure to take some pictures
ste (03:02.909)
yeah that sounds great well yeah i haven't had any elk elk stories so is for me i guess
adam (03:15.34)
yeah
well aside from aside from that what have you been up to on hard cover last last week
ste (03:23.689)
well i was thinking we could go through some hard covering news because we've been having some really cool updates lately for my end i've been building our library which will help with the future improvements for the site that we're planning and i'm hoping that would be like really good
m
asset to have once we get the apse out there because the sap is out there i know if we had actually we said that or did we in okay okay so one week the sap has been in the store for one week and we're getting the android ap ready which will happen in the coming weeks so i'm guessing working around
adam (04:06.68)
i think we said that last week
ste (04:26.509)
h was really interesting and we've also got very cool development that of course you might happen and followers on our disquard saw it for in the realm of having book recommendations made by a i which is something we're testing now
and it's showing so far really interesting results i think i've been trying it and i'm curious to see if other people try it as well because i think it works pretty neat for cases when you want to have book recommendations you know and a certain area that stuff or research i guess
for me it's been rendering some really cool results now how about you
adam (05:32.48)
yeah so yeah the experience we're talking about is the ask jewels experience on hard cover which is effectively using the open a chat a p that they released last week on march first and even before that experience was released stay had already done some mock ups and figma on like what an a i bought would look like with like how that experience might work and then like
like a week after he made those prototypes open the released this chat but a p i i'm like well this timing is too good and so on yeah last thursday i just like spent four hours and threw together i put a type of it um and share my screen as we talk about this
ste (06:20.949)
yeah let's go through it do some actual life testing here we go ask jewels
adam (06:26.5)
yeah yeah so yeah behind the scenes effectively what we're doing is like at this point like nothing has hit at g p t we're just like this is all hard coated at this point but then as soon as you send any question any uh it could be it could be a question it could be like here's some of my tastes whatever you want to say at this point is sending it to check
ste (06:29.349)
life
adam (06:56.36)
g p t a p i and it's including some some scripts on the back and along with it to let chatcpt know like who jewels is what the personality that you should it should be answering in is and i think like that even that like a couple of sentences of description kind of gives it a little more personality than just like a flat personality answering questions
um yeah like my my wife and i were just having a conversation about like what are like what are some good science fiction books that featured like really good romances so i was actually kind of curious to ask that in here
ste (07:42.849)
yeah we go let's see
adam (07:49.5)
m m
adam (07:54.04)
o s the time machine done
yeah so i feel like some of these are the ones that we we were talking about like we were alking about time travelers wife that's the one i'm currently reading as you can see
ste (08:08.929)
nice okay
adam (08:12.26)
but i feel like there's a lot of work that can be done here because i don't i don't know if i consider hungry games to be like a great romance i feel like it's it's like a a light romance but i feel lik there's a world of difference between like a an outlander like a time traveller's wife romance and like i guess i would be like a coming of age romance i guess tha would be the difference a coming of age romance versus like a like a relationship later in life
ste (08:36.668)
okay yeah yeah that's a good distinction because yeah yeah that's really good i mean there's plenty of romance in the hunger games but it is a coming of agromance more than anything else i guess
adam (08:53.28)
yeah what if i say um ones that aren't coming of age romances
ste (09:00.489)
yeah let's see
it's all in the question with this i mean asked for some lesser known books and it's given me some uh here we go
adam (09:16.76)
i keep seeing this one recommended
adam (09:21.66)
i mean i should move that one up
ste (09:21.909)
well the long way to a small angry planet okay seems like it has some potential romance in there just by the title
adam (09:30.4)
yeah and i still haven't read this ursula laquinbook like it keeps coming up
ste (09:34.969)
yeah yeah it's been quite popular ursula on he heard cover i've seen some of the of her books in the feet which is like really really interesting yeah it's the results i guess are debatable asked for some lesser known book
adam (09:37.56)
and the host
adam (09:51.26)
yeah
ste (10:04.849)
from lesser known authors and it gave me some books which of course threw some like really hard core readers or let's not say how but people reading a genre might seem okay this has i've seen this ol over the place but for instance from for someone who was outside you know certain gen wasn't like into that that niche specifically
i think could have been valid so i guess it comes down to how it interprets you know things like less known or romance for that for that matter i'm guessing it would give the popular books first so i'm wondering what prompt for bringing up books which are more obscure with sound like
