WEBVTT

NOTE Sentence-level transcript of https://www.youtube.com/watch?v=Ot4OPrPH4xY

NOTE One cue per sentence. Cue ids are the line anchors on /transcripts/Ot4OPrPH4xY.html. A cue ends where the next begins, or 2 s after its last word.

s1
00:00:01.309 --> 00:00:03.309
[music]

s2
00:00:12.880 --> 00:00:14.400
So hi everyone.

s3
00:00:14.400 --> 00:00:16.560
Thank you so much for taking the time to join this session.

s4
00:00:16.560 --> 00:00:22.080
I hope I'll or at least I can guarantee I'll do whatever it takes uh to make it worth your time.

s5
00:00:22.080 --> 00:00:23.199
My name is Om.

s6
00:00:23.199 --> 00:00:26.080
I lead the product marketing team over at Bright Data.

s7
00:00:26.080 --> 00:00:31.119
Just by maybe a quick show of hands, who here is familiar with Bright Data?

s8
00:00:31.119 --> 00:00:32.320
Okay, we can do better.

s9
00:00:32.320 --> 00:00:34.320
I'll pass it on to our brand team.

s10
00:00:34.320 --> 00:00:37.840
Bright Data is a web data company.

s11
00:00:37.840 --> 00:00:45.760
Basically, we help more than 20,000 teams around the world, including more than 70% of the world's biggest AI labs to extract data from the web.

s12
00:00:45.760 --> 00:00:52.559
Just to put this in perspective of what scale we're talking about, we're talking well over 50 billion pages, HTMLs every day.

s13
00:00:52.559 --> 00:00:56.559
More than 20 pabytes of video, audio, and other media data.

s14
00:00:56.559 --> 00:01:01.120
So that's just a perspective of the type of work that we do at Bright Data.

s15
00:01:01.120 --> 00:01:02.079
But enough about us.

s16
00:01:02.079 --> 00:01:14.400
Personally, I joined Bright Data about three years ago, which essentially gave me front row seats at everything around AI and the web and how they started to actually connect.

s17
00:01:14.400 --> 00:01:26.080
It sounds very old but if you think about it only maybe less than two years ago right we were able to start using access the web and search the web through

s18
00:01:26.080 --> 00:01:42.159
cloud or through uh chpt that option didn't even exist in the earlier versions right so that connection that way in which both AI and the web are starting to converge is something that is still evolving and evolving rapidly and that's part of what I want to try and shed some light on today

s19
00:01:42.159 --> 00:01:44.880
and talk about a new emerging breed of companies.

s20
00:01:44.880 --> 00:01:48.240
That's coming out of this connection.

s21
00:01:48.640 --> 00:01:54.320
I think we can basically agree that the web is by far the world's greatest source of data.

s22
00:01:54.320 --> 00:01:57.280
At least historically when it comes to bright air, that's all we cared about, right?

s23
00:01:57.280 --> 00:02:00.320
It's helping our customers extract data from the web.

s24
00:02:00.320 --> 00:02:06.000
But with the emergence of AI and the emergence more recently of of AI agents that need to do knowledge work, right?

s25
00:02:06.000 --> 00:02:07.840
The web is no longer just a source of data.

s26
00:02:07.840 --> 00:02:11.039
We can actually start looking at it as a source of context.

s27
00:02:11.039 --> 00:02:25.200
context in the sense that if I do knowledge work and I have knowledge agents that support my work I want to go out to the web find the information that I need use it as context but kept keep on working so the data itself is only a step in the process for something bigger for the actions I

s28
00:02:25.200 --> 00:02:41.360
need to take for the conclusions I need to draw and for every downstream application that follows uh the first one's to figure this one out oh sorry even before that but one thing that I want all of us to bear in mind because this is going to follow us through the rest of this conversation

s29
00:02:41.360 --> 00:02:42.400
The web is messy.

s30
00:02:42.400 --> 00:02:43.360
It's unstructured.

