Latest / The Learning Corner by Precursor / Episode #33: Do $1B Outcomes Matter?, Vibe Investing, The Agentic Web and The Original Sin

Transcript

Episode #33: Do $1B Outcomes Matter?, Vibe Investing, The Agentic Web and The Original Sin · The Learning Corner by Precursor

• 0:00
  And look, a 500,000,000 to a billion dollar company is still very significant.
• 0:04
  It just isn't as meaningful to the largest venture capital funds as it used to be.
• 0:07
  I think we're gonna have to rebuild all the infrastructure for how we support and advise
• 0:11
  those companies because it's a different world.
• 0:12
  It's the week of May 19.
• 0:15
  Welcome to the Learning Corner, a Precursor Ventures podcast.
• 0:22
  Welcome to the Learning Corner by Precursor.
• 0:24
  I'm Mia and I'm joined by RGP and managing partner, Charles Hudson, where we will be
• 0:29
  talking through the articles and Goodreads that caught our team's attention over the past week.
• 0:34
  And we've also got Leon back in the studio this week.
• 0:37
  Excited to kick off.
• 0:38
  So our first topic was written by David Clark, who's the CIO of VenCap International.
• 0:43
  He was reflecting on Paki McCormick's post about the growth in these back to exit values
• 0:48
  and, as a result, fund sizes.
• 0:49
  In Paki's post, he highlighted how $1,000,000,000 outcomes today are less
• 0:54
  extraordinary than they used to be.
• 0:55
  A $1,000,000,000 exit in February was in the ninety ninth percentile.
• 0:58
  By today's standards, it's roughly in the 80 percentile.
• 1:01
  David builds on this analysis and projects the data forward to estimate that top percentile
• 1:06
  exits could reach 20,000,000,000 by 2030, '40 billion by 2035.
• 1:11
  So we see that the VC power law is alive and well, and David implies that funds today should
• 1:16
  prepare for the exit environments of tomorrow.
• 1:19
  Charles, question for you.
• 1:20
  What do these assumptions mean for GPs looking to raise funds in today's market and LPs
• 1:25
  allocating capital, and how do their considerations differ?
• 1:28
  I think as we've talked about before, these funds are ten to fifteen year enterprises.
• 1:33
  So if you are raising a fund today, most of your companies are gonna exit in the mid twenty
• 1:38
  thirties.
• 1:39
  I think there's value in saying, what do I think the world looks like?
• 1:42
  Also, I think every fund size has a set of assumptions that have to be true in order for
• 1:46
  those funds to be attractive.
• 1:47
  And if you're gonna raise a multi billion dollar, multi stage venture fund, and you say,
• 1:52
  hey, I wanna give people a reasonable return, you're talking about returning tens of billions
• 1:56
  of dollars to them.
• 1:58
  And the only way to do that, I believe, is through really large outcomes.
• 2:01
  Two things got to be true.
• 2:02
  There have to be 30 to $40,000,000,000 outcomes, and your specific venture firm needs
• 2:08
  to be in those.
• 2:09
  And my hunch is that it's zero sum.
• 2:11
  Since there's only a % to buy, it is impossible that every large venture firm can own 20% of
• 2:16
  every meaningful exit.
• 2:18
  Interesting thing will be, of those really large firms, who does the best job of capturing
• 2:22
  meaningful ownership in the companies that ultimately matter.
• 2:25
  And I think if you're a really big venture fund, that's the only game that matters to you
• 2:29
  because it's the only game that moves in as the Post says, the median exit doesn't do anything
• 2:33
  for a big venture fund.
• 2:34
  It simply isn't big enough.
• 2:35
  Even if they own 200% of that company, it wouldn't matter.
• 2:37
  To that point, there's only been around like 10 non IPO related exits that have surpassed 20
• 2:43
  plus billion dollars.
• 2:44
  And so that's one per year if you kind of average it out to even max that out to what
• 2:49
  Deva was mentioning, to $40,000,000,000.
• 2:51
  I'm assuming also for the large funds, you'd want also exposure in as many companies as you
• 2:56
  possibly could within the areas that you think could possibly return that.
• 2:59
  How do you marry to?
• 3:00
  Bring exposure into and then that you can
• 3:02
  have that I think the scary thing and look, this is not a problem I have to deal with
• 3:06
  because Precursor is not a multi billion dollar multi stage fund.