adam (11:02.26)
m yeah
ste (11:02.829)
maybe actually like writing more i know if obscure is a good word what's wonderful about this is that you can put like really anything in this i've tried like really complex pumps like non fiction books written by women authors who that involved nature or you know any kind
termanit's given me some really like i mean i couldn't have found those results anywhere else because you know it's not something that would be reading in a block post so that was pretty pretty nice well about this i mean you tell me adam your apparently some of these books are ready in your one to read list
adam (11:52.7)
yeah
adam (12:00.undefined)
yeah i think like the city we became is one that i've wanted to add to my wont to read because it's by and kagent jemison yeah these and this this one spacebuteen where worlds i've i've definitely heard about
ste (12:07.869)
oh yeah
adam (12:17.4)
so maybe i'll this one i don't even i don't think i've heard about it's just an ancient society of which is an a hipstertechnological start up go to war as the world is tearing itself apart okay yeah that sounds funny yeah
ste (12:23.489)
wow
ste (12:32.869)
yeah that sounds really fun look at that wow okay that's actually pretty bales
adam (12:37.64)
yeah
adam (12:41.04)
yeah this one this one seems more a war and soldier so might skip that one yeah this a the way that this ends up working behind the scenes is that we have a rule that tells that g p t to return the book but then also return the i s b n in a form at like isbn colin and then like the full is b n and what we do is like we
we remove that so that it doesn't show up here and then we make another request to the hard cover servers to get the books matching those is bends and then the current users like a status for that book in the second request so at this point like chat g p p doesn't know anything about the users preferences or what books they've read or haven't read but we were we were talking about it some this weekend and that would be kind of one of the logical next steps
like loading it with the books that you already enjoy as you're browsing this and having that context so it doesn't recommend things that you've already read or it uses what you've read to recommend things that it knows you'll like
ste (13:58.229)
yeah this in conjunction with the match score that we already have i think it's going to be like a really powerful tool to discover new books and for this kind of guide the discovery where you just mentioned words or moods or vibes or key words or anything really so far it's been proving like really powerful so yeah it's really interesting how it would mix
and when we will actually be able to send each one's preferences so this would basically mean that you're reading preferences would actually matter in how those results would be generated in real time right
adam (14:53.14)
yeah
ste (14:55.329)
that sounds pretty i mean on the promise of finding books that you might like i think this really it delivers even in this state but i'm really like looking forward to how it can deliver when it actually has those reading preferences because like that's the whole having someone who knows all about what you like to read and knows about all the books ever read
adam (14:55.64)
yeah
ste (15:25.669)
that is actually able to recommend something that is outside you know the usual look of books that you see in every feed you see on every network you see everywhere else so that's kind of that's kind of exciting and we have lots of ideas from all of our team on how we can
deliver up dates to this so that's a pretty interesting space to explore i guess and it's been amazing that you put it together so fast it's
adam (16:10.48)
it's it's it's basically one react component and one a pin point on our server and that's it everything else like yeah everything else was using our existing graphic i didn't have to make any changes to the back end just you know
ste (16:17.789)
yeah it's a me i mean
ste (16:27.109)
yeah see the benefits of good architecture they're finally bring up i guess
adam (16:33.02)
that was that was one of the things that i'm like i want to figure out how we can do more of and how we can even leverage the community to do more of it's like um just like labs experience one of the things i was considering was like what if we make labs like a sub domain as like an open source project that anyone can do pars against and it's it's like its own little
ste (16:44.149)
m hm
adam (17:02.98)
um playground for hard cover a pi concepts and then like yeah because ideally i'd love to just make the entire front end open source but i think we're we're still a little ways away from that
ste (17:10.929)
yeah that sounds yeah
ste (17:22.229)
yeah i'm guessing for that we need more documentation right and for it to be more robust but this lab thing sounds great because we've already had some people in this square that who are working on improvements to the way we do imports for instance that start date and date conundrum with we hit and it would be like
adam (17:29.14)
yeah
adam (17:52.undefined)
yeah
ste (17:52.029)
interesting because i guess one of the long term plans is to have it as an open source project but that just requires us bringing it to a level where it's robust enough so that it wouldn't be at risk to you know actually had in way in directions that you
we wouldn't want it to but that goal is there which is this could work up to it really well i'm guessing so i really like that idea having this playground for people who want to contribute because i mean clearly there are people who want to to contribute and it would be like interesting to see how that how that happens