s31
00:02:43.360 --> 00:02:46.239
And most importantly, it changes all the time.

s32
00:02:46.239 --> 00:02:49.040
This is a chart that shows data decay, right?

s33
00:02:49.040 --> 00:02:51.360
That it's an analysis done by our team that shows data decay.

s34
00:02:51.360 --> 00:02:58.080
Basically, how long after a new page, a new piece of content goes live, it is no longer relevant, right?

s35
00:02:58.080 --> 00:02:59.760
So, social media, it's easy for us to understand.

s36
00:02:59.760 --> 00:03:01.840
It's far less than a day.

s37
00:03:01.840 --> 00:03:09.040
But also news, finance, retail, 30 days later, data that was collected, it's mostly no longer relevant.

s38
00:03:09.040 --> 00:03:16.800
And when we acknowledge that this simple notion, we understand that extracting context from the web or relying on the web is not a snapshot.

s39
00:03:16.800 --> 00:03:17.840
It's not a one-time effort.

s40
00:03:17.840 --> 00:03:19.760
It's not even a monthly effort.

s41
00:03:19.760 --> 00:03:21.680
It's something that we need to keep on doing.

s42
00:03:21.680 --> 00:03:29.680
It's something that we need to look at as an ongoing process and something that we need to be mindful of.

s43
00:03:30.239 --> 00:03:34.640
The first ones to figure it out were of course search companies, right?

s44
00:03:34.640 --> 00:03:38.959
Only what three years ago we were on the far left, right?

s45
00:03:38.959 --> 00:03:41.040
This is this is in our lifetime, right?

s46
00:03:41.040 --> 00:03:42.879
Three years ago, we were on the far left.

s47
00:03:42.879 --> 00:03:43.920
Everything was Google.

s48
00:03:43.920 --> 00:03:50.879
There was no there was complete and total dominance up until three years ago for the past 20 or so years, right?

s49
00:03:50.879 --> 00:03:53.599
That's what we're talking about purely for humans.

s50
00:03:53.599 --> 00:03:58.000
Search something, go collect the information you need, and carry on.

s51
00:03:58.000 --> 00:04:07.840
Then fast forward maybe one and a half years ago, two years ago, search began to appear within the LLN within the chat bots, right?

s52
00:04:07.840 --> 00:04:13.840
Which already started blurring the line between humans and and agents because now the same bots also have the same web search available.

s53
00:04:13.840 --> 00:04:18.079
So the same LMS have the same access through API for the bot.

s54
00:04:18.079 --> 00:04:20.959
So we started seeing that convergence happening, right?

s55
00:04:20.959 --> 00:04:28.960
And so for the first time, we're seeing more and more traffic flowing down, search traffic, search intent flowing down these channels, not only to Google.

s56
00:04:28.960 --> 00:04:33.759
And last but not least, right, we now see a whole breed of companies, the AI search companies.

s57
00:04:33.759 --> 00:04:34.960
I'm sure you're familiar with them.

s58
00:04:34.960 --> 00:04:41.360
I caught a talk yesterday by Will, the CEO of Exa and Parallel and you.com and Tavili and a bunch of others.

s59
00:04:41.360 --> 00:04:45.440
They are purely built and indexing the web especially for agents.

s60
00:04:45.440 --> 00:04:48.639
They're not even looking at the humans involved anymore.

s61
00:04:48.639 --> 00:04:55.040
So Google's dominance when it comes to search, if Google was synonymous of web search, that is very much shaking.

s62
00:04:55.040 --> 00:05:00.479
When there's blood in the water, the sharks come just last week.

s63
00:05:00.479 --> 00:05:13.120
Amazon Amazon announced, I don't know how many of you saw it, that they developed their own index and started allowing the ability to retrieve data from the web to retrieve contacts for agents on agent core.

s64
00:05:13.120 --> 00:05:17.039
Amazon developed their own search engine.