• 3:09
  But if you think about it, you need two things.
• 3:11
  You both need to be in we'll just take Wizz.
• 3:13
  You need to be in Wizz and you also need to own enough.
• 3:16
  And this is the scary part.
• 3:17
  Those outcomes are large and relatively infrequent.
• 3:20
  So like, I don't think the frequency of outcomes of that scale is gonna change, which
• 3:24
  really means when they happen, you have to own enough of them to matter.
• 3:28
  So the two ways you can do that, one is what you suggest in me, which is everything that
• 3:32
  seems interesting, go and try to buy 10 or 15% of it at c or at series a and hope that you
• 3:37
  stay close enough to the stuff that's working that when it starts to grow in scale, you can
• 3:40
  maintain or expand your pro rata.
• 3:43
  Or if something appears to break out and you missed it, grab it as quickly as you can.
• 3:47
  So if you didn't win the a, make sure you really pile in in the b.
• 3:51
  And maybe you can't buy the 15 to 20% that you wanted, but you can get something.
• 3:55
  I would imagine also all without maybe having full conviction in the metrics, etcetera.
• 4:00
  Oh, yeah.
• 4:00
  You just wanna be part of
• 4:01
  those Yeah.
• 4:01
  You wanna be part of those rounds.
• 4:02
  And I think it's gonna create an even more stark feeling of have and have not in venture.
• 4:07
  There will be companies that people are like, oh, for whatever reason, it could be metrics,
• 4:11
  it could be team, it could this one feels like a winner.
• 4:14
  Everybody wants to be in it, and you have other businesses that are perfectly fine and good,
• 4:17
  and nobody cares.
• 4:18
  Leaning back into some internal conversations that we've had, which has been wonderful on the
• 4:21
  team.
• 4:22
  We spent a lot of time thinking about and discussing these market shifts.
• 4:25
  We've covered the steep drop off in the seat to series a graduation rate, how most founders are
• 4:31
  subject to this three year and 7 to $10,000,000 runway to test how venture scalable their ideas
• 4:36
  and then kind of like the PE off ramp concept.
• 4:39
  So tying this all together, how do we advise founders on navigating this environment while
• 4:44
  also thinking about fund managers and their expectations and what they need to see from an
• 4:48
  outcome perspective?
• 4:49
  I think the question's going to be if you're a fund manager and you see a company where the
• 4:54
  math works for your fund, so maybe I would be very happy owning 5% of a billion dollar
• 4:58
  outcome given our fund size, that'd be great.
• 5:00
  If you know upfront the company that you're dealing with is big, but not capital b big, as
• 5:04
  in it will catch the eye of multistage venture funds.
• 5:07
  Like, what is the funding ecosystem instead of options gonna look like for those companies?
• 5:11
  And I'd argue we don't have that solved yet.
• 5:14
  If you wanna be a hyperscaler or you wanna become one of these companies that raises a ton
• 5:18
  of money and shoots for the moon, we have a pretty good set of products for you to do that.
• 5:21
  If you're like, no.
• 5:22
  I'm building something that I think will be pretty and look.
• 5:25
  A 500,000,000 to a billion dollar company is still very significant.
• 5:28
  It just isn't as meaningful to the largest venture capital funds as it used to be.
• 5:32
  I think we're gonna have to rebuild all the infrastructure for how we support and advise
• 5:35
  those companies because it's a it's a different world.
• 5:37
  And I assume it'll be more around coaching our founders into walking through how big TAM looks
• 5:42
  like and kind of how to storyboard that out for those types of investors who need to know that
• 5:46
  that outcome is possible.
• 5:47
  Whereas before, I feel like that was something you could gradually grow into because the
• 5:51
  outcome just didn't have to so early on.
• 5:53
  And I and I think my hunch is investors are gonna make decisions also just about teams.
• 5:57
  Like, I believe this team?
• 5:58
  And that's obviously highly subjective.
• 6:00
  Do I think this team has what it takes to build and run a multibillion dollar company?
• 6:04
  And I hope we don't miss out on some awesome people because they don't look like they came
• 6:08
  on a central casting on day one.
• 6:10
  I think in the majority of people's minds, that archetype looks like two, three time ex founder
• 6:15
  who's already raised and exited.
