YT YouTube2Text

YouTube2Text → Transcripts → The Future of the Open Internet Starts at the Impression

The Future of the Open Internet Starts at the Impression — Transcript

by ExchangeWireTV · 4,525 words · 689 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:01[music]
0:06>> Good morning.
0:06>> Good morning. Yes, yes.
0:08Great to be back in Singapore.
0:10Um
0:11I think Andrew, you were scheduled to
0:12speak here 6 years ago. Is that right?
0:15>> 6 years in the making.
0:17Uh I guess 2020. Uh and then something
0:19happened. I can't remember, but a lot of
0:21flights were canceled.
0:22>> there was a little flu thingy.
0:24Man flu.
0:26>> [laughter]
0:27>> And so
0:28for people who don't know you, I'm sure
0:30you do, but please introduce yourself
0:34um and say a little bit about who you
0:36are and a little bit about the company
0:37as well.
0:38>> Yeah, well, super happy to be here. Uh
0:40good morning, everybody. Uh I'm Andrew
0:42Casale. The The best way for me to frame
0:44myself is I kind of affectionately refer
0:46to myself as an ad tech lifer.
0:48Uh all I know is ad tech. Uh I was the
0:51the kid in the '90s making websites.
0:53Some of them got popular and ad tech
0:55found me and I fell down the rabbit hole
0:58and never left. Um but I had the great
1:00pleasure of building the company that is
1:01Index Exchange. Uh we sit on the sell
1:04side of the market, uh which means that
1:06we represent the interests of media
1:08owners globally.
1:09Um we're also really proud to have been
1:11uh here with boots on the ground in
1:13Singapore for the last several years. Um
1:15but we're also aggressively growing
1:17right now. So, there's a few open recs.
1:19If you know anybody, uh and we hope to
1:21be around seven uh by this time uh I
1:23should say by the fall. So, aggressively
1:26growing right now in market, too.
1:27Um and uh yeah, sell side.
1:30>> Right. We've been doing this for 16
1:32years and uh
1:34we've been talking about programmatic
1:35since then. I mean, the very first event
1:37had uh
1:3850 people at it and uh it's grown to
1:411,100 people. So, I want to talk to you
1:44about, you know, where we where we've
1:46been and where we are now, cuz we're at
1:48this really interesting sort of um
1:50interesting sort of point in the
1:52programmatic uh timeline. And I want to
1:55talk about a little bit about where we
1:56are and cuz particularly as it pertains
1:59to this this part of the world.
2:01So let's talk about that and talk about
2:03where we are right now and and the kind
2:05of stuff that you guys are doing.
2:07>> Yeah, like there's there's a broader
2:10trend that's underway right now that we
2:12refer to with the term sell side
2:15decisioning
2:16but it's the premise of saying a lot of
2:19our market in programmatic has been
2:21built historically at great distance
2:24from the actual impression opportunity
2:26itself
2:27and so if you think of the way we've
2:30created programmatic with protocols and
2:32standards we send lots of requests all
2:35over the internet and we wait for DSPs
2:38to make decisions and send responses
2:40back
2:41and that's got us here as a market but
2:43at the same time the overall scale of
2:46the internet has never stopped growing
2:48we started on the web we obviously have
2:50mobile
2:51and increasingly CTV is is taking over
2:54so much of programmatic and so as it is
2:56growing the complexity has also swelled
2:59dramatically and that old construct of
3:02doing everything at great distance has
3:04turned out to be really really
3:05expensive. There's an inversion event in
3:08the market happening right now
3:10which is starting to localize value
3:13creation closer to the impression so
3:16closer to the origin of ultimately
3:18what's going to drive an outcome because
3:20what we've recognized is that on one
3:22hand
3:23if you do more next to the impression
3:25you have better proximity to signal you
3:28have a far more efficient transactional
3:30environment and you can also
3:32dramatically drive down the cost of
3:34compute which you know in this age of AI
3:37is also at record highs
3:40and it's giving us an opportunity to
3:41kind of rethink and challenge the way we
3:43built this market.
3:44>> Yeah, cuz like in in this market
3:47the walled gardens are pretty entrenched
3:50Um, I had a conversation with a
3:52a unknown person last night about how,
3:54you know, the data they had they wanted
3:57to activate it in programmatic and the
3:58agency said, "Ah,
4:00let's just use Meta.