s65
00:05:17.360 --> 00:05:20.080
Two weeks before that it was Microsoft.

s66
00:05:20.080 --> 00:05:21.840
Microsoft always had skin in the game.

s67
00:05:21.840 --> 00:05:30.800
I know one or two% of the world search traffic went to Microsoft but they have now repackaged it and launched it again as part of web byq as part of their suite

s68
00:05:30.800 --> 00:05:34.800
for agentic development and orchestration.

s69
00:05:34.960 --> 00:05:40.160
So we're seeing more and more this space becoming crowded.

s70
00:05:40.320 --> 00:05:48.800
But when we're talking about context and we're looking at this through the lens of search, I believe it only tells us part of the story, right?

s71
00:05:48.800 --> 00:05:54.880
I can search for, I don't know, what's the cost of a certain pair of sneakers this morning, right?

s72
00:05:54.880 --> 00:05:57.199
In on a certain website.

s73
00:05:57.199 --> 00:06:04.319
I cannot really search for how has that price changed over the last six months, what discounts it had, right?

s74
00:06:04.319 --> 00:06:08.240
I can search for what open job positions we have at Bright Data.

s75
00:06:08.240 --> 00:06:08.720
We do.

s76
00:06:08.720 --> 00:06:17.840
I urge you to go have a look but I can't see how how that that was a chart and how that changed over time and how the headcount of the company changed over time

s77
00:06:17.840 --> 00:06:21.680
and all of that information existed on the web simply back then.

s78
00:06:21.680 --> 00:06:34.960
So when we actually start to think about it we understand that there's much more context in the web that than what web search allows us to extract and this is what we started seeing in the recent years a whole new breed of companies

s79
00:06:34.960 --> 00:06:35.680
rising.

s80
00:06:35.680 --> 00:06:39.919
We like to call them internally casts, context as a service because that's what they do.

s81
00:06:39.919 --> 00:06:53.120
They allow agents to tap into them MCP, CLI, uh just pure good old API and actually start extracting data to to retrieve data so they can reason over for whatever knowledge work they are responsible for.

s82
00:06:53.120 --> 00:06:55.039
We see this happening in e-commerce.

s83
00:06:55.039 --> 00:06:56.160
We see this happening in travel.

s84
00:06:56.160 --> 00:07:03.759
We see this happening in in finance, in market research, in uh in HR, in real estate, in a bunch of other domains.

s85
00:07:03.759 --> 00:07:05.599
I'll show a few examples in a second.

s86
00:07:05.599 --> 00:07:06.000
Right?

s87
00:07:06.000 --> 00:07:15.759
What all of these have in common is that they don't only just discover the web, you know, in terms of think crawling, think searching, think all of that, accessing, extracting the data and indexing it.

s88
00:07:15.759 --> 00:07:17.440
They take it a step further.

s89
00:07:17.440 --> 00:07:24.880
They actually develop knowledge graphs to start structuring all of the entities and to ddup them and they start enriching them with a lot of different sources.

s90
00:07:24.880 --> 00:07:27.280
So they actually start merging all that data.

s91
00:07:27.280 --> 00:07:31.599
If you think about it, they kind of behave like vertical search engines, right?

s92
00:07:31.599 --> 00:07:45.919
they are a very very very good search engine for something very specific and and it's it's already in full motion right so as I said we see this in finance and in market research and retail and e-commerce and GTM and sales intelligence

s93
00:07:45.919 --> 00:07:56.160
what all of these companies by the way have in common they're all part of brighta's startup program if you are a builder and this is a hot space to go in because I think we're only tapping the surface

s94
00:07:56.160 --> 00:08:04.800
I invite you to scan this and apply up to $20,000 in credits and all sorts of co-arketing but that's enough self-promotion.

