00:05hi listeners welcome to no priers today
00:07we have Ivan Z co-founder and CEO of
00:10notion the Beloved productivity
00:12application for notes tasks and
00:14knowledge base they recently launched an
00:16AI Q&A interface as well as a calendar
00:19application we're super excited to have
00:21Ian thanks for being here again Ivan uh
00:24we are going to start with the hardest
00:26question which is what is notion Notions
00:28always pretty hard to Define because can
00:30do so many different things um but
00:32that's also our goal we want to give
00:34people one tool that they can do their
00:36most work with for personal user that
00:38means all your personal notes all the
00:40planning for a trip for for your wedding
00:42for business for Enterprise for company
00:44that means all your documents all your
00:47task all your issues calendaring
00:50knowledge basing in one tool um the
00:53reason we want to do that because just's
00:54just so much fragmentation in the market
00:56today we wish like it wouldn't be nice
00:59as one place to do your most work and
01:02the our approach here is rather than
01:04trying to cram all of use cases into one
01:06product uh what are the underlying
01:09software building blocks what are the
01:11Legos that power those use cases can we
01:14give users those Legos so they can be
01:17creative with software themselves they
01:18can create and Tinker their perfect
01:21workflow for their personal life or for
01:23their company and none of this is new by
01:26the way like people back in the
01:2880s even' 70s tried this kind of
01:31building blocks approach to software
01:33we're just trying to take a modern spin
01:35with cloud and with AI uh to what it's
01:39like to break the prison of application
01:42based software it's dramatic to think
01:44we've been living in a prison of SAS
01:46fragmentation the last two decades but I
01:49I do think it's actually um uh you know
01:53surprising to hear eight points of view
01:56that is so obvious which is like of
01:58course we want one tool where the data
01:59was interconnected why do you think
02:02people um why do you think more people
02:05don't try that to have unified tools and
02:09underneath I think people try for
02:11different angles like even fairly
02:14recently there's this thing called NOCO
02:17right NOCO is like coming from this kind
02:19of like power user developer angle of
02:23wouldn't be nice everybody can modify
02:24this underlying software they use every
02:26day that's one angle that it wasn't
02:28coming from the angle of the the
02:30knowledge and data wants to be in one
02:32place right and language model sort of
02:35give another angle the underlying
02:37knowledge in the bending space wants to
02:39be one place wouldn't be nice in one
02:41place right and the macro is also coming
02:44from the uh the budget place wouldn't be
02:47nice rather than pay for five different
02:49vendors and all C based business just to
02:51pay one vendor and save some money so
02:53they have different angles from
02:55different times um I would say we are
02:59more come from this kind of computing
03:02and medium and literacy angle like you
03:06and me go through school to learn how to
03:08read and write know English and Chinese
03:11we spent years to do that we all know
03:13how to do that the world the same
03:17MacBook for most people are are very
03:20rich as I more like a machine to do
03:22typew writing or watching YouTube uh not
03:26much more beyond that it's not very
03:28creative right um wouldn't be more nice
03:31that more people can use their software
03:33more creatively right because there's a
03:36separation be between people who can
03:38make software and people who use
03:40software that's why sf's rent is so
03:43expensive because we're the modern day
03:45Detroit or Manchester right we're the
03:47factory of the world um Notions largely
03:50come from that angle which is the
03:53original angle we were inspired by early
03:56Computing Pioneer they thought about
03:57that angle right they thought about
03:59Computing could just be like literacy
04:01one day everybody can do it I guess they
04:04didn't expect AI might make that even
04:07give a really interesting twist to it
04:10because now language model AI can not
04:13only to create software but also do a
04:16lot of thinking working for you so the
04:19future is pretty interesting no so for
04:22someone who thinks on you know um span
04:25of like Decades of you know what should
04:28Computing look like and what were um
04:31what were the most ambitious plans for
04:32personal Computing you know uh three
04:35four decades ago like what are you most
04:38excited about seeing from AI broadly
04:40over the next decade I think 34 decad is
04:44bit too long if AGI happen that time
04:46like Computing might not be
04:51decade I think a one sleeper category is
04:54the the drag thing embedding space the
04:57de case might be too long we're saying
04:59the next next year or two now the
05:01language model can understand what you
05:03put into a computer understanding so
05:06rather than you to the organization to
05:08make you retrieval retrieve the
05:10understanding more easily machine can do
05:12that better than anybody else can right
05:14so before that uh we use keyword-based
05:17search where you find your coworker who
05:19remember that that cue where does that
05:22information sit now just ask notion Ai
05:25and you get that in second so that's one
05:29I'm personally really excited about I
05:30think not enough people talk about it
05:32and of course the other one is like the
05:35agent the workflow side that that's has
05:37a lot of buzs already so that's
05:40too you and Simon uh you said bet the