adam (18:40.12)
no
adam (18:45.72)
yeah
adam (18:49.08)
and it could even be like one of the other benefits of it is it is it ends up being example uses of the a p i that people could use on their own and their own projects um so it's it's like an open source example library of usage so like someone could create their own like blogwidget for instance and that could be in hard cover labs and then someone could just copy that and use it on their blog or things like that
ste (19:18.809)
yeah that would be really really interesting yeah definitely i mean if you can set that that would be really great and i think we'll make some of the users on this card happy we have some really good developers who i think would be eager to have a try at at this it's also opening up some really nice opportunities really regarding book data
adam (19:19.22)
so some of them
adam (19:23.36)
yeah
ste (19:49.989)
adam and i have been talking about how leather box handles the descriptions for films and it's really neat how they for each film have a really short i think one sentence description than slightly longer description that captures the essence of the film which of course are films is way easier because you don't get as many film movies in general
as you do books books i think they may be like ten or maybe even one hundred to one out numbering movies which also tells you a bit about our mission to like build platform for books but coming back to that i think we can extract like a lot of interesting data for the books using this and using the chat
adam (20:26.7)
m hm
ste (20:48.809)
the a b i i've done some tris i've actually asked it to make me a prompt that would generate that kind of description and it's been working out really well the problem that worked really well was can you impress me with a brief and creative summary of book title that will make me want to read it shorter than ninety characters so that
actually gives you one sentence description for each book and it does it really well i've tided on the metamorphosis i tried it on like some of the books i've i've read on it pretty accurate if you ask it to give you options you can also like choose among the options and the options are really good some of them are like describing in ninety characters which is about well i guess
adam (21:25.84)
interesting
ste (21:48.709)
six seven eight nine about ten words in ten words it sums up the book really nicely and it tries to actually persuade you to read it without offering any spoilers and then you can do that for even longer description so you can ask for that one sentence book description so you can like see what the book is about at a glance and then you can ask it for four hundred character description so that's you know
about two twits long and it tells you all you want about the book that's pretty amazing
adam (22:19.58)
m yeah
adam (22:25.52)
yeah and yeah that would be that would be amazing to store those in our data base and then to show those for every book and it feels like that's like the start of it like having those would be an improvement on the descriptions that we have today since so many of them start with like new york times a winning series from author who has done this book this book in this book and something else all in cap
letters and then it starts the description of the book like it's like yeah yeah we don't most people don't need that like that's that's what you need on the back cover of a book when you're viewing it for the first time and you don't know anything about the context but i feel like the readers that are looking at a book page on hard cover like if they want to look at the author they're gonna look at the author what they're interested in is the book content and so it feels like this is getting much more improved book content for the description than
ste (22:58.609)
yeah
adam (23:25.32)
what we currently have for sure and what
ste (23:27.569)
yeah and i love that it's also yeah it's also normalized so it's like the same length and it gives you that predictability and this you can find like on any platform because what's crazy about book data is it all comes from the same sources i mean it's either google books or open library which is kind of a mess i mean it's a great library but the descriptions from there are really really massy
adam (23:31.04)
oh go ahead
ste (23:57.689)
uh or from i mean the good red pa has been discontinued so there's not even that which is crazy to think that you know the data about the books is i mean basically i guess it's mostly google would that be right
adam (24:15.34)
there's there's one other source which is the paid isbn service and that's what a lot of a lot of a lot of services are paying for that book data and the other one is man i'm forgetting the name of it but a
ste (24:20.409)
m
adam (24:35.66)
another book tracking website that is massive but now amazon owns part of it also
ste (24:41.949)
okay so mason
adam (24:42.96)
oh like like library thing
ste (24:46.069)
oh library thing they got the stake in that i didn't even know of those tentacles
adam (24:49.4)
yeah yeah they bought like a i don't know if it's all of it or just part of it like a ten plus years ago
ste (24:58.669)
well okay yeah that's
adam (25:00.undefined)
but but library library thing actually sells their reviews to um commerce sites so sometimes when you're looking at wallmar and there are reviews those reviews weren't done by wallmart patrons who bought the book their reviews from library thing that they're showing on walmart dot com
