0:00Good morning. Um, I'm Karly Kaiser. I
0:03lead the JP Morgan Innovation Economy
0:05Commercial Banking Team based here in
0:06Seattle. Um, and it is an absolute
0:09delight to see you all. So, thank you
0:10for spending your morning with us. Um, I
0:13have the pleasure of getting to talk to
0:16Jordan Tagani this morning who's CEO of
0:18Mother Duck. Um, I think probably one of
0:20our more visible uh, leaders here in the
0:22Northwest. So hopefully some of you have
0:24heard him speak before, but um getting
0:27the opportunity to kind of dig in a
0:28little bit more into his philosophy
0:30around the changing world of data and
0:32how we're going to analyze it and access
0:33it and all of that kind of stuff. Um so
0:35maybe actually to start, would you give
0:36a little bit of your background and kind
0:37of how Mother Duck became?
0:40
>> Sure. Yeah. So um I, you know, I've been
0:44sort of a big company, small company
0:45kind of kind of person, you know,
0:47throughout my career. Worked at
0:48Microsoft for a long time, worked at
0:49Google for a long time. Um, but always
0:51sort of had the startupy bug, but I
0:53never thought like I really had it what
0:56was in it to be a founder. Um, a couple
0:58of times I like larped at it, you know,
1:01I like, oh, it'd be cool to like do this
1:03and then like you just feel like you're
1:04playing and um, so this time, you know,
1:07I I actually just got fired from my job.
1:09I was the chief product officer at
1:11Single Store and um, I was like, what am
1:14I going to do? And I kind of was like,
1:17you know, thinking, well, I should just
1:19hack on something for a while. And so I
1:20I was like, you know, this duct DB
1:22database is like is amazing. Nobody's
1:25really heard of it yet. It was, you
1:26know, several years ago. And um and I'm
1:29like, you know, somebody should really
1:30build a cloud service out of this. And
1:32and I'm like, well, hey, I've built two
1:35cloud data analytic services. You know,
1:37I helped start Google BigQuery and then
1:38I helped um get single store SAS service
1:41uh up and running and turned it into a
1:43business. And I'm like, hey, maybe this
1:45should be me. And so I just started, you
1:47know, I think I got about two days into
1:50it and I got introduced to, you know,
1:53Hanovs and Mark, the founders of of the
1:54open source project like, "Hey, we'd
1:56love to partner with you." And um and
2:00then somebody said, "Hey, you should
2:01talk to my friend. He invested in my
2:03company. He will give you some feedback
2:04on the idea." And uh I talked to him for
2:07about 15 minutes and he's like, "Yeah,
2:09it's a good idea. I want to fund it."
2:11And you know, and ever since then, we
2:12were sort of off and off and running.
2:14And you know, it turns out, you know, I
2:16ended up becoming becoming becoming a
2:18founder and it's been it's been an
2:19amazing journey. It's been almost uh
2:21almost four years since then. Um but
2:23it's been uh I think it's like having
2:26kids, you know, it's not every day is
2:28fun, but like it's certainly it's
2:30certainly rewarding.
2:32
>> I love that. Um so in 2023, you wrote
2:36kind of a blog post that got a lot of
2:38attention about the death of big data.
2:40And then recently you've kind of kind of
2:43updated maybe that perspective based off
2:45of a lot of what mother duck is doing
2:47around kind of how the modern data stack
2:49is evolving. Um given the amount of
2:52capital that has been raised against
2:53this big data kind of problem that we've
2:56been tackling for the last decade. What
2:58do you see on the horizon for the next
3:00kind of handful of years around this
3:02category?