05:44company on AI and are you know have real
05:47conviction and as you are building out
05:49the team like um what does the talent
05:52look like you like have or need to make
05:55um notion and AI first company and I
05:57kind of argue you folks are one of the
05:58earliest adop of AI at scale as an
06:00application so part of the question in
06:02some sense is you built so many
06:03interesting things like what are the
06:05people that you now need to sort of
06:06build the next level stuff in addition
06:08to what you already have on the team in
06:09early days kind of just Bo Force slim is
06:12really good thing k a lot of thing and
06:13learned really quickly right I would say
06:16notion is a company where largely people
06:18interested in interface and design a lot
06:20of full stack front then heavy folks and
06:23back and people who scale we have
06:26somewhat a small team of search but we
06:29don't have too too many ml folks almost
06:32nothing and in at least my learning our
06:35learning in the past year or so building
06:36for AI is okay you m Mo folks
06:40are important it's kind of like you no
06:44longer do a this deterministic thing
06:46that you can see how it works it's
06:48almost like a I don't bake but feels
06:50like a baking right you have to like do
06:52something get the thing ingredients
06:54ready run through rinse press the button
06:56and wait for a while and see does it
06:58come up it's a different way different
07:00sense of patience and different type of
07:01personality to do that well a lot of
07:06preparing my friends call that
07:08probalistic software engineering kind of
07:10like I think it's it's morphed into this
07:12sort of stochastic world or at least
07:14partially stochastic yeah so one is like
07:17maybe gardening feels like that way I
07:19don't Garden either uh so that's
07:22category people are to to me is pretty
07:25necessary the other category is like
07:27people who are curious and learn really
07:30fast right it's like okay um like the
07:34group of promp like language prompt
07:37engineer language model sort of make
07:39everybody like a a real time machine
07:42learn learning engineer you just prompt
07:44write then you can get your stuff right
07:47and there's a lot of trick and
07:48techniques uh um and how does that plug
07:51into user interface uh I think this
07:54category of people call AI engineer or
07:56something there's an terminology form
07:58they tend to be pretty young they tend
08:01to be like we have someone like under
08:03drinking age working ocean uh they they
08:06fit into that bucket and I think both
08:08seems to work quite well um we don't
08:11have too many researchers and notion
08:13that's another one I think uh um will be
08:17important but we're fundamentally sit in
08:20the application layer so it's more about
08:22apply side of things more manipulating
08:25the models and making sure you can scale
08:27them to like user outcomes and things
08:29about that and another part is like the
08:32scaling part right how do you scale to
08:34like tens of millions 100 million users
08:36it's bit it's a problem on its own to
08:39it's beyond just a demo on Twitter right
08:41It's Tricky no yeah so you have often
08:45said that uh notion is less a
08:49productivity company than an application
08:51Building Company how do you think about
08:53the initial use case and like how what
08:55makes you believe people want to build
08:56more application I don't think people
08:58want to build more applications what got
09:01me started in notion got I started in
09:02notion it's um last year college I read
09:06a paper by uh one of the Computing
09:08Pioneers Douglas angelar and he talked
09:12about his papers name augmenting human
09:14intellect so every day we use software
09:16today is very much like application when
09:18you go into one application do one thing
09:20but for that generation of computing
09:22people in the 60 70 80s computers are a
09:25lot more software are a lot more
09:26malleable you can actually Tinker it
09:29modify right small talk you can go into
09:32and change the how operating system work
09:33on the Fly um that really inspired me
09:37like today people software are so rigid
09:40can we create a new breed of software
09:42that people can modify can and can
09:45change and customize and bring back some
09:47original esos of those early Computing
09:50Pioneers that's why we started notion um
09:53the hard lesson for us is like like you
09:56mentioned most people don't want to
09:57create software they don't wake up said
09:59hey I want to create my perfect project
10:01management tool my project perfect
10:03knowledge base they boss ask for
10:05something they just have to get the work
10:06done right um so the in some sense our
10:10learning and pivot is instead of giving
10:13people those um software building tool
10:17we have to package the software building
10:18blocks together as ready to use
10:20templates as ready to use use cases then
10:24people can adopt really quickly so you
10:26were one of the earliest adopters of AI
10:28in terms of appc with any real scale and
10:31I think it's impressive how quickly noan
10:32ended up starting to work in this area
10:34how do you think about how that impacts
10:36different aspects of what you built and
10:37what you're building going forward and
10:40how does that impact that vision of
10:41saying okay we have this um this
10:43effective platform that allows people to
10:45both interact with uh documents or core
10:48use cases in simple ways add things like
10:50calendar but then also go in very
10:51interesting directions in terms of both
10:52a set of applications and templates they