ste (25:16.849)
okay wow that's really shady
adam (25:21.24)
i mean it's in their t s but it's a i mean it's a
ste (25:24.029)
yeah but still i mean i wouldn't want my review to like be poached by wall mark to sell more books i can tell you that yeah that's yeah that's pretty crazy how i mean even this i mean considering books i mean it's a huge thing and it's like what five six actors who own the data and it's being gated by huge sums of money so to get it from is beyond you have to i'm wondering if like
adam (25:31.78)
it's true yeah yeah
ste (25:53.809)
getting data viachagipt for most of the popular really popular let's say one hundred thousand books would actually be cheaper to run than actually getting it via de s b service
adam (26:12.92)
oh for a hundred per cent like the isbn service is like multiple thousand per month after the first year and he and like the t p t a p i is like a penny per like thousand hits so like we could like once we had our algorithm and our query correct are prompt that we were asking chattpt correct we could potentially fill all of our data for like a book
ste (26:20.589)
for months
ste (26:42.689)
wow that that would be
adam (26:43.18)
so it's it's gonna yeah
ste (26:47.169)
okay i'm mean i'm seeing why we should do this even if it's like for a mission to get better book data for for books that's like not gated i'm guessing just have something mind book data for let's say the most popular books on high cover because we know what those are in the background knowing that would be like really
adam (26:49.3)
but
yeah exactly
ste (27:16.889)
really great especially because we can generate like really cool snipes i don't think one the one sentence description for the most popular like one hundred cape books of all time as something you can get from anywhere so it would actually be nice to have it i really like that about leather box
adam (27:37.74)
yeah
yeah and yeah yeah i'm a fan of it too i think one other like potential future idea for it is that like this description like when when when you were prompting it it's a it's a generic it's a generic description of the book that's going to apply to everyone but maybe the description could be different if the person looking at it is a big
ste (27:46.229)
yeah
adam (28:09.82)
a romance books or they're a big fan of cytha books so it's like write a description of this book for someone who really loves book this book this book in this book and loves this dana this dana and the strnera and then that's going to change what the description talks about
ste (28:22.909)
that old be like so next level who okay okay that's like that's freaking i mean it's crazy that that's achievable right i mean that's actually like one prompt way and yeah it would be really nice i'm actually wondering if we can ask first suggestions on how to write that algeremfrom tragepity itself i mean
adam (28:28.84)
yeah
adam (28:38.02)
yeah
adam (28:51.1)
probably
ste (28:52.769)
yeah that would be like really met a thing to for it to do but i mean it does write its own prompt apparently so i mean of course the thing knows itself and the way it works better than we do so yeah i'm actually i've at some point i've i've wondered if they actually wrote the
adam (29:16.2)
uh yeah
ste (29:22.809)
documentation for the whole a pi using charge because i think that would be like totally durable and maybe that's one of the reasons why it's so easy to employment plus they have access to the newer models so i mean they could
adam (29:44.46)
my guess is that was written by a human but i could be wrong i mean i could be wrong
ste (29:47.629)
okay i think i'm close to like yeah we could bet on that but yeah maybe it was written by a human but but it's so easy yeah
adam (29:55.undefined)
yeah
adam (29:59.54)
but that's not to say yeah i wouldn't be surprised if some of the documentation was written by just yeah
ste (30:07.229)
yeah
i think in any way we're played pretty close for to the moment when the model could actually like write its own documentation and improve it i've seen that it actually can self correct when you asked for cold problems while that wasn't chegipitat that was another one i think it was cold or something now it wasn't cloud anyways it can self correct before presenting either
adam (30:10.12)
yeah
ste (30:39.629)
ult so before this you actually get the result you get the error you tell it that through another and you get the correct result back but now it actually anticipates that error as well but i don't think it was pitt was another thing but yeah that's i mean it would be interesting to see if you can get at least some suggestions on how to write the algorithm
adam (30:58.66)
m
ste (31:10.049)
for whatever data bases we have
adam (31:15.26)
yeah and and bringing this back to book data one of the one of the things that i'm hoping we can do in the next month that i'm going to start like sketching out is like the next phase of our are like book data set up so like right now the way our book data works at at a high level is we we have a book
and a book has many editions these are editions in different languages audio books m you know physical book book that way you know if you're reading um enders game there's one book that represents enders game and there are many many editions that contain information about the book and behind the scenes we're hitting different apis and storing information about each edition and then we're figuring out