4:02It's just easier." Uh,
4:04uh,
4:05but there is a broader conversation
4:07around, you know, walled gardens versus
4:10the open internet, right? Um, everybody
4:12likes to take pot shots at the open
4:14internet because, you know, uh,
4:16cost, right? Uh, the fragmentation.
4:19But actually,
4:21what's happening right now is that the
4:23open internet is starting to become uh,
4:26competitive
4:28on not just the cost, but on performance
4:30now because of all the innovation that's
4:32happening right now. So, let's talk a
4:33little bit about that. I'm, you know,
4:36cuz people would like to spend more
4:37money away from Meta and Google. They
4:39would really like to do it. I know every
4:40day I talk to people said, "All we're
4:42doing is retargeting our own users. I'm
4:44sick of giving money to these two
4:45companies."
4:46Same standards, closed walled gardens,
4:48etc. But let's just talk about how you
4:51see sort of the open internet fighting
4:54back or being able to be competitive
4:57with those walled gardens.
4:58>> Yeah, I would say we're relatively early
5:00into this and it's part of this
5:02inversion in value uh, event that's
5:04underway, but maybe to frame things uh,
5:08in like a slightly different way. Um,
5:11you know, I think
5:12it's often uh, a common refrain like if
5:14you look at the walled gardens and you
5:15look at their growth and you look at how
5:17much ad spend they continue to capture,
5:19there's typically like a narrative
5:20that's like, "Well, that's true, but
5:22they grade their own homework and if
5:23they didn't maybe the the result would
5:25be different." And I think that's that's
5:26actually a failing narrative. I think
5:28the fact of the matter is they've
5:30created an incredible machine that does
5:32drive outcomes at scale. Works really
5:34well, it's really easy to use. And if we
5:36were to ever do anything about that, we
5:38have to match them in kind. Um, when I
5:41think of the open internet, on one hand
5:42what's incredible about it is that it's
5:44inherently democratized and there are no
5:46walls, which is wonderful.
5:48Um when I think of the walled gardens,
5:50um I also think of something that's very
5:52easy to buy. And unfortunately, when I
5:54think of the open internet, I think of
5:55something that today is very exhausting
5:57to buy. And so to some degree, while we
5:59don't have walls,
6:01um our garden is almost like a messy
6:04garden today.
6:05Uh anyone can freely access it, you can
6:08traverse it, um you know, you can be
6:10empowered to use any platform you want,
6:14um but nothing is orchestrated at all.
6:15It's a little bit manic. Whereas inside
6:17the walled gardens, it's quite lovely.
6:19It's a closed ecosystem. It's very
6:20simple. I think the next trick for the
6:23open internet is to keep the advantage
6:25we have, which is there are no walls, um
6:28but to bring orchestration into the
6:30market, um so that it's not messy, it's
6:32not exhausting, and it can actually
6:34drive outcomes. And what's exciting is
6:36this is now underway. I think the advent
6:39of the portability of models, uh which
6:42is coming out of optimization vendors or
6:43the custom bidding algo uh category is a
6:46pretty exciting event coupled with the
6:48inversion of value to the sell side
6:50because we're getting to a place where
6:51we can actually start to match outcome
6:54performance from the walled gardens. If
6:56I look at How is so? Well, if I look at
6:58something like Google or or Meta,
7:00um you know, their their products,
7:02whether it be PMAX or Advantage Plus,
7:05um on one hand, a marketer will be like,
7:07"Yeah, it's a bit of a black box." But
7:08on the other hand, I often hear, "But it
7:10works. It drives great outcomes." Well,
7:12we're starting to see the same thing
7:14happen where if you use the collective
7:16scale of the internet, if you use all
7:18the signal that we have that
7:20historically has been caught in a bunch
7:21of fiefdoms everywhere,
7:23um and you reinforce that with the same
7:26mechanisms that Meta uses, you know,
7:28Meta is trained on the sales and
7:30conversions events of all of their
7:32customers' activity instead of just
7:34pixels on the internet tied to cookies
7:35that nobody can even trace anymore. If
7:37you start to bring that tech in, um you
7:40actually start to drive outcomes. And
7:42we're seeing it. There's a growing
7:43number of case studies that are
7:44connected to um, portable models that
7:47are being driven from the cell side uh,
7:49with Max Signal. It's early days, but
7:51it's really encouraging. I think it's
7:52going to get us to a place where we're a
7:54lot more organized as an open internet,
7:57even though we're still uh, without
7:59walls.