• 6:17
  So in an optimistic way, like, you are a smaller fund, you now have this awesome window
• 6:22
  of opportunity to capitalize on stellar founders who don't look good for that.
• 6:25
  Well, our next topic is a piece written by Colin Wallace, GP at Lobby Capital, opining on
• 6:31
  the growth of Vibe Investing.
• 6:33
  It's the process of feeding sentiment and historical data into AI models to generate
• 6:37
  investment decisions.
• 6:38
  He highlights that the value created by Vibe Investing is limited if investors don't
• 6:43
  understand why specific decisions are being made.
• 6:46
  The rise of AI has also dramatically increased startup creation and homogeneity, making it
• 6:51
  increasingly difficult to identify outlier opportunities.
• 6:55
  A large portion of his piece focuses on the importance of a founder's ability to generate
• 7:00
  perishable knowledge and an investor's ability to identify it.
• 7:04
  Ultimately, his conclusion is clear.
• 7:06
  AI should be a tool that enhances but does not replace investor judgment.
• 7:11
  So getting more specific into perishable knowledge, Colin compares the processing power
• 7:16
  of the latest AI models against that of a human brain.
• 7:20
  And obviously, the AI model currently surpasses the human brain from a speed perspective.
• 7:24
  However, it's the context and quality of insights that have been processed that AI
• 7:29
  perhaps has trouble accessing today, which makes a human in the loop important for today's
• 7:34
  investment decisions.
• 7:35
  So I would love your thoughts on that.
• 7:37
  Yeah.
• 7:38
  I mean, I like this piece because I always like reading about how other people are integrating
• 7:42
  AI into their decision making process, whether it's as a substitute for decision making or a
• 7:47
  compliment.
• 7:48
  And I'm of the mind that it's inevitable that there will be programmatically deployed capital
• 7:53
  in private markets, including early stage venture, just because there will be somebody
• 7:58
  who's like, we can't afford to not run this experiment.
• 8:01
  And I think right now, we're in a place where I don't know many VCs who aren't using AI in some
• 8:06
  way, shape, or form to augment their workflow.
• 8:08
  It might be as simple as, oh, we just dump all of our work in here, and it writes a memo that
• 8:14
  synthesizes a decision we've already made to, hey, generate a bunch of due diligence
• 8:18
  questions, or what did I meant?
• 8:19
  There's a bunch of ways people most people I know haven't yet said, I'm gonna allow the
• 8:24
  algorithm to make a decision for me.
• 8:26
  But we have lots of other areas of finance on the public equity side where we've essentially
• 8:31
  done that.
• 8:31
  In some cases, it's fully automated.
• 8:33
  In some cases, there's a human supervising, but we're already letting computers make a lot of
• 8:36
  decisions, and I think it's inevitable that somebody will do this.
• 8:39
  And the thing that's curious to me is, will the AI perform better or differently than a human?
• 8:45
  And if I think about a lot of our best investments, there were obvious flaws in them,
• 8:51
  and the thing that got me excited was something about the person, and the question is, will AI
• 8:55
  be able to find that spark about people?
• 8:58
  There's no doubt in my mind that on a quantitative basis that AI will be faster, more
• 9:02
  efficient, more thorough than a human at simply crunching through data.
• 9:05
  But at the early stage, the thing that is taste and feel, I think, is people.
• 9:09
  And for me, the question is, what if somebody builds a really good AI clone that asks the
• 9:13
  same questions and understands how you think?
• 9:15
  And maybe there will be a point where the synthetic you can meet with the synthetic you
• 9:20
  of the founder and make a decision that would approximate yours.
• 9:23
  I don't think we're there yet, and I think that's still the missing piece.
• 9:26
  Because I think particularly for people who are doing it for the first time, a lot of who we
• 9:29
  back, it's actually hard to know who's gonna rise to the occasion.
• 9:32
  Sometimes there are things in their background that would give you clues, and sometimes the
• 9:36
  people just turn out to be fantastically awesome.
• 9:38
  But I'm pretty sure AI could replicate the median VC investment decision, but the median
• 9:42
  investment decision is to lose money.
• 9:44
  Right.
• 9:45
  So that's actually not that useful.
• 9:46
  The article also talks about perishable knowledge.