which edition is the best edition and then using that data on the book itself on the book page so we might have fifty different descriptions about a book from fifty different editions and then while figuring out which one to use on the book and i think that's that's that's a good start but i feel like having that g p t description would be a better description for one but also
one of our one of our problems is that as we have all these editions we sometimes don't know what the best edition is like we have like our internal algorithm does like a score for how good an edition is based on how much data we have so like if if we have a cover that's worth so many points if we have a title that's worth so many points what we always have a title if we if we have a release date if we have a format all these pieces of data
us know that this edition has more data and then the better the edition is then the more likely we are to use that as the canonical edition for that book but going this route we wouldn't have to worry about scoring editions anymore we'd just be able to like ask chatgpt for the best one but we would still have to most likely associate what
adam (33:45.38)
we're going to be asking chat to t about a book that means asking it about an edition because the edition is what has the i s b n so we're still going to say like this edition is the best edition or maybe maybe we ask like ou know we have we have a book we ask chattpt what is the best audio book what is the best book what is the best hard cover book what is the best mass market paper back and it tells us the i s b ns for those and then in our data base we mark those
and our as like this is the best audio book and that's the one that people will automatically use if they select that they've read it in audio format so i think like that step is gonna improve our data a lot because then we'll know what the best edition is because of chatbchatgpt sothat's one thing that i'm excited about
ste (34:37.249)
yeah
yeah that sounds great and obviously i think the way we're thinking about editions and we're approaching editions is the most complete and i mean the most comprehensive way i encountered i know what you ap can let you choose edition but let me just share my screen to show people what we're actually working on
just sick
ste (35:15.069)
so we actually have the book tracking over here and
basically each edition would be associated with each date read so each date each read of a book would have specific edition associated with it so when you would change a book you wouldn't do like you normally on other book apse where you would change the edition for the whole let's say instance of that book for you on the network
actually like say which one you actually read r listen to or read as a new book which i think is more reflective of the way reading is done because you are reading certain certain edition and you know there is an edition which is the most popular but it might not be your edition it could be we're selecting the most popular
are the best let's say edition as a default here when you actually read it but you can also change that there's lots of people who read the books in foreign languages translated books or our books that are
a certain edition that's you know rarer or more special and this lets you actually mark that edition in your read so yeah more preview of what we're actually doing now for the read and the book tracking dialogue so yeah that's pretty exciting how it will pan out
adam (36:57.undefined)
m hm
adam (37:06.1)
um m
yeah yeah i'm excited about using open the eye for that now especially now that we've we've tried it out on a tools and kind of have realized how easy it is i think i think one of the differences is that for this kind of book data step we're going to be using a different a pi from open the i the one we're using right now is the chat a p i and the chat p i has like a series of messages back and forth where you're saying
the system says this and then you have an idea of an assistant so our assistant is saying like return books with the i s b n like that's not something the reader ever sees and then you have the readers messages so it's like a conversation while this a p i would be using just the regular prompt a pi where where we're saying um give us back this data here's a big jason file of what we want the format to look like just give me back all the data in this form
and if we can do that that will be amazing because it will basically be using it as an a p i just like we're using open library or just like we're using google books except it's an a p i that's being dynamically created by what we ask it for
ste (38:11.509)
yeah
ste (38:26.349)
yeah that's goin to be crazy i'd actually like really see if we ask for like that exactly same thing you said a just put it in a prompt and see what it output because it might offer some some tips and yeah it's crazy to i mean for all sorts of tasks raging from creative ones to programming ones i think it's a real good accelerator so
not a replacement but an accelerator when you're stuck when you're like yeah that's that that's really that's really good and i've been trying to use it i've been trying to actually see how it can be used for designs as well so far i've been getting like feature explanations for it so i've been asking if i have an apcoltharcover which does book tracking what
the things that are most likely to be good features for it or something along those lines and it's cool because it actually gives me most of the features were building so props to us because i mean apparently we've been as good as the hive mind so far r some things which are on our road map so it's giving me like social features a really good book tracking experience setting goals