s95
00:08:06.960 --> 00:08:20.000
So CAS as as an industry is already in full bloom and as always with these situations right also the traditional players aren't left too much behind there at the bottom

s96
00:08:20.000 --> 00:08:30.720
you see good old data as a service you see zoom info right by researching for this uh presentation today I also saw that they launched that thing at the top it's called gtm.ai AI.

s97
00:08:30.720 --> 00:08:33.519
You can only imagine how much they paid for that domain.

s98
00:08:33.519 --> 00:08:39.360
But they launched as a secondary brand for Zoom Info that is catering specifically for the need of agents.

s99
00:08:39.360 --> 00:08:41.120
Look at look at the wording, right?

s100
00:08:41.120 --> 00:08:42.479
They talk about GTM work, right?

s101
00:08:42.479 --> 00:08:49.760
That knowledge work, that research that you do when you need to prospect, when you need to do headhunting, when whatever it is you need to do that involves people mostly

s102
00:08:49.760 --> 00:08:54.959
straight from cloud code, straight from codeex or any other agent, they understand the gap, right?

s103
00:08:54.959 --> 00:09:01.920
So yeah, it's funny to think of Cass as an evolution of DAS and it is in a way, but it's catering for a very specific need.

s104
00:09:01.920 --> 00:09:09.040
As much as the AI search engines are different than Google, when agents need them, it's different than people.

s105
00:09:10.800 --> 00:09:21.360
When we let this one sink that at the very least we have two different types of paths to complete knowledge work as an agent, we can start thinking about this in terms of web context engineering.

s106
00:09:21.360 --> 00:09:32.720
We can start thinking about this in terms of how do I optimize for the specific task or more importantly when things come as it a as it is how do I optimize this for

s107
00:09:32.720 --> 00:09:45.760
breeds of task how do I I do this for various parts of the organization that I'm building for if I'm an AI engineer I need to serve different teams they may have different needs it's very tempting to throw AI search at all of them but maybe that's not optimal

s108
00:09:45.760 --> 00:09:52.240
maybe I need a combination of both maybe I can start seeing all sorts of cost efficiencies emerge emerge from that.

s109
00:09:52.240 --> 00:09:56.959
So for the second half of this presentation, we actually went ahead and created a test.

s110
00:09:56.959 --> 00:09:58.320
This is not a benchmark.

s111
00:09:58.320 --> 00:10:10.080
You won't see any any something concrete that I can say with great confidence other than the actual research that we did because we wanted to start unraveling the different considerations and how do these two stack up against each other.

s112
00:10:10.080 --> 00:10:11.440
So we designed a test.

s113
00:10:11.440 --> 00:10:12.560
We went for something basic.

s114
00:10:12.560 --> 00:10:18.079
We said okay let's take a company an entity and try and enrich it across 25 different fields.

s115
00:10:18.079 --> 00:10:19.200
Some of them are very easy.

s116
00:10:19.200 --> 00:10:25.440
you know the company domain, the name uh the headquarters, but some are more challenging, right?

s117
00:10:25.440 --> 00:10:27.040
Things about hiring and people and something.

s118
00:10:27.040 --> 00:10:39.519
And we build a simple agent, a loop in a loop that knows that uses Opus 4.8 as the harness and it starts to go over field by field, go out, search for it or retrieve it from the cast,

s119
00:10:39.519 --> 00:10:47.839
do it again and again and again until it completes and brings back set some guardrails, you know, like budget and stuff just to keep it fair.

s120
00:10:47.839 --> 00:10:49.680
and I want to share with you the results.

s121
00:10:49.680 --> 00:11:00.079
So the first thing that we would care about right being knowledge work would be uh sorry we ran it 100 times on all of the sponsors of today's event.

s122
00:11:00.160 --> 00:11:04.880
So the first thing that we saw in terms of coverage is that there's pretty good convergence.

s123
00:11:04.880 --> 00:11:07.200
They all did fairly well, right?

s124
00:11:07.200 --> 00:11:09.600
I'll get to the two at the bottom in a second.