10:54can use yeah I think we're lucky like I
10:56mentioned we're not trying to build
10:57specific use cases right we're trying to
10:59build a Lego bricks that power those use
11:01cases um what are those Lego bricks text
11:04editing is a one fundamental Lego bricks
11:07most software have that piece relational
11:09database a table is one fundamental Lego
11:12bricks right um different form of
11:14permission commenting so we've been
11:16spent five plus years building those
11:18Lego bricks and feels like boom AI just
11:21jumps in almost like a brand new car
11:23engine and can power those Lego braks in
11:25brand new ways uh so feels very lucky in
11:28that way and that because we've been
11:30building those Lego bricks and refining
11:32those allow us to ship features uh
11:34pluging with AI really quickly we one of
11:37the earlier one to launch AI writing for
11:40productivity software at scale because
11:41we've been spent years building a text
11:44editor uh we can do uh AI power database
11:48table features really quickly because
11:50we've been building a relational
11:51databases um we've been building a
11:53knowledge base for a long time so we
11:55launch AI Q&A really quickly fairly
11:57quickly uh the rack system on top of oce
12:00because we have those Lego breaks uh so
12:03in some sense kind of like at just right
12:05moment right time for us right how did
12:07you um begin to like resource and
12:11prioritize this effort because you're
12:12like ah magic we have this engine it
12:14applies for our Lego bricks um and then
12:17he started shipping pretty quickly but I
12:19think there are a lot of organizations
12:20right now trying to figure out what to
12:21do with AI and so you know in terms of
12:24like designing the features prioritizing
12:26that effort versus everything you're
12:28already doing and a rapidly growing
12:30software company yeah I think I had the
12:34conviction my co-founder Simon actually
12:37they all had the conviction um funny
12:40because we all live in the mission right
12:42and openi initially still is in the
12:44mission and uh some of my friends
12:46especially Simon's friends work at open
12:48remember we go to their office they
12:50during the DOTA days they like what is
12:52this company doing kind of interesting
12:54and Simon and some people in notion saw
12:57early that most gpts like what is this
12:59thing spin out text sometimes gibberish
13:04um I personally I have to admit I slept
13:07on it on on gpt3 even saw gpt3 feels
13:11like what is this thing useful for it's
13:14like yes for marketing for Content
13:16writing Crea first draft didn't really
13:18click for me um personally for me was uh
13:22fortunate enough to start early preview
13:24gbd4 and that's like oh wow this thing
13:29it can reason it can know how to do
13:32things has this little bit workflow
13:34power to it that's a big a heart for me
13:37and like it just give me it personally
13:39give me so much conviction like this is
13:41going to change everything if you think
13:43about what knowledge work is in right
13:45why do we use software fundamentally SAS
13:49software is all we all in the same
13:52information people paper pushing
13:56activity right it's like a piece of
13:58paper coming in front of you a
14:01human like change a couple bits push to
14:03another human Lang model can do some
14:06form of this now so uh so that just like
14:09give me the conviction like this going
14:10to completely change everything we do
14:12with a computer and after that we sort
14:16of just bet the company on it uh like
14:19we're lucky enough to have those Lego
14:21breaks and then what come which legal
14:22breaks can works well with AI which
14:25doesn't we're trying to figure that out
14:27who inside a company are good with this
14:29technology we have search but it's not
14:32like we don't have a lot of ml folks so
14:34need to hire more ml folks need to get
14:37people inside a company to have similar
14:39convictions so we can move in the same
14:41direction it's quite interesting it's
14:44like so must dinosaur feels like when
14:47Astro hit the earth and like what do
14:49they do there yeah yeah there's a lot of
14:50change coming for sure it's a lot of
14:52change yeah what do you think is missing
14:53from the capability set because to your
14:55point I think a lot of people weren't
14:58really thinking about AI too much until
14:59chat GPT and gp4 came out and there was
15:02a period of time where three and then
15:033.5 and you started to see the
15:05capabilities incrementing up and
15:06entirely new businesses are suddenly
15:08able enabled with each sort of step with
15:10the next GPT level model you know GPT 5
15:13or de your point rag adds a lot of
15:15capabilities what are the biggest
15:16missing gaps for you to take full
15:19advantage of this technology is it
15:20future reasoning is it better uh
15:24thinking and knowledge like what's the
15:26yeah I think all about that above to me
15:30like technology is fun we're in the tech
15:32business technology is fundamentally
15:34about tradeoffs right it's like the
15:38plastic can do things that wood cannot
15:40do and we discover plastic and then we
15:42figure out new things we can bottle
15:44water like this before you cannot bottle
15:45water with a wood table right so um all
15:49of a sudden we have this thing called
15:51language model you have a new
15:52characteristic that deterministic
15:55software cannot do in the past and and
15:59we don't really know how it's made fully
16:01so every month every week if you're on
16:03Twitter people discover new techniques
16:05to get more out of this and for
16:09companies entrepreneurs they're also
16:12making tra off discover what how the