these are things that were actually generated and you know it's good because knowing having access to all human knowledge is probably like like in a position to give you like an exhaustive list of all the things that people are looking for so yeah for design definitely usable in that regard i'm wondering what if and how it would match with the ideas we've generally
it for recommendations and suggesting books in general to people but yeah it's interesting it's really good to accelerator for work so far
adam (40:41.62)
and and there are i feel like with the book data we're going to hit a lot of points where were like we need to make decisions about like book data like for instance i was going through in cleaning up um the brendon sanderson the author his author page for like all the books that have been added under him and there are a lot of books that have been added that we're in foreign languages and our algorithm on the back end
didn't pick those up as being an edition of an existing book so it created a new book with the foreign language title while we also have the english title but we want those to all be the same book but we just want them to be an international version under the book that we will have its own page and we'll have its own translated into later but we don't want it to have we don't want it to say like rene anderson has ten thousand books we wanted to say he has exactly the unique amount of books
ste (41:16.009)
yeah
ste (41:22.589)
yeah
ste (41:38.269)
yes
adam (41:41.54)
that he's written across all languages in one language so one of the like the way is that we can handle something like that i was playing with using g p t like like i was going through it and as i was going through it i was like copying a title like and asking it like this book by like this this foreign language title by brandon sanderson what is the english title of this book and for the most part it was able to come
ste (41:45.689)
yeah
adam (42:11.42)
back and find out what the what the title was even though if i searched for it on google or if i searched for it on good reads it wasn't giving me the title which is crazy to think it was it was better than google or good reads at determining what book it was
ste (42:21.269)
oh wow okay
ste (42:27.129)
well okay that yeah that's that's actually major i mean yeah i wasn't expecting the second part uh yeah that's actually pretty cool especially because you can actually do that on algorithm basis so you can tell it that for any book title and maybe we can use it for those associations i've noticed that as well we have some extra books in foreign languages which would be great to
associate under an edition so yeah for for book dat a definitely i'm wondering how this will this will play out definitely exciting to to see it to see it in action
adam (43:19.72)
yeah trying one out now so like this is this is what i just
adam (43:33.86)
did yeah like i was i was going through and
so like i asked i like this is the title that we have on hard cover for this book and this is what it says and i could say like i could probably say like a what is the i s b n of the english version
ste (43:45.669)
oh here we go yeah
adam (44:00.12)
and i have a feeling
ste (44:00.369)
here we go
adam (44:04.66)
okay so if i take this and if i mean let me share this window instead
ste (44:06.009)
here we go
adam (44:15.08)
so
that was that one so if i go to fire flight
uh fire fight and go to editions and it's the first edition so yea like it like it matched it to the addition that we already have in our data base which is yeah like
ste (44:30.329)
oh here we go
ste (44:35.729)
yeah that's crazy yeah
yeah it's going to be great i actually asked it for one of the other important book data items that we could improve which is book covers and author images and this one isn't connected to the internet of course so we couldn't search for ur of the picture of the author or the picture of the cover but i'm guessing some of the other a p s that are connected to the internet
adam (44:54.76)
m
ste (45:10.769)
i'm thinking for the other other models those could search so i'm going to try it on their other side to see if we can actually get you better you are else for those because it would save a lot of time and yeah basically it would i guess up date the covers dynamically i know someone has been asking in this quare
if there will be a time when the users can suggest edits for the book covers or uploadtthetheir book covers as well so i'm hoping to do something for us to be able to do something even better which is actually like get better book covers from like over their internet because they're bound to be somewhere so yeah that's also interesting
adam (46:06.78)
that does bring up that does bring up a good point though about like all this data that we get from open open eye i would still consider it to be like unverified data like no user has looked over it and said like this is correct so the more the more specific data we ask for the more of concern that's going to be so for instance if we're asking we're asking
ste (46:27.249)
yes
ste (46:33.849)
m hm
adam (46:36.84)
it for like what is the gender of author name and we're adding that to our data base i think we would really need to make sure we also allow the ability for users to go in and mark that as wrong let us know what the actual data says because otherwise yeah we we can't rely on that data being accurate when it's based on the internet opinion so yeah i think
ste (46:50.989)