s125
00:11:09.600 --> 00:11:12.320
So search were consistent performance.

s126
00:11:12.320 --> 00:11:15.279
One of the major cast providers were also very well.

s127
00:11:15.279 --> 00:11:17.680
The third one by the way you can see unlocker and SER.

s128
00:11:17.680 --> 00:11:21.200
SER is good old data uh good old Google.

s129
00:11:21.200 --> 00:11:26.160
We basically did the same thing just with Google and it performed pretty well in extracting that information.

s130
00:11:26.160 --> 00:11:32.399
Native is cloud's own uh search and you see that they converge really well.

s131
00:11:32.399 --> 00:11:36.320
I was originally surprised about the two cast solutions at the bottom.

s132
00:11:36.320 --> 00:11:37.440
It was counterintuitive.

s133
00:11:37.440 --> 00:11:44.320
I expected cast to dominate this thing because that's you know you had one job right to map out all these companies.

s134
00:11:44.320 --> 00:11:52.800
But but after diving into into it a bit more you you understand that well they are limited in the sense that they know what they have about an entity.

s135
00:11:52.800 --> 00:11:57.680
If I ask it a question that is beyond that, they will never have that data, right?

s136
00:11:57.680 --> 00:12:00.720
Unlike a searcher can go out and continue searching and exploring it.

s137
00:12:00.720 --> 00:12:05.360
If they didn't collect data about the recent job hiring, it will never be there, right?

s138
00:12:05.360 --> 00:12:14.800
So, it makes sense that they are a bit behind, but I'm sure at the same time that they have a lot of other advantages that we simply didn't ask for, a lot of other fields that they didn't have that aren't represented.

s139
00:12:14.800 --> 00:12:25.279
So again it it creates some complexities when what how do we measure coverage when it relates to the specific job that we need to do rather than um in general

s140
00:12:25.279 --> 00:12:38.160
when we the second thing we looked at was cost of course here it's we started seeing it spread out a bit so you can see that massive bulk in the center most of the search and the and the cast and even using uh Google right h

s141
00:12:38.160 --> 00:12:46.160
converge to pretty much the same cost only different right the cast was just about the service itself what you pay the vendor, right?

s142
00:12:46.160 --> 00:12:55.839
All of the other search solutions, you also needed a lot of token burn it burn to actually structure that data so you can actually act on it and and and use it as something retrievable,

s143
00:12:55.839 --> 00:12:56.000
right?

s144
00:12:56.000 --> 00:12:59.600
So, it's the same output native obscenely expensive.

s145
00:12:59.600 --> 00:13:04.320
And the the cast on the right, I'm sure you're all familiar with the by far the most expensive in the industry.

s146
00:13:04.320 --> 00:13:07.600
I will not name and shame them.

s147
00:13:07.600 --> 00:13:08.079
Interesting.

s148
00:13:08.079 --> 00:13:13.680
You see that small cast there at the the left that cast number two that were very cheap.

s149
00:13:13.680 --> 00:13:26.480
They're also the ones that are here at the bottom, which is funny because what I believe is happening there is that we're seeing even within this industry niche players that have, you know, lower quality data but much cheaper.

s150
00:13:26.480 --> 00:13:33.839
They're already carving that niche of the longtail, right, of small shops or small usage that don't want to pay as much and don't need as much data.

s151
00:13:33.839 --> 00:13:37.760
And we're all seeing it them branch out there.

s152
00:13:38.480 --> 00:13:41.440
Most of you here, I presume, are engineers.

s153
00:13:41.440 --> 00:13:50.079
So there's a very evident question that we did not ask here which is what is the one thing that an engineer would care about?

s154
00:13:52.000 --> 00:13:52.480
Thank you.

s155
00:13:52.480 --> 00:13:55.200
Let's talk about scale.

s156
00:13:55.920 --> 00:13:58.560
This is the cost not for the whole 100.

s157
00:13:58.560 --> 00:14:04.160
This is the cost per one for one record.