16:14market react to this capabilities of
16:17this new W this new language model and
16:19so it's a constant evolutional cycle
16:21happening really really fast right now
16:26um I think if with that mindset uh what
16:30are the dimensions um on the technical
16:33side on the technology side itself yes
16:35the model gets larger context Windows
16:38more reasoning better speed smaller
16:42footprint um those are all great like
16:45notion uh to power the work force I
16:48would really need like we learn like gp4
16:50is smart a cloud to a Smart we need that
16:54intelligent to do reasoning or for a
16:55tech summarization cheap fasts better
16:58right that's the technology side and in
17:02my opinion there's so much about human
17:04behavior as well just like inertia in in
17:07our personal Behavior companies risk
17:10tolerance and that's slowly evolving as
17:14well right like like what Steve Jobs
17:17always was talk about you cannot make
17:19something too new you have to be largely
17:21the same and change one thing and two
17:23thing right virtual Appo the off white
17:26guys is like 3% difference just push the
17:28boundary so people can accept it but
17:31still also new right to me it feels like
17:33language model power application are
17:35kind of in the same pH if it's too
17:37different people like what do they do
17:39with it right such alien Behavior it has
17:42to rag is pretty nice because the larg
17:44the existing Behavior but better output
17:47right can you describe the um the Q&A
17:50product for people who have an
17:52experience right essentially everything
17:56notion notion help you
17:59remember right and this is not just
18:01apply to notion pretty apply to most
18:03rack systems like why do we use computer
18:05when you to store things I need to
18:07recall things before language model and
18:09rack the recall largely happen based on
18:11keywords right you the keyword has to be
18:14precise or there's some lexical Tech
18:16tricks you can like recall easily um uh
18:21imprecisely what rack happens language
18:23model can actually understand what
18:25you're put into there so you no longer
18:27need to organize your information in
18:30notion whatever you throw in there you
18:32can find it later what that means is for
18:35a person or for a company for a team you
18:38can have perfect memory and not only
18:41have Perfect Memory the the right piece
18:43of information if we design our software
18:45right can push to the right person at
18:47the right time right that's probably
18:49more than 50% knowledge work right we're
18:52still perfecting the system I think
18:54we're one of the first on the market
18:56that apply a skill we still have a some
18:59somewhat a waiting list because it's
19:00hard to do this at scale still um but
19:03for a company for a team before search
19:05one of a weaker point but with rack you
19:07completely change that I changed how I
19:09use notion I can just ask a question to
19:11notion like um how large when I removing
19:14out of the SF office to a new office and
19:18someone in the company wrote in some
19:20documents I don't have to Ping five
19:22different three different people to find
19:24the answers if it's in notion we'll find
19:25it for me right um everyday Engineers
19:30designers operation people just keep
19:33asking each other on slack or in email
19:35such a question each question is 10
19:37minutes writing the answers 20 minutes
19:40to find answers and there's a delay in
19:42the Middle with notion Q&A you can
19:45completely cut that right into in
19:47seconds we're just at the beginning of
19:49what rack can do for work it's pretty
19:51amazing I feel like Rag and edings are
19:53very under discussed or underappreciated
19:56in some sense relative to the impact
19:58they really seem to be having or
19:59starting to have and I think notion um
20:01Q&A is a great example of that I guess
20:03the other thing that you folks just
20:04launched is calendaring and if you can't
20:06talk about it or if there's nothing to
20:08talk about that's fine too but um I feel
20:10like one of the really interesting
20:11things that people are talking
20:12increasingly about is agents and sort of
20:14the agentic world and there's a lot of
20:15capabilities missing to really make
20:17those valuable but in the can context of
20:19a calendaring application you could
20:21think of all sorts of ways that having
20:22AI act on your behalf or help understand
20:25things can be incredibly valuable and so
20:27I curious how you think about the
20:29application of AI relative to calendar
20:31versus you know some of the core um
20:33information related things that you just
20:34talked about maybe we can group AI stuff
20:37into like at least in my mental model
20:39it's the rack the knowledge information
20:41retrieval is one bucket knowledge bucket
20:45then there is this workflow bucket right
20:47use the word agent uh um that's in that
20:51bucket CER somewhat in that bucket um
20:55why do we need to meet why do we need
20:57Canada because we need to meet and we
20:59need to schedule time we need to figure
21:00out exchange some kind of bits between
21:02my brain to your brain right um can that
21:05bit can that exchange be done by a
21:07language model maybe and can the meeting
21:11time be done by scheduling be done
21:13that's like a baby step right um and
21:16most things we do has this kind of time
21:18Dimension to it uh can language model
21:20help us shuffling our schedule yeah it
21:23feels like there's also the information
21:24retrieval piece of it because you know
21:25if my calendar autop populated
21:27everything I need to know about the
21:29meeting or the people attending or other