3:04Well, so I think until a few months ago,
3:07the conventional wisdom among data
3:10people was like I know AI is changing
3:13everything in software engineering and
3:15like it's kind of um you know this huge
3:18disruptive disruptive force that's
3:20coming in um but it's not going to
3:22affect us and here's all the reasons why
3:24it's not going to affect us and in fact
3:25I used to you know get on stages like
3:27this and say here's why you know uh text
3:30to SQL is never going to is never going
3:32to be great. here's why this, you know,
3:34what everybody's talking about is going
3:35to be harder than expected. And then,
3:37you know, the next generation of tools
3:39came out, the next generation of models
3:40came out in like December. And then all
3:42of a sudden it was like, well, actually,
3:44this just works. Um, and I think there
3:46was a couple of reasons that we were
3:47thinking about things wrong. But I think
3:49we all of a sudden went from like this
3:51isn't going to impact us to like, hey,
3:53this this just works. And then like
3:55within you know about 15 minutes of
3:57playing with clawed not even clawed code
4:01and you know the ability to just sort of
4:02ask questions about data I'm like hey
4:05wow when I ask it about my revenue and
4:06it it plots this graph like this is
4:09actually a better graph than like the BI
4:10tool that we use. It's actually it's
4:12better graph than any BI tool I've ever
4:13seen. Um and I can also do a lot more
4:16interesting things and I can like uh I
4:18don't have to like bang my head against
4:19the wall to figure out how to get the
4:21the the metric or the thing that I want.
4:22I just ask it. Um and it's like wow this
4:25is going to change this is going to
4:27change everything in in the data world.
4:29Um and so I think that the you know the
4:31modern data stack which is kind of
4:33typically aligned around a set of swim
4:36lanes you know you ingest your data you
4:38transform your data you uh you analyze
4:41your data and you visualize it. I think
4:44those are all sort of blurring. I mean
4:45you can see how like the you know
4:47snowflake data bricks uh etc are kind of
4:50like saying hey we're going to build a
4:51BI tool we're going to build the
4:53ingestion you know Fiverr and DBT are
4:55merging so there's been a lot of like
4:58coalescing of the of the of the market
5:00but I think that that even predates AI I
5:03think AI is just going to sort of
5:05rapidly accelerate it and you know and I
5:08think then it becomes like it's super
5:11hard to know what you know what what
5:13shakes out but I think as a you know as
5:14as a founder of a software company. Um,
5:17and you know, I think it's it's a hugely
5:20exciting time because because nobody
5:22knows what's going to happen. I think
5:23you know, you look at a year ago, you
5:25felt like, hey, I kind of know what's
5:26going to happen. you know, this is going
5:27to happen and this is going to happen
5:28and it's sort of a relatively sleepy
5:30part of the um you know, of of
5:35technology and now it's like oh my god
5:37like all this stuff is new because if
5:39you look at like you know data pipelines
5:41those are highly vibe codable like you
5:43can you know do you need you know Fiverr
5:45when you can just ask you know claude to
5:47say hey um you know uh pull data from
5:50the HubSpot API and like put it into my
5:53into my database. Um, so I think there's
5:56lots of opportunity, lots of things, you
5:58know, I think agents are going to kind
5:59of dramatically dramatically change how
6:01things work. Um, but it's, you know, I
6:04think times of great, uh, difficulty, I
6:08think, are also times of great
6:09opportunity because it's like, um, you
6:12get sort of pushed out of your comfort
6:14zone and then you you can recognize,
6:15well, hey, there are these adjacencies
6:17or there are these things that that I
6:18thought were would never work and now
6:20all of a sudden all of a sudden they
6:21work. Well, and you guys are obviously
6:24investing quite heavily in AI into the
6:27top of the warehouse for Mother Duck. As
6:30somebody who used to lead product
6:31exclusively, now you are leading this
6:33organization. How are you guys thinking
6:34about product strategy given all the
6:36uncertainty you just shared about like
6:38how do you navigate that as a leader and
6:40as a company?
6:42
>> Um, so one of our investors, uh, you
6:44know, Martin Casado, like he's kind of a
6:46a legendary infrastructure investor and
6:48he's, you know, he likes to say there's
6:50basically three phases of a company.
6:52There's the kind of the build phase
6:53where you kind of it's all about
6:55product. Uh then there's the go to
6:57market phase which is really all about
6:58like optimizing your go your sales and
7:00go to market and like really making sure
7:02that you can scale that. Then there's
7:04the scale phase where you just become a
7:05big company and you kind of try to um
7:07you know make it make it much bigger.