8:00>> Let's talk a little bit about um,
8:02containerization, right? Because
8:04everybody's talking about it. Uh,
8:07uh, well, everybody's talking about AI.
8:08And then everybody's talking about
8:09containerization. But let's talk about
8:11what cuz it's it's
8:13Let's take it from, you know, from the
8:14abstract and talk about it from a
8:16application point of view, from a
8:17day-to-day point of view, from a
8:19marketer agency point of view, or
8:21publisher point of view.
8:22What Give us just an overview of what
8:25the concept of containerization is and
8:27why it's important for this narrative
8:30around open internet competing on cost,
8:33competing on performance versus walled
8:35gardens.
8:37>> Yeah, like I can give you the story of
8:39containerization cuz it started at Index
8:41about 3 years ago. Um, and it was really
8:44intended to solve a a problem that was
8:47cost prohibitive to bring use cases into
8:50the market that we thought would be
8:52really important. Um, and so I'll try to
8:55be I'll try to avoid being overly
8:56technical. Um, and so I'm going to
8:58oversimplify it. Um, but
9:00containerization in effect is the idea
9:03of saying that if you run code locally,
9:06um, it's hyper-efficient to compute. And
9:08so the way to think about this is
9:10historically being, you know, back to
9:12the kind of my commentary earlier, ad
9:14tech and programmatic has been built at
9:16these great distances where um, if you
9:19take just the interplay between an
9:20exchange and a DSP, um, we have our own
9:23infrastructure, DSPs have their own
9:25infrastructure. Often that
9:27infrastructure is in the public cloud on
9:29GCP or AWS. And so for us to
9:31communicate, we have to send a lot of
9:33requests out and they have to send a lot
9:35of responses back in. Um, so these are
9:38two kind of infrastructures uh,
9:39interoperating. Unfortunately, it's
9:41really expensive to do that. And to do
9:43that at internet scale is even more
9:45expensive. Um, containerization is the
9:47idea of saying, "Why are we doing that?
9:49Why are we making all these calls
9:51outbound to only get all these responses
9:53back inbound? Instead, can we just run
9:55their value or their code right at the
9:58exchange?" Um, and this is mimicking um
10:01a little bit of parallels from the
10:03financial markets as well. If you've
10:04ever read Flash Boys, you're
10:06>> I love that book. Michael Lewis, great
10:08book.
10:09>> You're well aware that there is such an
10:11arbitrage advantage in finance to
10:14localization. If you apply it to ad
10:17tech, it's not about creating arbitrage
10:18advantage. It's actually about cutting
10:20an unnecessary cost, which is the public
10:22clouds, to be perfectly frank. They've
10:24been incredible at being able to spawn
10:27tech businesses overnight, but at the
10:29same time, they're a huge convenience
10:31tax on the market.
10:33And we don't need them anymore. And so,
10:35containerization is just the idea of
10:36saying, "Instead of sending a request
10:38outbound to code that sits in AWS, send
10:40the request right locally on the
10:42exchange and have the code run there."
10:44And when you do that, it's really cheap
10:47to process.
10:48The reason why this is a significant
10:50event and the reason why this is now
10:52taking on new forms is we started this
10:55with custom bidding algorithms. Um,
10:57because at the time, a lot of the
10:59companies in the space were startups.
11:01And for them to be able to make really
11:03smart decisions to drive outcomes, they
11:05needed to see everything. And the thing
11:07was, when we started to think about
11:08these designs with these platforms, to
11:10send them everything, to be perfectly
11:11frank, would bankrupt these companies.
11:13If I were to send the firehose of index
11:15to an AWS endpoint for a custom bidding
11:18algo, probably have to spend a million
11:20dollars a day to listen. Like, it would
11:22just be completely cost prohibitive. But
11:24today I can do this because their code
11:26runs on top of Index. That's what got
11:28the party started. Um, then we started
11:30to look at other use cases that were
11:32also very cost prohibitive. We moved
11:34into data, which is also very expensive
11:36to run in the public cloud, and then
11:37most recently uh the advent of
11:39containerization for DSP bidders. Um and
11:42it's starting to generate a lot of
11:43excitement because we're actually
11:45starting to rethink the way we built
11:46this whole market. Um in a way, we have
11:49this invisible tax. Like people refer to
11:51ad tech and programmatic as a tax. What
11:53they don't recognize is that we have
11:54this invisible tax on the market, which
11:56is the public clouds are really
11:57expensive and they power almost
11:59everything and we don't need it anymore.