21:30things that's incredibly valuable as a
21:32user of a calendar and so I just feel
21:34like there's a lot of these things that
21:35kind of tieen together both in terms of
21:36the coordination which you mentioned and
21:38the workflow and then separate from that
21:39there's just what do I know about this
21:42the calendar part it's the simpler part
21:44of the workflow like the holy grills
21:46kind of like can just the agents robots
21:48do all our knowledge work for us right
21:50it's a really interesting framing that I
21:52didn't have before of a bunch of the
21:54work you're doing at notion actually
21:55eating into like communication
21:58right it's it's sort of obvious in
22:00retrospect but like if you look at what
22:02you describe like like am I really going
22:04to slack back and forth about this thing
22:05about you know when we're moving if I
22:07can just know I'm in oce help me know or
22:10with calendaring like you know the most
22:13intelligent version of it is like well
22:14do I need to have that meeting or can
22:16you tell me what I was going to tell me
22:18I know like why do you need to
22:19communicate because there's something
22:22the work cannot be done asynchronously
22:24or by the software itself right then
22:26that's why you talk uh
22:28yeah it's kind of interesting maybe it's
22:31interesting question like are we kind to
22:32communicate more or less with language
22:35model I probably feel it's probably less
22:38uh uh the agent side essentially bet on
22:42language model that's the communication
22:43one question I have for you just going
22:45back to like the implication of rag and
22:48like you can be my brain and do my
22:50Organization for me um like what if my
22:52brain is really disorganized like do you
22:54do you think that uh this changes the
22:57amount of work people should do in
23:00systems like notion at input right like
23:04you know should I be designing my
23:06knowledge base in the same structured
23:08way or can of just dump it all in stream
23:11of Consciousness in theut I think
23:13organization might be we might be moving
23:16away from the organizational world world
23:19why do you need to organize because you
23:21retrieve why do you have index like
23:24index initially are file cabinets and
23:27little IND are sitting on top of so you
23:29can find things quickly right and they
23:33certain uh names or certain Dimensions
23:37but edding and rag sort you just you
23:39have in semantically connection of all
23:42the thing you throw into this file bag
23:45and you can find bring it out however
23:46you like so I think we might be moving
23:49past the need for organization uh that's
23:53really liberating that means on my phone
23:54imagine this experience I'll have a new
23:56idea I see a whiteboard behind me I just
23:58take a picture or write something dump
24:00it and notion going to organize for you
24:03right so that's then that's become my
24:05perfect memory to start later could be
24:07my perfect assistant to help me do
24:09something with this knowledge and that's
24:11the vision we're moving towards right
24:14that's super exciting to me so you are
24:19um this is a question from h m foral
24:21you've been long time friends with you
24:23know you are a student of History you
24:25mentioned duck angelart earlier I know
24:27um you think about you know the
24:28transition in terms of like Alan K and
24:31what he did in terms of simplifying any
24:32of those concepts for like a broader
24:34audience around Computing Vernon's
24:37question was what lessons in history do
24:39you take that um inform your point of
24:42view of like how to treat AI strategy
24:44with notion now like from a prior
24:47revolution in Computing you know how
24:51does it help you decide what to
24:54do a lot is intuition I think understand
24:57understand history give you a sense of
25:00History doesn't repeat itself but Rhymes
25:02so like okay which phase are we in um I
25:07personally think we're sort of in this
25:08kind of bundling phase like um um who
25:12said this like there's only two way to
25:14do business bundling un bundling right
25:17and uh actually the during the breakouts
25:20reading a uh Chinese novel uh Romance of
25:25three kingdoms and the opening sentence
25:27of for that it's uh the Empire long
25:31divided must unite long United must
25:34divide that's has always been business
25:36is the same way too right um we're in
25:40the bundling phase I would say the stas
25:43it's sort of this unboundly
25:45fragmentation phase if we Trace back to
25:48SAS why is SAS happening in the late 2
25:52mid 2000s before that everything's
25:55running on Microsoft that was a bundling
25:57phase early days at PC there's so many
26:00different applications the first version
26:02of the the World Star World perfect
26:04different text editors dbas different
26:07database software the funny fact of
26:09dbase is like they start with dbas 2
26:11because there's so many company go bus
26:14that it sounds like if they start with
26:15dbas 2 people has more credibility it
26:18feels like this this product has been
26:19around for a while so that's the 80s
26:22'90s was this kind of bundling phase
26:25because Microsoft has OS layer on the
26:29and the sad is because the web becomes
26:32good enough to run software right then
26:34so then we have this un bundling phase a
26:36fragmentation phase and then with the
26:39the last 10 15 years it easy really the
26:41money is cheap easy to create company
26:45much too much now feels like there's
26:48like information so fragmented and now
26:50the new technology happening is AI
26:52language model and if you build more
26:55with it or just think more with it