7:09And I think we had progressed from the
7:12kind of the build to the and we're
7:13solidly in the sales phase. And I think
7:16AI shows up and it basically says hey if
7:18we continue just sort of like saying how
7:20do we optimize what we're doing then we
7:22we're going to be toast and then we had
7:24to really go back into kind of product
7:27mode and build mode and and figure out
7:29okay what are the you know what are the
7:31opportunities and a lot of it was
7:32looking at sort of the adjacencies. Um,
7:35you know, I guess one way of, you know,
7:36a metaphor I like is, you know, there
7:38were like these houses on a cliff and
7:41like the the tide came and they washed
7:43away the house on the cliff, which was,
7:45I think, BI and uh and now all of a
7:48sudden we have we have a, you know,
7:50waterfront view, which is awesome,
7:52except like we're also the next house
7:54for the tide to come and and hit. So, we
7:56kind of have to, you know, you figure
7:57out what do you have to do to to not
7:59get, you know, uh not get washed away.
8:03Um but then also how can you take
8:05advantage of like of this this amazing
8:07waterfront waterfront view and you know
8:09specifically like you know we were able
8:11to build this technology
8:13in a month which we call dives which is
8:16you know you start from Claude and you
8:19you know Claude uses our MCP server um
8:22you ask a bunch of questions and then
8:25when it builds a visualization which
8:27it's quite good at then we just say hey
8:28can you save this in motherduck so what
8:30the the visualizations claude usually or
8:33or ChatGpt or Gemini usually does is is
8:36they have this drawback of the data is
8:38embedded so it's static and if you want
8:41to like have a revenue dashboard that
8:43you keep looking at well it's never
8:44going to change so you need you need to
8:46you need something that will update it
8:48and so what we did is we kind of built a
8:50tool that would replace the data with a
8:52live query and then host it on our
8:54website and this turned out to be super
8:57powerful um in being able to understand
9:00uh and and allow allow people who
9:04typically wouldn't be writing queries,
9:06wouldn't be interacting directly with
9:07the database to um to get to get just so
9:11many things done like our sales team,
9:13our head of c our customer success team,
9:15our support team, our engineers, like
9:17marketing, everybody is everybody is
9:19sort of building these things. And
9:21they're also kind of more than they're
9:23not just they're not just dashboards.
9:26they're actually full-blown full-blown
9:28data apps and you can skin them and
9:29they're they've been super interesting.
9:33So, I think there's, you know, some
9:35product adjacencies that we can that we
9:37can look at on kind of on that side on
9:39the ingestion side. Um, and then really,
9:42you know, the as a as a product person
9:44kind of the lens is like which is how do
9:46we make our our customers jobs our
9:48customers lives better? How do we make
9:50it kind of easier to do things that
9:52people already want to do? And then once
9:54you kind of focus from that level and
9:56from and even step back is like what is
9:58the thing that people are actually
9:59trying to do well people are trying to
10:01get insights about their their business
10:05and using data and so how do we how do
10:08we help help them do that better and I
10:10think that just sort of opens up a bunch
10:12of of new opportunities at the same time
10:14like we have to make sure that the other
10:16stuff that we've been doing the building
10:18an amazing data warehouse uh also stays
10:21you know that we that we keep that we
10:23don't lose focus of that and just like
10:25focus on the new shiny thing because
10:27that's also a a great recipe for uh you
10:30know for for falling on your face.
10:32
>> Well, that kind of transitions. you
10:34recently shared something you shared
10:36with your mother duck team in Slack kind
10:38of about how you think about leveraging
10:40AI internally as a company and maybe
10:44share with the audience a little bit
10:45about kind of what your perspective is
10:46on that and then how are you guiding
10:48your team to both find the efficiency
10:50and not give up the creativity that kind
10:54of comes with that like where are you
10:55asking people to find that balance
10:57
>> so on one hand like you know AI and is
11:02is a huge way to like help accelerate
11:04the stuff that you can do. But one of
11:06the things I was seeing is is a lot of
11:08people were sort of using AI as a
11:10substitute for actually thinking through
11:11a problem. And you know, people would
11:14write an AI generated doc and they'd
11:16send it and like and then you you're
11:18reading through like these eight pages
11:19of just sort of like AI generated
11:22generated slop and you'd ask them about
11:24something and they wouldn't know because
11:25they didn't like all they did was sort
11:26of like they had like a two paragraph or
11:28a one paragraph prompt and you know AI
11:30actually did the thing or you you'd kind
11:32of push back and like they wouldn't be
11:34able to push back because it because it
11:36wasn't actually their their thing. And
11:38so, um, you know, to me that's just, you
11:41know, it's laziness, uh, intellectual
11:44laziness. Uh, especially if you're
11:46writing, you know, if you're trying, if
11:47you're writing to communicate with
11:48somebody else, um, you know, one of the
11:51things that ways that I write personally
11:54is, uh, I write to just help me
11:56understand my thinking about something.