s158
00:14:05.600 --> 00:14:07.680
What happens if we need a million?

s159
00:14:07.680 --> 00:14:16.160
Now yes a million records will not fit in the context with obviously we're not talking about a single run that needs a million you can think about million in terms of the frequency

s160
00:14:16.160 --> 00:14:31.920
right if I am a market research I do due diligence for private equity I revisit these companies all the time I ask more questions about them as the time goes by was there any new news about them was there anything that changed somebody joined somebody leave do they have new hires I keep on asking the same thing

s161
00:14:31.920 --> 00:14:46.480
so when I'm talking about this the multiply by a million it's not just about the number of companies it's the frequency in which I'm asking it frequency is the cost killer when we talk about these and we need to acknowledge that right we're we're thinking about this in terms of web context engineering we're starting to look at

s162
00:14:46.480 --> 00:15:01.360
it differently every repeated query costs the same as the first even if it brought back the exact same answers nothing changed pay up right no is false positives for sure go in

s163
00:15:01.360 --> 00:15:14.160
token cost right we saw them all that there's very high token we know that doesn't that doesn't shrink think well over time there's always some volume element in terms of the cost but it's not the same as you know flatlining

s164
00:15:14.160 --> 00:15:29.360
right and if we bring this back to knowledge work this is where we see teams that are starting to cut corners so I won't research this company every day I'll look at it once a week or once a month I won't ask that question now I will I don't want all the results I'll only take 10 results

s165
00:15:29.360 --> 00:15:36.560
20 results something we start so we already have the setup we have what we need to do the knowledge work but at the same time.

s166
00:15:36.560 --> 00:15:40.959
We're not extracting all of the value because we're starting to be conscious about cost, right?

s167
00:15:40.959 --> 00:15:43.279
Basically, we're renting context.

s168
00:15:43.279 --> 00:15:45.360
We're not owning the context that we use.

s169
00:15:45.360 --> 00:15:48.320
That is a very important distinction.

s170
00:15:48.720 --> 00:15:51.680
Again, if we're good engineers and we ask ourselves, what about scale?

s171
00:15:51.680 --> 00:15:57.199
The second and most obvious thing that will come to mind now, so how about we build it?

s172
00:15:57.199 --> 00:16:06.320
What if we take all of that web data ourselves and stick it in some vector database and try and see what comes out of it?

s173
00:16:06.320 --> 00:16:10.399
So I asked my AI engineer to do exactly that.

s174
00:16:10.399 --> 00:16:12.399
Again, this is a test.

s175
00:16:12.399 --> 00:16:14.639
This is not a benchmark or a full-blown operation.

s176
00:16:14.639 --> 00:16:24.399
This is a day's work at best just to illustrate the concept and to show something about the cost efficiencies that you can generate potentially by doing it yourself potentially

s177
00:16:24.399 --> 00:16:26.959
in specific scenarios.

s178
00:16:26.959 --> 00:16:43.440
The test simple take the company name nothing but run it through Google find the relevant entries the relevant URLs of that company in various websites that have all of that information right you use search when you don't know the source but when we're talking about company enrichment we all know these sources we all know where that data

s179
00:16:43.440 --> 00:16:53.199
comes from the cast also bring it from them zoom info bring it from them it's the same thing over and over again why not just go straight to the source why are we doing that middleman thing

s180
00:16:53.279 --> 00:16:57.519
LinkedIn companies LinkedIn jobs crunchbas Right there we have scrapers for those.

s181
00:16:57.519 --> 00:16:59.839
You just tap in and you start paying as a pay as you go.

s182
00:16:59.839 --> 00:17:01.920
We build two dedicated scrapers.

s183
00:17:01.920 --> 00:17:04.480
We have a new AI tool called scraper studio.

s184
00:17:04.480 --> 00:17:09.039
It's basically lets you build a scraper for any website in less than 5 minutes.