26:57language model wants information to be
26:59one place wants the end points to be
27:02connected so it's easier to is it hard
27:05enough to at current version of language
27:07model do do what you want but imagine
27:08talk with different endpoint that's even
27:12so and so we're in the bundling phase
27:15because the macro but we're also in the
27:17bundling phase because language model I
27:19believe wants the things to be together
27:21think makes sense I also feel like we're
27:22in the bundling phase because the nature
27:23of how Founders think about their
27:24businesses shifted how so um I think
27:27that uh it's interesting because I
27:29remember I don't know 10 years ago I
27:31used to argue with people about oh you
27:32should really buy other companies or
27:34integrate or sort of pull all these
27:35things together and in consumer that
27:37actually happened right like Facebook
27:38bought Instagram and WhatsApp and other
27:41things and they effectively created like
27:42a bundle of social products that they
27:43could cross use in different ways for
27:45distribution or other things but I feel
27:47like what happened is we had a series of
27:49Highly technical Founders because we
27:50shifted in the Facebook era from Cheryl
27:52becoming CEO to coo and you went from
27:55business Centric uh CEO in the 90s in
27:58some cases although there's people like
27:59Bill Gates who learned and adopted as
28:01technologist to very technology and
28:03product Riven Founders who often thought
28:05no matter what product I build it always
28:06has to be better and so I I can't just
28:09think of distribution as my wedge I need
28:11to think of every product as being
28:12Superior and so I'm not going to build
28:13certain things and now I feel like
28:16people are both building great products
28:18as well as bundling them but also
28:19they're much more aggressive about
28:20saying it can be 80% is good it could be
28:2250% is good but I'm going to have a
28:24bundle and that's HubSpot and that's
28:26Rippling and they have very high quality
28:28to their products it's just they realize
28:30they don't need every single edge case
28:31and every feature as long as they're
28:33able to cross sell yeah I think the YC
28:35school's philosophy of build one thing
28:37use internet to find the
28:39distribution that was I think overlap
28:42quite a bit with the rise of internet
28:44right and and feels like there's a value
28:47to create on the other dimension which
28:48is like you mentioned it doesn't have to
28:51be as good 90% that's good but because
28:53the Synergy of things just make a lot
28:55easier a lot cheaper less tabs you open
28:58your browser yeah it's all integrated
29:00you have the information flow or or the
29:02system of record for whatever thing that
29:04you're dealing with yeah I think a lot
29:05of people also just um perhaps lack that
29:08uh sort of historical context right if
29:10you look at the strategy of company is
29:13like horrible right it was very much for
29:17a decade and a half like a dominant at
29:19least commercially attitude of like okay
29:21we're going to buy the second best
29:23product in this additional software
29:24category we want to be in and then sell
29:27the heck out of it worked great right
29:29actually because um was very hard for
29:31customers to deploy these things or
29:32there're just advantages to everything
29:33being attached to a single database at
29:35some point and I I do think there is
29:37some analogy to as you said language
29:39models because having things in the same
29:41edting space is very useful very useful
29:43yeah I think there's bundling of
29:45distribution and bundling of information
29:48the what you're describing to me is more
29:50of Microsoft more like bundling of
29:52distribution langu wants the bundling of
29:54data bundling information so I remember
29:58hearing from Dylan at figma early on
30:01that there was um one crazy user who was
30:04in the product like 14 hours a day it
30:06was you um early on in the uh notion
30:09Journey um being really design obsessed
30:11I think the company has a reputation for
30:13that do you think of notion as like a
30:15design Centric company and um is it
30:18important how do you scale
30:21it I think depends on what you mean by
30:24design design is to us at least to me
30:27it's less about how it looks is how the
30:29system plug together right and then in
30:32that case the trade off you make do you
30:34centralize that thing or do you
30:35decentralize that thing um certain
30:39company work well or certain business of
30:40product work well being decentralized
30:43like operation heavy company could work
30:45that way and notion like I mentioned
30:47we're sort of in the bundling business
30:49our value provided having this one
30:50information space one works space for
30:53people do all different kind of things
30:55so things need to be work well together
30:57it's almost building notion it feels
30:58like building an operating system
31:00building a programming language right
31:03you don't you don't you don't Farm out
31:06to like 50 people to design a
31:09programming language you programming
31:10language are done by one person so that
31:12means the design here is very much
31:14centralized energy so kind of like apple
31:17how they build OS integrated with their
31:20hardware and right like what is the
31:22Apple for software doesn't quite exist
31:25today right it truly doesn't quite
31:27exist so that's what I'm interested in
31:30what we're interested in so in that case
31:32means to build a good product a good
31:34customer user experience we need to
31:37think things more horizontally more
31:39holistically that means the decision-