11:58And I find if I don't write it down, my
12:00thinking has all sorts of jumps and
12:03gaps. And you know, I'm like, well, we
12:05should do A and then A leads to B and
12:07then B leads to C. And then you start
12:09writing it down and like actually
12:10there's no way to get from B to C
12:12directly. And you have to like, well,
12:14maybe it's, you know, well, maybe you
12:15have to go through like B prime or, you
12:17know, maybe actually it's not C at all,
12:18it's D. Um, and and so that that process
12:23I think is is extremely valuable. And I
12:25and it's one that kind of I'm pushing
12:27our team to um to to do more and more.
12:31Uh, and it's not because I like writing
12:32and like I I hate writing. Like whenever
12:35I have to write something, I like I
12:37procrastinate and I you know I go and I
12:40you know read surf the web and go get
12:43gummy bears and go like whatever I can
12:45to avoid to avoid writing. But like
12:47because it actually thinking about
12:49something is hard and and and I think um
12:52you know that's one of the things that
12:54as more and more stuff becomes AIdriven
12:56and and as the sort of the cost of
12:58building goes to zero I think the one
13:00the one thing that is going to be
13:03important is like is clarity of thought
13:05and being able to sort of communicate
13:07clearly what you want or what the
13:08problem is or what the customer customer
13:11needs are. And um and I think that you
13:14know the only way to do that is is by
13:16practice. And plus it's just like you
13:17know if you send somebody an AI
13:19generated thing it's like from three
13:21bullet points like you might as well
13:22have just sent them to three bullet
13:23points because like because that's
13:25actually the information content that
13:26you've you've um that you've shared. Um
13:29at the same time you know like for
13:32building for building things AI has
13:33enabled us to to to move a lot faster.
13:36So, I don't want to completely
13:39completely denigrate it, but I just
13:40think that there's a time and a place
13:42and a and a and a way to to use AI and
13:44and um and you know, and AI can be a
13:48good can be an editor, can you know, you
13:51know, can poke holes in your ideas. You
13:53can write something and, you know, have
13:54AI point out where you went from A to C
13:57without
13:59like having a clear a clear path there.
14:01But, you know, just having AI design
14:04something for you or write a, you know,
14:06write a doc, I think is laziness.