s185
00:17:09.039 --> 00:17:10.319
All powered by AI.

s186
00:17:10.319 --> 00:17:12.480
And then it also have a self-healing function, right?

s187
00:17:12.480 --> 00:17:16.720
So if the website changes, it fixes itself and keeps on going.

s188
00:17:16.720 --> 00:17:19.520
Merge it all into one entity.

s189
00:17:19.520 --> 00:17:20.400
Basic heristics.

s190
00:17:20.400 --> 00:17:22.640
If there's conflict, choose that over that.

s191
00:17:22.640 --> 00:17:28.799
And eventually we have a data set of these uh 100 companies.

s192
00:17:28.799 --> 00:17:31.760
Zero AI cost involved.

s193
00:17:31.760 --> 00:17:34.080
There's no tokens.

s194
00:17:34.080 --> 00:17:36.880
Coverage fairly well.

s195
00:17:36.880 --> 00:17:38.160
Not amazing.

s196
00:17:38.160 --> 00:17:41.440
Not the best that we saw here, but stacking up pretty well.

s197
00:17:41.440 --> 00:17:44.320
And again, this is just a day experiment.

s198
00:17:44.320 --> 00:17:46.080
Probably not even as much.

s199
00:17:46.080 --> 00:17:52.320
Okay, so again and again, very specific task, very limited context, very limited situation.

s200
00:17:52.320 --> 00:17:58.160
you know tread uh tread lightly and then proceed with caution when it comes to conclusions.

s201
00:17:58.160 --> 00:17:59.919
The real story is not this.

s202
00:17:59.919 --> 00:18:02.640
The real story is this.

s203
00:18:03.679 --> 00:18:08.240
That's what it cost to just go and fetch that data that is out there.

s204
00:18:08.240 --> 00:18:15.520
We think about knowledge graphs, we think about entities, but if you think about for example LinkedIn, the data is already structured in form of entities.

s205
00:18:15.520 --> 00:18:21.200
There's an entity for a company, there's entity for a person, there's entity for a job and they're connected between them.

s206
00:18:21.200 --> 00:18:22.799
Sometimes the ontology is already there.

s207
00:18:22.799 --> 00:18:29.120
Again, this is not the most complicated of scenarios, but this is pretty damn good.

s208
00:18:29.120 --> 00:18:34.480
Now, yes, it took time to set up.

s209
00:18:34.480 --> 00:18:39.520
So, it's not really fair to compare apples to apples when it comes to the cost because these are out of the box.

s210
00:18:39.520 --> 00:18:41.200
You can just tap into the API.

s211
00:18:41.200 --> 00:18:44.640
That one that I just showed you required some setup.

s212
00:18:44.640 --> 00:18:45.679
Let's say it's a week.

s213
00:18:45.679 --> 00:18:48.400
Let's price it at $5,000 just to give us some perspective.

s214
00:18:48.400 --> 00:18:52.480
We can actually start thinking about this in terms of a tipping point.

s215
00:18:52.480 --> 00:19:00.320
We can actually start thinking about what is that tipping point in which it makes more sense for me to build it myself, right?

s216
00:19:00.320 --> 00:19:01.919
Then keep on renting it.

s217
00:19:01.919 --> 00:19:08.000
Now again, everything to the left of that dot in this case it was just over 15,000 entities or queries, right?

s218
00:19:08.000 --> 00:19:11.840
When we think about it, so it made sense to do it at this point.

s219
00:19:11.840 --> 00:19:15.600
Maybe it's not 15, maybe it's 30, maybe it's 100,000, maybe it's 10,000.

s220
00:19:15.600 --> 00:19:17.039
Really depends on the use case.

s221
00:19:17.039 --> 00:19:22.080
But there is a tipping point in which it actually makes sense to do it yourself.

s222
00:19:22.080 --> 00:19:31.919
H which leads us to the fact that both AI search and casts and all these solutions they're very get good in the sense that you can just plug and play.