yes
ste (47:02.609)
yeah that's a really really good point yeah that's that's like an excellent yeah i'm guessing over here what's good is that you know for key things like descriptions and more sensitive data which you know would take like a lot like may be an army of people like working tirelessly to generate descriptions for one hundred thousand two
adam (47:07.48)
i think that that will be like a big step
ste (47:32.309)
hundred thousand books it will do like a really good job i think in our case the job of actual librarians would actually be flagging stuff which is wrong so that would be like way easier than actually generating or uploading covers or actually you know uh doing editing that data themselves so i think for every item in the
uh data on the book page there should be a way of i mean do we have flagging now so i'm thinking of a way that in design we can
mark the things that are inaccurate and that would tefinitely be way easier for us to correct rather than generate because i think in most cases it would be accurate but of course there would be those cases in which it won't be because it's still making stuff up it's getting improved i mean this thing was released in november so from
and basically many of the things wrong with it were fixed because basically prompted to not say those things so patching it up actually worked really well but it's still making things up so for that data that it does make up because usually when it doesn't know something unless you tell it to not make things up if it doesn't know something
it makes things up sometimes i mean lately it's been saying i don't know or i don't have access to that information but yeah it can't make things up so for those cases i'm guessing flagging with flagging any piece of info but on a separate page so you wouldn't have a little flag next to everything would be an option
adam (49:22.1)
yeah
adam (49:50.74)
i asked it how confident are you in this on a scale of zero to a hundred percent like as an a language model i am highly confident that the isbn of the english version of firefly by brandon sanderson is my confidence level is a hundred percent as this information has been obtained from reputable sources and is widely available so i i'm really curious on like if we'll be able to trust like the confidence level like like if we make that a variable like you know if we ask
ste (50:13.749)
yeah yeah
adam (50:20.46)
an author's gender and we say like how confident are you that this author's gender is x like will that data be will that will that help us or is that just going to muddy the water even more i think i think either way
ste (50:21.049)
yeah
ste (50:32.189)
no it might actually help us because if you put in a prompt that actually like pushes it to tell you if it thinks it's wrong it's actually going to give you an accurate answer or it gave me an accurate answer as far as i could tell but yeah that's actually like a good good idea
i'm wondering if four sources where it's i'm wondering if for anything it could say like i'm eighty percent confident where i'm twenty percent confident
i'm guessing not yeah
adam (51:14.42)
i asked it about
i asked it about some one's sexuality and how confident they were and it said they were hundred per cent but they said it is important to know that people's identities are personal and can be complex so it is possible that she may not identify with the specific label or may use different labels to describe her sexuality at different times so very cabot
ste (51:35.849)
okay that's actually a good disclaimer that we can put next to like an author's gender because i mean that's the case many times i mean that's that could actually be a good disclaimer for for us as well okay
adam (51:48.86)
yeah
ste (51:52.289)
it's a
adam (51:55.48)
yeah i'm a i'm going to be doing some some mock ups in a figmafor some sketches on some how i see like some of the work flow of some of this working um very low fidelity just here's here's kind of how i see the interaction working from a user standpoint with the idea that where we're getting a lot of this data from opening eye and external sources loading it into our system and then giving people the ability to you know tweak
errors on it edit it and i think i'd like to even lean away from reporting and more make it just anyone can just edit the data and instead we give like a system for um like rolling back errors if we have bad actors and automatically pro actively banding people who are um maybe making inaccurate edits so i feel like that
ste (52:53.609)
that ud be cool are you thinking to open this up ye do you think this should be open to librarians solely or to like any reader on hard cover would you have to have a status for
adam (52:56.36)
yeah
adam (53:08.42)
i think i think it would be open to everyone
ste (53:11.349)
okay that's that's going to be an interesting experiment i mean if we have like back up yeah i don't see why why not everybody could be a librarian i actually on the top i wanted to ask you because i found i was looking into libraries and i found some really interesting people on tick tock who either have bookblogs or you know
uh taking this role of being book curators so for instance i found the really nice blog which i'm urging encouraging everyone to visit is called solo books it's x l o books and the it's non lined book store for intersectional and an anti colonial list books and uh i was thinking having some
adam (54:08.86)
m
ste (54:11.329)