12:01>> So, let's just say that you have all of
12:04that hosted inside the Index instance uh
12:08cloud instance. Uh you're cutting out
12:11cost and latency.
12:13I want to talk about timing cuz that's
12:15interesting as well, right? I mean, we
12:17talked about this as an X innovation.
12:19What does that mean when you've got all
12:21that time? Cuz you know, the the the the
12:22the the tests you've done with with with
12:25on the on the campaigns you run with
12:26Bedrock and Index are showing that
12:28there's incredible you know, time to do
12:31other things within that sort of uh you
12:34know, within that time frame. Um and I'm
12:37I'm quite interested to hear about that
12:38because
12:39you talked about like you talk about how
12:42we compete with, you know, the big
12:44walled gardens. That's kind of where we
12:47start to kind of like crank it, right?
12:49That's where you start really competing
12:51and doing some incredible things beyond
12:53what they can do.
12:54>> Yeah, it's it's really hard to
12:55appreciate this um but in tech um time
13:00is a cost. Light speed is a cost.
13:02Distance is a cost. And so, um if you
13:05look at the way we've operated this
13:06market historically, um we run these
13:08auctions that drive programmatic in
13:10about 200 milliseconds. Sometimes we
13:12have a little bit more time. Sometimes
13:14we have a little less time. Most of that
13:16time has just been lost in that transit
13:19time that I mentioned earlier um going
13:21out and coming back in. Um with
13:24containerization, we're able to now
13:26repurpose almost all that time. When we
13:29call a model locally, it typically
13:31responds in 5 milliseconds. When we call
13:34a data provider locally, it responds in
13:35about 1 milliseconds. And now that we
13:38have the beginnings of containerized
13:39bidders, we're not publicly revealing
13:42how fast the first containerized bidder
13:45is, but it's remarkable. And I think
13:48some case studies will be dropping over
13:49the summer to communicate this more.
13:52But imagine almost two orders of
13:54magnitude faster in bid speed. So then
13:56you end up in an environment where
13:58instead of losing about 200 milliseconds
14:00in just transit time, we've gained back
14:03190. So what are we going to do with
14:05that time? Well, it all comes back to
14:07performance in the walled gardens. If
14:10anyone in this room has ever worked at
14:11Google or Meta or Amazon and has been
14:15close to any of these wonderful
14:16performance products, you've probably
14:18heard that there's an obsession inside
14:20of the cultures about latency
14:23and that there's a direct correlation
14:25between how long they can let the
14:27algorithm wait and its ability to drive
14:30better outcomes.
14:31And literally milliseconds matter inside
14:34the walled gardens. We've been burning
14:36all of our milliseconds just
14:37communicating over the internet. So what
14:39starts to happen next?
14:41Well, if I look at a growing number of
14:43transactions on Index,
14:45we say today approximately a third of
14:47Index is sell-side decisioned, which
14:49means some signal or some container on
14:52the sell side directly influenced the
14:54transaction that we cleared.
14:56Today what often happens is requests
14:58will
15:00look at a model. A model might make a
15:01decision, communicate that decision
15:04downstream to a DSP by changing the bid
15:06request. This can now happen so fast
15:10that we're getting to a place where
15:12models are potentially going to be able
15:14to communicate with each other. And like
15:16before this gets too weird and like
15:17agentic hype, I'll give you a very
15:19practical example of the first use case
15:22that we're focused on. Uh we call this
15:24and I didn't even coin this phrase. This
15:26came from one of the agencies that we're
15:27working with. It's solving one of the
15:30impossible problems in this market. And
15:32an impossible problem in this market
15:33that directly affects yield and
15:35performance is the fleeting opportunity
15:38that gets a no bid because there's a
15:40lack of information. So, what do I mean
15:42by that? Um take CTV. CTV is such an
15:45incredible channel. It's about half of
15:48Index now. It's growing like a rocket in
15:50this region as well for Index. It's
15:52growing at about 100% year-on-year. Um
15:55it's just swallowing so much of the
15:56market's growth. But, CTV has a huge
15:59metadata problem. Um while we have a
16:01bundle in CTV, so we can know that maybe
16:03the consumer is watching a fast app, we
16:06don't always have the metadata to know
16:07the show or the episode that a user
16:09might be watching or any of the other
16:11associated attributes. What
16:12unfortunately ends up happening in CTV
16:14more often than not is a request will go
16:16out to bid and we'll get a no bid back
16:18because there's a lack of metadata. And
16:20it's a lost opportunity.