s223
00:19:31.919 --> 00:19:43.039
But if your knowledge work needs right are persistent and consistent and to a certain degree may even continue escalating and growing then this is perhaps a direction to start considering.

s224
00:19:43.039 --> 00:19:46.400
Maybe I can just go ahead and build my own.

s225
00:19:46.400 --> 00:19:55.360
Cuz the nice thing about it is that all of the things that we see on the left up until the third part is upfront investment.

s226
00:19:55.360 --> 00:20:02.240
And the most important thing that whatever retrieval happens later on from the from the agents, right, is free.

s227
00:20:02.240 --> 00:20:04.160
Not really free, but you get what I mean, right?

s228
00:20:04.160 --> 00:20:05.200
There's no added cost.

s229
00:20:05.200 --> 00:20:07.120
I can just ask that question over and over again.

s230
00:20:07.120 --> 00:20:08.400
I did not like the first answer.

s231
00:20:08.400 --> 00:20:09.600
I'll ask it again.

s232
00:20:09.600 --> 00:20:11.840
I'll ask it a hundred times until I get what I need.

s233
00:20:11.840 --> 00:20:16.160
I I have no more fear, no more cutting corners, which is maybe the most important thing.

s234
00:20:16.160 --> 00:20:18.960
And I'm leaving aside the fact that this is also custom business logic.

s235
00:20:18.960 --> 00:20:20.400
I can connect it with my own data.

s236
00:20:20.400 --> 00:20:22.799
There's all sorts of other advantages of owning it.

s237
00:20:22.799 --> 00:20:26.559
You know, we'll keep it to the imagination.

s238
00:20:26.829 --> 00:20:28.799
[snorts]

s239
00:20:28.799 --> 00:20:32.080
Remember, we asked about a million, not about 15,000.

s240
00:20:32.080 --> 00:20:35.520
This compounds, this compounds greatly, right?

s241
00:20:35.520 --> 00:20:36.799
We need that horizon.

s242
00:20:36.799 --> 00:20:38.159
Remember, the web keeps changing.

s243
00:20:38.159 --> 00:20:41.039
We saw the stainless data and how the data decays.

s244
00:20:41.039 --> 00:20:50.080
So we need to be thinking about this in the long run and how this will evolve when we keep on asking the questions about the entities that we care about

s245
00:20:50.799 --> 00:20:52.000
just to wrap it up.

s246
00:20:52.000 --> 00:21:00.000
So AI search CAS they can get you very far when what you need is ad hoc when what you need is always changing when sometimes you look at different things

s247
00:21:00.000 --> 00:21:13.360
even the mix and match of them for certain task use this for certain task use that you can I'm sure again I just showed a test there's a lot of ways to optimize it just like any other context engineering and make and use uh lighter models and use other stuff there's a lot of great stuff to be

s248
00:21:13.360 --> 00:21:23.760
done but eventually the frequency will come and bite you in the ass when it comes to cost and that's something to be mindful of and there's a fair chance that that tipping point

s249
00:21:23.760 --> 00:21:25.440
is much lower than you think.

s250
00:21:25.440 --> 00:21:32.400
And that's something that when as we design these systems, when we think about web context engineering, we need to be mindful of that.

s251
00:21:32.400 --> 00:21:39.919
And last but not least, the last slide we showed owned context compounds while rented decays, right?

s252
00:21:39.919 --> 00:21:41.440
It's not a one-time task.

s253
00:21:41.440 --> 00:21:43.280
Again, if it's a one-time question, use AI search.

s254
00:21:43.280 --> 00:21:44.960
It will be amazing.

s255
00:21:44.960 --> 00:21:57.679
when you need to do it over and over again, there's a fair chance that it will not that you're missing out on potential compounding effect and you are losing out.

s256
00:21:57.679 --> 00:21:59.840
Thank you very much.

s257
00:21:59.840 --> 00:22:01.840
[applause]