14:10
>> Well said. As somebody who works in a
14:11large corporation, I think that I could
14:13I could I share your sentiment. Um given
14:17your current role at mother duck as well
14:20as all of your experience at Google, how
14:23do you think about preparing now your
14:25data warehouse against all of kind of
14:28the um agentic queries that are going to
14:30start to come at it and how do you think
14:32about scale and all of those different
14:34things that are going to be a new world
14:35order given what we're also creating. So
14:38there was a an interesting interview I
14:40saw the other day um it kind of went
14:42around the you know Twitter sphere um
14:45with had had James or Jeff Dean who was
14:47you know one of the you know architect
14:50wizards at uh at Google talking about
14:53you know the the impact some of the
14:56impacts of like of you know agent-driven
14:58computation and one of the things that
15:00he you know he said was like all of
15:02these tools that we're using like the C
15:04compiler was was an example or designed
15:06in an for an age where like it was a
15:08human that was driving it. And so if a
15:10human is typing, you know, um is typing
15:12make, you know, they uh if it takes the
15:16C compiler a second or two to start up,
15:18it's not a it's generally not a big
15:19deal. Um but an agent can can run things
15:24so much faster than a human. So if it
15:25takes two seconds to start up, then
15:27that's actually that's actually a
15:29problem. And that's, you know, a big
15:31limiting factor in how much you can
15:32actually get done. Um so for for data
15:36and analytics and dealing with your data
15:38um for a long time there's been like
15:41this sort of truism is like well latency
15:43doesn't really matter. It's throughput
15:45that matters like hey I'm trying to
15:47crunch this giant thing and like the
15:49fact that I can do that in like a few
15:51seconds versus a few minutes is
15:52important. Um but if you know when
15:55things are being driven by agents and
15:56agents are possible are are are um are
15:59able to run a 100 queries per second um
16:05then but if it takes 5 seconds or 500
16:08milliseconds even to run that query then
16:10you're really going to slow down the
16:11agent. So um a data warehouse or an
16:14analytics engine that can run you know
16:17in run your queries in 10 milliseconds
16:20is going to be you know is going to be
16:22the one that that that works. And one of
16:24the nice things is we've sort of have
16:25built on top of of ductb and we've made
16:27a bunch of choices that that mean that
16:29we can do super low latency. Like I
16:30think our median query is six
16:32milliseconds our you know 80th
16:35percentile
16:36is is under 20 milliseconds. Uh and so
16:40um you know we I think we're really in
16:42really good in really good shape for
16:44this sort of agentic world and like kind
16:46of a architecture also means that every
16:48agent or every end user gets or every
16:50workload gets their own separate
16:52instance so they're all isolated and so
16:54you can have lots of agents all
16:55hammering things in uh in parallel but I
16:58think you know kind of the the the the
17:01shape of the problems are changing with
17:03with agents and I think the the shape of
17:06the systems that have been built over
17:08the last, you know, couple of decades
17:10are designed for the shape of the
17:12problems with with humans. And I think
17:14uh I think some of that's going to is
17:16going to change. And I think it's likely
17:17going to change uh change very quickly.
17:19And you know, I think um there's kind of
17:21a truism in in in software that like
17:24yes, there's things things change, but
17:26they change a lot slower than you
17:27expect. And I think you kind of throw
17:29all that out the window when you're
17:30dealing with with AI is like some of
17:31these things are changing just so much
17:34faster than I ever I ever ever would
17:35have thought. like something, you know,
17:37whereas just like four months ago, it
17:39was almost unthinkable that like, you
17:41know, you could just ask questions to
17:43your to your database. Like I think
17:45internally at Motherduck, I don't think
17:47any of us ever write SQL anymore. Like I
17:49used to be our number one SQL writer and
17:51now like and now I just ask ask Claude.
17:54Uh and it's um you know, so things
17:57things change things are changing
17:58quickly and um and that's what's that's
18:02what's exciting about it.
18:03
>> Um I have one more question. So just
18:05prepping you all so you can start to
18:07prepare your questions for Jordan. Um
18:09and the final question I had for you is
18:11as you think can you kind of walk
18:13through the trade-offs in your mind
18:14between data lakeous and data warehouses
18:17in the world of AI and maybe for the
18:18nontechnical people in the audience kind
18:20of help break those two apart as in that
18:22answer.
18:23
>> Yeah. So I think you know tra
18:24traditional data warehouses or
18:26traditional way people store their data
18:27is is in the database. Uh, and I and I
18:30think um, which is great, but then your
18:33database owns it. And and one of the
18:36things that people are wanting to do in
18:37the age of, you know, with AI or even
18:39even without AI is they want to say,
18:41hey, well, I've got a bunch of different
18:43tools. I want them all to be able to
18:45access my data. And, um, and if it's
18:48locked up in my Snowflake instance or my
18:49Redshift instance or BigQuery, uh, then
18:52that doesn't that doesn't work. And so
18:53there's been the sort of this, you know,
18:56flowering of open data formats, iceberg,
18:59delta, duck lake that are kind of making
19:02it easier to store your data elsewhere.