16:22Uh what we're now starting to see is if
16:24we take um data providers like take uh
16:27Nielsen's Gracenote, um based in the US,
16:30but they have this incredible data set
16:32of schedules from live broadcasts all
16:34over the world. I think they also have a
16:35few patents. They literally know what
16:38you're watching even if there's no
16:39metadata. Now, imagine in a world where
16:43we first have the request come into a
16:45model and the model says, "I can't buy
16:47this today because there's a lack of
16:49information." But, instead of passing,
16:51it can now ask a question. And a model
16:53can say, "Hey Gracenote, what do you
16:55know?" And Gracenote might come back and
16:56say, "You know what? This is the World
16:57Cup. Messi's playing." Um and the model
17:00might now say, "I like this. I'm now
17:02going to run a campaign." Now, we go
17:04from a no bid to a question and a bid.
17:08This is how we start to fix this entire
17:10channel because there's just so many
17:13yield opportunities that are squandered
17:14because of this race over time and and
17:17these lost signals that we're going to
17:19start to gain back. So, in short, we're
17:21pretty excited with what we're going to
17:22be able to do to repurpose all this time
17:24that we've now found.
17:25>> Nice. So,
17:27everybody's trying to strip away budget
17:29for the big players in this market,
17:30right? It's going to be meta, it's going
17:31to be YouTube. So, how would you
17:34position this for them to go to agency
17:37brands and say, "This is fundamentally
17:39going to shift,
17:41you know, what we can do in
17:42programmatic." So, like, what would your
17:44sort of advice be to to to to your
17:46clients, particularly the ones who are
17:48doing curation or data or
17:50campaign-driven stuff? So, what would
17:52you say to them?
17:53>> Well, what's really exciting is that on
17:54the sell side, we historically have had
17:57to do that, which is come up with
17:59innovations and then call agencies or
18:02brands and try to convince them that
18:03they're good ideas. Um what's happening
18:05right now is totally different. Uh while
18:08all this innovation is happening on the
18:09sell side, it's being entirely driven by
18:12the buy side. Um and so, like, one of
18:14the coolest whiteboards that I was in
18:16just 2 weeks ago in New York,
18:19uh was with Major Holdco.
18:21Um they had uh data science people in
18:23the room. Um and we were literally
18:25talking about what we're going to do
18:27with the 192 milliseconds that we've
18:29gained back and their prioritization
18:32of questions that they want their models
18:34to start to answer. Um and so, I guess
18:36my my answer to my answer to your
18:37question is we're we're going from um
18:40trying to pitch this as like a new
18:42opportunity to collaborating on a future
18:44for this market that is solving real
18:47problems, and everybody wants the same
18:49prize, which is to drive better outcomes
18:51from the open internet. I think
18:52everybody realizes that the internet is
18:54enormous. We have more scale than the
18:56walled gardens. Y- yet, we've done such
18:58a terrible job making it easy to drive
19:01performance in an orchestrated way that
19:03we're failing miserably. Um everyone is
19:06invested in solving this problem, and
19:08what's really cool about this particular
19:09trend is we have the buy side driving
19:12it. And and that's like the the there's
19:14like a stateless element to this. We
19:15call it sell side decisioning, but it's
19:17not about the sell side. Uh it's just
19:19about fixing the way we built this
19:20market.
19:22>> Very good. And one last question around
19:25agentic, right? It's like the term that
19:27is just running around the industry like
19:32a lad with a shirt off shouting and
19:33roaring, drinking cans of beer, you
19:36know,
19:37like a typical England fan really. Uh
19:39>> [laughter]
19:40>> Sorry, I had to get that shaming. I'm an
19:42Irish man, obviously.
19:44Uh soccer hooligans.
19:45Um so,
19:47let's talk about how you see that
19:49affecting the market because everybody
19:51says it's going to revolutionize stuff
19:53and there's lots of talk and, you know,
19:56there's Brian O'Kelly and all the rest
19:58of them trying to say we're going to
19:59rebuild the the whole thing. What's
20:01what's your take on it? And where do you
20:03think that the human plays a part in
20:05that in that process?