31:41making tend to be centralized in or
31:43design team or like tend to be
31:46centralized right so um less like more
31:50so more Apple like less Amazon like it's
31:52funny cuz when I first met you it was
31:53just you starting notion and it before
31:55you brought on Simon and and you talked
31:57about things that way even then and I
31:59felt that one of the reasons I was lucky
32:01enough to invest or you know I I came on
32:03board was because you had such a
32:05cohesive view of how you wanted to build
32:07software and you had such a cohesive
32:09design aesthetic and it was your mock
32:12but it was also how you were dressed and
32:13how that reflected into the product I
32:15felt like it was extremely striking you
32:17know like you're one of a very small
32:19number of people I've ever seen where
32:20that design aesthetic has just kind of
32:22permeated everything in a very cohesive
32:23way and so that's one of the things that
32:26got me excited at the time I was like
32:27wow this is capturing a uh aesthetic
32:31that could be an incredible product
32:33platform but you also talked about
32:34things even then I remember in terms of
32:35like okay what's the what's the cohesive
32:38Apple like thing that you can do for
32:39software things so I think it's kind of
32:42amazing to see that consistent thread so
32:44I was just stricken while you were
32:45talking you know by that yeah thank you
32:47yeah like I study cognitive system
32:49cognitive science which is kind of just
32:51like a degree for everything in some
32:53sense it's like a little bit philosophy
32:55a little bit l State computer science I
32:58learned how to Cod when I was a kid and
33:01I did a lot of Art and also in school so
33:05like try not to like there's so many
33:07things you can steal from all of
33:09different places right and it's like the
33:11Bund are sort of man-made
33:14and in notion was most our designer can
33:17call majority of 80% of our designer can
33:20call because the moment you can as a
33:22designer or as engineer you can Cod or
33:24you can design you can make really
33:25interesting tradeoffs right at the end
33:27of the day Technologies at least in my
33:29opinion is about trade-offs what kind of
33:31trade-off you can make that unlock new
33:35user behaviors that's valuable but if
33:39you can do more things you can more make
33:41more interest and trade off the other
33:43people cannot make as a designer if you
33:45can code you know how to change your
33:47design to make easier to build as
33:48engineer if you can design you can like
33:50do the same thing almost like squeeze
33:52the air bubble to whichever direction is
33:54easier to squeeze right um
33:57and therefore I think being more
33:59holistic help at least notion as a
34:01company energy we're trying to be
34:03holistic also help our keep our company
34:05team very small like we're usually one
34:08of the smallest relative to our business
34:10scale because people can think can do
34:13more can be more holistic and people
34:15enjoy that too because they can do more
34:17things it doesn't feels like they have
34:19one role they have to be doing that
34:21repeatedly has a lot of different
34:23benefits but it's much harder to find
34:25such people and important question here
34:27please so if I think about the first
34:30office and this may or may not be uh
34:32true still um no is a no shoes was a no
34:36shoes place did this contribute to the
34:39energy I'm Asian so when you go home you
34:42take off your shoes our first office no
34:45shoes lasts us to 10ish people second
34:49office 20ish people no shoes third of us
34:52no shoes it actually has heated floor so
34:55even better oh very those all in the
34:57mission and the fourth office we try to
34:59do no shoes still in the mission um I
35:03think I made a wrong choice in the rock
35:05the rock kind of hard when you it's a h
35:08based Rock so when you step on it
35:10without shoes with socks only it's just
35:12all hurts so we decide not to do no
35:15shoes at forth office and uh so far has
35:18been stuck that way yeah P intuition has
35:20socks and uh slippers at the front so
35:23that way if you need the padding I know
35:25but then where the question is like
35:27where do you store your slippers it
35:29become stinky like it's aesthetic yes if
35:32you come to our office it's like we're
35:34still try to be um not corporate it's
35:38like they we're trying to use the
35:39furniture that people use for homes I'm
35:41pretty picky about what kind of
35:43furniture is in office like ideally
35:45designed classic that last 50 plus years
35:47so Inspire us to build software that way
35:50right so they made they also made
35:52trade-offs people who design a chair
35:54make t they made really interesting
35:56trade off Force to solve certain
35:58problems if you know the history of it
36:00so we try to in the office use good
36:04software good chairs good lighting so
36:06this back to the aesthetic point that I
36:08made earlier I actually felt that in the
36:10offices as well there's that ongoing
36:11cohesion even the music I remember I
36:13think it was in the second office was it
36:15was always jazz in the background and I
36:17just felt like it all kind of was this
36:18consistent Vibe you know so it's pretty
36:21cool within notion are you using a
36:24singular um underlying l l or are you at
36:27this point using multiple different
36:28things for different use cases you
36:29mentioned sort of the highle reasoning
36:31versus the fast Jeep sort of synthesis
36:34we try everything open air and Tropic um
36:38are the high-end model
36:41uh we want reasoning which is we work
36:44with the high-end model right uh yeah