19:05And then you have a, you know, a
19:06plurality of tools that actually can can
19:08access uh access that data. And I think
19:11as as you know, you're going to see more
19:13agentic tools so that your agent can
19:16basically use whatever engine it wants
19:18or whatever the the the best the best
19:20way to access the data data is. Uh I
19:23think we're going to instead of having
19:24these things locked up in warehouses,
19:26they're going to be um basically
19:29federated out or they're basically
19:30centralized in um you know there are
19:34just you know parquet files and some
19:36metadata uh that live somewhere on a uh
19:39somewhere on a data lake. Um, but it
19:41also gives it gives other tools access.
19:43So like if your data is locked up in
19:45Snowflake and you're like I want to use
19:47Mother Docker, I want to use BigQuery, I
19:48want to use something else, then it's
19:50it's this big migration project versus
19:52if it's an iceberg, then you know we
19:55have the same, you know, all these tools
19:57have the same the same level of access.
19:59Uh, and I think it's also going to
20:00cause, you know, a flowering of
20:02additional tools and additional ways to
20:04uh to look at look at your data.
20:08opening it up to questions for Jordan.
20:17
>> Hi. So uh for so to me obviously uh the
20:20duck DB it's a huge momentum for Maduk
20:24uh but for for you uh so over time do
20:27you see that there's a line that you
20:29should be cautious about uh if like how
20:31do you define the boundary between um
20:33duct DB being more popular and also
20:37mother ducks still remain uniquely
20:39valuable.
20:40
>> Yeah. So, I mean, that's a good
20:41question, and I think it's something
20:43that pretty much every every kind of
20:47open-source
20:48commercial version of open source uh um
20:52wrestles with and and I think, you know,
20:55to us is like there's value in in having
20:57somebody run infrastructure for you. I
20:59mean, it's the same reason that, you
21:01know, um uh you know, running Postgress,
21:04you know, RDS, Aurora, Neon, uh etc. is
21:08a is a multi-billion dollar business is
21:11like yeah I could I could run Postgress
21:13on my own it's this great open source
21:15project but like it's also a lot easier
21:16if somebody else somebody else runs it
21:18for you and I think particularly with
21:20analytics um analytics is kind of
21:23capable of of using up really whatever
21:26resources you throw at it and then and
21:28in doing so very fast and um and so like
21:33you can either scale you know if you can
21:36run DB on your own you can run it on a
21:37giant machine to be able to run things
21:40at at speed. But that also means you
21:42have to you have to provision a giant a
21:44giant machine. Um and you and analytics
21:47also tends to be very spiky. Or you can
21:49run it on something like motherduck
21:51where you know you basic you know
21:53basically can do sort of uh you know
21:55real time real time provisioning of much
21:57you know much larger resources than
21:59you'd actually do. And I think
22:00particularly for for running agents like
22:02agents are um probably not going to have
22:06tons of compute on their own and tons of
22:08tons of memory on their own and you
22:09might want to might not want to devote
22:10that memory to you know buffer buffer
22:13pools for your analytics. Um so it's
22:15just something that's becomes really
22:16easy uh and you know makes a lot of
22:18sense to to offload that to a to another
22:21service and once you offload it to a
22:22service then you know we can run we can
22:24run that really well. But that's a good
22:26question. Thank you.
22:28uh kind of a followup to that question.
22:30Uh so you you kind of mentioned you know
22:32there is value in managed infrastructure
22:34and I think you know larger enterprises
22:36are willing to pay for that ease and
22:40user friendliness but I'm curious like
22:42from mother duck's perspective is there
22:43any kind of tipping point that you're
22:46seeing as you know customers are moving
22:48from duck DB to mother ducks uh various
22:51products and whether that's monitored
22:53internally like from a threshold
22:55perspective of how customers engage with
22:57you guys.
22:58
>> Um I mean most of our Most of our
23:01customers are actually smaller you know
23:02smaller businesses like we um you know
23:05we think we can provide we can provide a
23:07lot of value for um for companies that
23:09just you know they want they want some
23:12analytics and they want somebody to run
23:13their you know they want a data
23:15warehouse ductb is not on it not on its
23:17own a data warehouse that has no concept
23:19of users or permissions um it's very
23:22hard to sort of share things between
23:23customers and yeah you can kind of build
23:25some of the frameworks around it and you
23:26can you know run a you know caching
23:29caching servers etc. Um but you know
23:33most of our customers you know just are
23:36like hey it's easy it's easier to run
23:38have somebody else somebody else run it.