20:07>> Like I I think AI is going to
20:10revolutionize the market and it already
20:12is and it is going to increasingly. Um
20:15when we think at Index about agentic, uh
20:18we think about it in a very basic way,
20:21which is
20:22um we are creating uh the or we are on
20:25the advent of creating agents that can
20:27do work. Um but in your brain, cuz I
20:30think agentic is this wonderful world
20:33word that is like a black uh hole that
20:35just swallows everything and feeds this
20:38narrative of a hype revolution. Um
20:41an agent is a replacement for a human.
20:44Um and so, when we think of what agentic
20:47will do long term, it's work that humans
20:50don't want to do or shouldn't do or
20:52monotonous work uh that we can kind of
20:54hand off over time, um which is going to
20:57be very valuable. But when we think of
21:00what the frontier is on fixing the open
21:02internet and driving performance, um we
21:05think it's making really smart decisions
21:07really fast. Index bursts at 21 million
21:10requests a second. Unfortunately, if you
21:12try to spawn 21 million agents per
21:14second to make those decisions, if
21:17Anthropic's market cap today is going to
21:19be a trillion, you'll probably make it 2
21:21trillion just with that one use case.
21:23It's completely cost-prohibitive. It's
21:25even worse than
21:27>> busts the whole electric grid.
21:28>> It's yeah.
21:30It's worse than the crypto bros were
21:31were pitching at tech like, you know, 8
21:33years ago on moving everything to the
21:35blockchain.
21:36We think agents are going to have a
21:38great application for removing or or or
21:41or assisting human work. Uh but we think
21:43models do a far better job making really
21:45smart decisions in real time, and that's
21:48where we're focusing all of our effort
21:49right now. And we think that is um what
21:52AI is going to do to this market, not
21:53what Adgenta is going to do to this
21:55market.
21:56>> Particularly with the the conversation
21:58you had about models talking to each
21:59other, it it could play a role in that
22:01sort of that's that time left over from,
22:04you know, the the the conversation you
22:05had with the agency in New York.
22:07>> Yeah, like it the the next frontier for
22:09modeling, if you've heard the term and
22:10maybe you haven't yet, um it's all
22:12moving to vectorized embeddings now. And
22:15so, what they are is n-dimensional
22:17spaces. So, if we think of the models
22:19today that have been driving performance
22:20in programmatic, um think of it as just
22:22like a series of columns, you know, we
22:24know the device, we know the bundle, um
22:27you know, we're we're trying to create
22:28this table to to figure out where
22:29performance is.
22:31Vectorized embeddings are that and I can
22:33a quadratic quadratic space. Like it's
22:35just an an unbelievable number of
22:37pointers are going to be used now to
22:38make these decisions in a really smart
22:40way, and they're going to be reinforced
22:41by neural nets. That's the future of
22:43this market, but that's not Adgenta.
22:45That's just applying AI fundamentals to
22:48what we've been doing historically.
22:49Adgenta has an application, but I I I'd
22:52really link it to the human. Uh anything
22:54beyond that is completely
22:55cost-prohibitive, and to your point,
22:57we'll probably make the planet even
22:59hotter um just with the with the amount
23:01of scale that we deal with.
23:02>> Well, on that note of
23:05climate calamity
23:07>> [laughter]
23:07>> we'll end it there. Well, so Andrew's
23:09going to be around for most of the day.
23:10I I do advise people to take time to
23:12talk to Andrew cuz he's not here very
23:14often. And I do think that the market
23:16would be really interested in the
23:18containerization piece. Like if you
23:20think about the cost base for running
23:22programmatic in this market and other
23:24markets like it that are price
23:26sensitive, it changes the whole script.
23:28It really does shift it away from the
23:31big sort of hyper scalers and to
23:34independent ad tech and publishers and
23:37agencies and brands. So, Andrew, thank
23:39you very much.
23:40>> Thank you.
23:42>> [applause]
23:47[music]

About this transcript

This page contains the full transcript of The Future of the Open Internet Starts at the Impression by ExchangeWireTV, generated from the public captions YouTube serves with the video. The transcript has 4,525 words across 689 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to translate the transcript into 100+ languages.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, translation to 100+ languages, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.