and like that use hair cover how it would benefit them and if it's useful to have certain let's say uh flare user flare for librarians or for book star owners that can benefit them in some ways i was even thinking of how it can tie into because
adam (54:35.66)
m
ste (54:41.709)
all of the books on this particular store are books i haven't like really i think they're they'd be really good to show case on hard cover so if someone had that account and let's say you know those books would appear in the few i'm sure they would find more readers and i was actually planning to do like a library run around like ball bok
adam (55:11.2)
hm
ste (55:11.209)
shops in london to see actually like talk libera if you know they're like the owners are around or if the cashier can tell me anything um to see how it could benefit them the main us case i thought could be useful this making book lists of the books they have in their inventory for instance the book store forks
to books i think is i have to check i think it's through water stones or no it's not water stones his bookshop but work i'll have to check don't take my word on that but anyway i think just book discovery not necessarily like the way books are sold could be like a good niche for for
adam (56:07.4)
oh
ste (56:11.349)
for us and i'm wondering for those books which might not have good book data uh this would basically allow those people to edit the book data themselves right
adam (56:25.48)
hm yeah
ste (56:28.189)
if we do this
adam (56:30.98)
and yeah they would they would be able to edit it just like anyone else
ste (56:36.869)
okay that sounds that sounds promising i'm wondering what sort of trouble will get into i mean right now we don't have like pam content so the users the readers have to go through on boarding and i on't i haven't i don't know actually if any bots can actually go through on boarding without us knowing um do we have a system in place to verify like
adam (56:42.26)
why not
adam (56:46.1)
yeah
adam (57:05.1)
yeah
ste (57:07.449)
if they're real people like who do that
adam (57:12.2)
we we do require that they confirm their email address um so at least like they would be able to use an account for i think forty eight hours and then after that they have to confirm their email address to use it but beyond that not really so yeah i think if we did allow anyone that's logged in to edit a book we would probably want to add we would want to we would want to do it with a lot of smart limit
ste (57:26.189)
okay
adam (57:40.82)
tations to prevent someone from just like using the a p i to just destroy our entire site like limit the number of limit the number of edits you're allowed to do per hour maybe don't allow don't allow someone don't allow someone to edit unless they have like x number of books in the library you know just like a basic basic set up or maybe like you have to have had an account for a week or or maybe like your the amount
ste (57:47.549)
yeah true eh
ste (57:52.109)
yeah that definitely
ste (58:05.029)
hm
adam (58:10.92)
able to edit increases over time so like the first week you do and edit you're only allowed to do five edits a day and then after you've done that you're allowed to do ten edits just like increase it over time
ste (58:20.109)
yeah
yeah yeah we can also allow supporters to do as many it's as possible because those readers are verified so and that would be the like oidinsnting for whoever wants to edit books that you know they're they're feeling strong towards
adam (58:28.92)
yeah
adam (58:32.1)
exactly exactly and
adam (58:42.26)
yeah and when you were talking about like the flare for librarians or if we had fair fair authors or bloggers it made me it made me think kind of about like a rotten tomatoes has like their audience score and their like critics score like we could very easily do something like that where we have like here's the librarian score for this book here's the loggers score for this book here's th readers score for this book or here's like the students score
ste (59:01.209)
m
ste (59:09.829)
hm
adam (59:12.06)
maybe you can tag that you're a student
ste (59:14.589)
yeah that would be great actually i mean let's think of some flares i mean or like i got like five or six just then and yeah i mean if that's as easy to implement it's basically adding more options for flares and i think that's really cool we could tied with great ings and it would be like really interesting to to see
adam (59:38.34)
yeah
ste (59:44.429)
it's yeah it would give
adam (59:44.56)
yeah
ste (59:51.549)
it would be interesting to see what librarians are on the platform what students are on the platform what bookplockers are on the platform yeah that's going to be like really really need
okay yeah
adam (01:00:08.22)
this was a fun talk i have i have some additional ideas for book data now that i'm gonna try to write up a write up some notes on because yeah that's it's going to be our next big one that i'm working on so yeah this is good brain storming thank you
ste (01:00:23.029)
yeah i was really good yeah it's been a really fun talk and i'm really liking how you know this direction for book data and for recommendations and for like what else we can do shapes up i think there's going to be a lot of opportunities with the things that you know are popping up everywhere it's really like exciting to like
you know be able to use this and now that we almost have the androythat out we'll be able to like actually show these up dates and you know make the more visible to readers which is definitely really cool yeah wait well until next
adam (01:01:18.62)
very exciting
cool
yeah talk to you next week have a good one see it
ste (01:01:24.109)
yeah okay have a good one my ready