23:40um you know if you're running things on
23:42your laptop then that's you know you you
23:45can run duct DB on your laptop but like
23:47very often you don't want to download
23:48all your data onto your laptop and sort
23:50of and and do all that work and if
23:52you're going to be hitting hitting data
23:54that lives in the cloud um it's just
23:55much easier to use a uh to use a cloud
23:58service and it's almost you know um
24:00irregardless of the uh of of the sizes
24:03and especially as as you know things uh
24:06things start to scale
24:15Hello. I haven't fully formul formulated
24:18this question, so I hope it comes out
24:20correctly. Um, as someone who used to
24:24write a lot of SQL who no longer writes
24:27any, do you, if you had a crystal ball,
24:31could you talk about how
24:35um the goodness of code is going to
24:39either crash where no programmers write
24:43code or is is it just going to continue
24:46to evolve? Like it just seems like it's
24:47happening so fast. What happens when we
24:50no longer code?
24:53
>> Um I mean that's a fantastic you know
24:55philosophical question but um I think
24:58you know throughout the history of of
25:01computers there's been um a progression
25:04of higher levels of abstraction you know
25:07I think you know first you know people
25:08would write on punch cards you know it
25:10was like um this was you know very very
25:13little abstraction you're kind of
25:14dealing with specific registers and
25:16memory etc. um you know there was
25:18assembly there was really low-level
25:20programming languages like you know
25:22forran um and and cobalt uh and then you
25:26kind of the you know see to then the man
25:29manage languages and kind of like there
25:32is um I think a desire to add add more
25:36levels of of abstraction just so that
25:38people can think about what is the what
25:40is important for their for the problem
25:42they're trying to solve versus like how
25:45do I teach the computer how to do kind
25:46of every every little thing. And I think
25:48AI is ushering in kind of the next step
25:51where I really I really think actually
25:52the the the
25:55end point is where source code is just
25:58English um or whatever you know language
26:01you uh you you speak. It might be some
26:05sort of stylized version of English. But
26:07um you know I think that's one of the
26:09interesting things about I don't know if
26:10you've seen the kind of the open claw
26:12project. Uh and you know the way to add
26:15plugins to that you don't add it in code
26:18you add it in English descriptions of
26:19the thing that you want to add. And um
26:22to me that's just sort of fascinating
26:23and it feels like that's the that's the
26:25future to me. Um, you know, I I was a
26:29software engineer writing, you know,
26:30hands- on keyboard writing code for
26:31probably 15 years. And I love writing
26:33code. And ever, you know, every every
26:36day that I'm not writing code, I sort of
26:38wish that I was because like the the um
26:42the boost you get from like, hey, this
26:44works. This is awesome. Like, you know,
26:46and like it's just like this this
26:48feedback loop is is great. And like
26:50people are so much harder to deal with
26:51and understand than than code. Um, you
26:55know, I think for a certain type of
26:56certain type of introvert, maybe not
26:58everybody is is the same, but um, so I
27:01do I do think that like, you know,
27:03probably programming, you know, just
27:06like there are still assembly language
27:07programmers, like I think that, you
27:09know, I think there's still going to be
27:10a market for C programmers, but but I
27:13think it's probably not going to be uh
27:14not going to be as many. Um, is that
27:17sad? Yes, it is. But it's also means
27:19that like a lot more people are going to
27:22be able to write code and get stuff done
27:24and build things and get the same like
27:26the same feedback of like hey I built
27:29this this is cool I want to do this and
27:31like and so I I think that that value uh
27:35outweighs kind of the you know like hey
27:37you know like I'm not going to be able
27:39to wrestle you know with uh in-memory
27:41data structures in the same way.
27:44
>> We're going to end it there. Thank you
27:46all so much. Um Jordan's here for a
27:48little while longer. The mother duck
27:49team is in some great shirts if you
27:51actually have questions too and want to
27:53continue this conversation. So, thank
27:54you Jordan.