36:48it's kind of like everybody building
36:50different flavors of this MH so yeah
36:53makes sense and then as you look at um
36:55it feels like with no there's a set of
36:56core sort of templates or use cases um
36:59you know there's things around project
37:00management there's other types of almost
37:03like applications that people have built
37:04to use there's knowledge based related
37:06stuff there's the things that you
37:08mentioned um are any of those you feel
37:10differentially impacted in terms of how
37:12you think about future AI R map or
37:13things that you know will really change
37:16the game dramatically in terms of some
37:17of these areas yeah I would say rack
37:19changed all the knowledge say
37:22fundamentally um you no longer need to
37:25organize so the notion one of thing
37:29people love is the life sidebar right
37:32the life sidebar you can organize your
37:34knowledge base organize your personal
37:36workspace maybe the future doesn't have
37:38to have that like what it's like to um
37:42you know not fall into your own
37:44Innovative Dilemma to double down that
37:47your ex Paradigm but just having a
37:49notion that you can just dump things and
37:51retrieve right that's knowledge sty
37:52that's actually really interesting at a
37:54high level to think that um everything
37:56sort of moves to a form of search over
37:57time move over to search over time like
38:00you're kind of losing organ you don't
38:01need to self-organize information
38:03anymore no in this new world you can
38:05just create a mechanism to interrogate
38:08it yeah at least like you you don't
38:09organize your brain you just dump into
38:11it then you wake up all you remember
38:13that thing right so like it's
38:15interesting some people do like the the
38:17art of lowkey the art of memory you
38:18actually or visualize your brain but for
38:21most people it just works without any
38:23organization and magically right what is
38:25like for software we're getting there
38:28right yeah it's kind of fun are there um
38:32areas of uh software more broadly that
38:35you think are outside of notion scope
38:36that you think are going to change a
38:37great deal from AI wow
38:42um in some sense it's kind of a race
38:46there's like the we're in the B Notions
38:48in the bundling business where are
38:53um we are uh we're in the bundling and
38:57front office business front office my
38:59definition our definition is what's
39:01happening in like imagine a 1960s office
39:05right what's in 1960 office on your desk
39:07you have a notepad you write on
39:09something you maybe have a typewriter um
39:12then you have your binders on the left
39:13and right to that's essentially the
39:15notepad is your documents your notes and
39:18notion your binders of things are like
39:20your weeky knowledge based in notion and
39:23behind you will be the file cabinet
39:25that's your relational datab is in Ocean
39:27right and you have a little push card to
39:28put thing into there then there's a back
39:31office whereas like the Librarians
39:33organize all the things and that's
39:35snowflake right that's the back in the
39:38days IBM we don't touch that uh we
39:40largely touch our strength like I
39:42mentioned is software interface UI ux
39:45which is largely what's in front of the
39:47human we're trying to bundle in this in
39:48one space at the same time there's also
39:52largely back office power use cases they
39:55tend to be vertic cides specific to
39:58Health Care specific to some kind of
40:02workflows very specific by very
40:04essential to store somewhere and the
40:06vertical integrate that use cases uh
40:09that could be aifi too and people in
40:11fact we see this in law we see this in a
40:14bunch very specialized thing that people
40:17have that domain knowledge and trying to
40:18figure out how do you um instead of
40:20human shuffling this let language model
40:23help a lot of that right um the front
40:26office type of things it's kind of
40:27open-ended the back office power things
40:30tend to be specific so I think it will
40:32be a race but the market is just so
40:34large and it doesn't it's not Zero Sum
40:37necessarily maybe you can talk about the
40:40market as you see it for notion so you
40:42know when you guys began I think early
40:45adopter startups were the first to um
40:48get onto notion for knowledge base um
40:51you're a much bigger company now we're
40:53also in a different macro where um you
40:55know startup budgets are are uh less
40:57robust like how do you think about the
40:59Enterprise and helping the Enterprise
41:00adopt AI or or you know do Knowledge
41:02Management yeah we're still very uh
41:05early uh not at scale yet um I would say
41:10uh bundling in trans do a lot of good
41:13things one is you don't have to jump
41:15between different tabs to do things
41:17second is save your cost right like we
41:20save a lot of customers bills for their
41:23project management Tool uh around their
41:26issue tracking tool and and that's
41:30Enterprise really care about that it's
41:32the very CFO friendly today um with this
41:35macro so yeah spond only has many good
41:39benefits besides M besides information
41:41there is also money well Ian I mean this
41:43conversation has spent so many
41:44interesting topics thanks so much for
41:45joining us today thank you yeah great to
41:47see you good to see you find us on
41:49Twitter at no prior pod subscribe to our
41:51YouTube channel if you want to see our
41:53faces follow the show on Apple podcast
41:56Spotify or wherever you listen that way
41:58you get a new episode every week and
42:00sign up for emails or find transcripts
42:01for every episode at nopi.com