• 9:48
  Did you have thoughts on how IEI can have an impact on that, maybe from a founder's
• 9:52
  perspective?
• 9:53
  I think one thing I'm always asking myself when I meet a founder is how much does this person
• 9:58
  know about the problem, and what's the half life of that set of insights?
• 10:01
  Because very few things are true forever, and some things are so generally true.
• 10:05
  Like, the Internet's gonna be big, yes, but what do you do with that insight?
• 10:08
  And I'm always trying to find people who have something that's gonna give them a two or three
• 10:12
  year window, that even if it ceases to be true at some point, or it's an insight that they
• 10:17
  understand that isn't widely understood yet, it'll be enough to get them a head start.
• 10:20
  Now what they do with that head start is really up to them, but it will get them a head start.
• 10:23
  And I think one thing I'm always surprised by is how quickly you can get up to speed on the
• 10:28
  broad strokes of a topic using AI, which is very different than I've done it on a handful
• 10:33
  of topics right.
• 10:34
  I think I know a fair amount.
• 10:35
  I'm like, wow.
• 10:35
  This is an 80 answer with no effort, but the 20% that's missing is actually the really
• 10:40
  important context.
• 10:42
  Yeah.
• 10:42
  And is it the effort required to get that is not trivial.
• 10:45
  Like, does a great job of synthesizing all the things that I I know and maybe puts it into
• 10:49
  words a lot quicker, but it didn't add as much value as I had been hoping for to get me to an
• 10:54
  end decision.
• 10:55
  I really love Colin's comparison of the importance of confidence in good investing.
• 10:59
  He compares it to grade school and the importance of showing your work for difficult
• 11:03
  math problems.
• 11:04
  Similar to investing, if you don't know how you arrive to an answer, it's difficult to have the
• 11:08
  conviction to act on or justify an existing position.
• 11:12
  And this is not the first time we've encountered people flocking to a thesis area or
• 11:16
  markets without a unique perspective or strategy.
• 11:18
  But what's at stake for Venture if people continue to fail to quote unquote show their
• 11:22
  work?
• 11:23
  Just talked to somebody last night who's in a pretty high profile company, and the person's
• 11:27
  like, I don't know if I can recreate the set of conditions that got me excited about that
• 11:31
  company.
• 11:32
  It was not an obvious investment, but it's doing quite well.
• 11:34
  And he's just like, I don't really know why or how I said yes to that company.
• 11:39
  And truthfully, I often look at some of our best investments and I'm like, I liked that
• 11:43
  person.
• 11:44
  It seemed to make sense.
• 11:45
  Maybe if I'd met them on a different day, I might have reached a different conclusion.
• 11:49
  And I think the importance of show your work is at least it gives you a chance to recreate.
• 11:53
  This is why I try to capture so much of this information at the moment that we're investing
• 11:57
  or before we've invested, because hindsight has a way of rewriting the past.
• 12:01
  People would ask me, well, what did you see in that person at a series c company?
• 12:04
  I'm like, I kind of remember, but I'm probably romanticizing what I thought because the
• 12:09
  company's been successful.
• 12:11
  Or I'm filling in the gaps to make it sound like I maybe knew more than I thought I did
• 12:14
  because the company's been successful and the person's asking me a question and demands an
• 12:18
  answer.
• 12:18
  And I found that like part of the reason I try to capture more of thoughts upfront is I don't
• 12:22
  actually know what's gonna happen.
• 12:23
  And that's one of our summer projects actually, to look at a bunch of the stuff.
• 12:26
  Yeah.
• 12:26
  And just think how good are we in the moment at predicting.
• 12:29
  And I think if we were better, the returns in venture would be much higher.
• 12:33
  For those specific founders that you're thinking of, Charles, I assume first time
• 12:36
  founders
• 12:36
  Many
• 12:36
  of them.
• 12:37
  That you were really taking a bet on.
• 12:38
  I just think in most cases, there was something about that person that made me believe that
• 12:42
  they had a ton of untapped potential, and that putting them in the CEO and founder position
• 12:47
  was gonna tap into something that they hadn't had to access in their previous jobs, and that
• 12:52
  I had a really strong conviction that that was the case.
• 12:55
  For our last topic, in his recent Strathecari essay, Ben Thompson challenges the view that
• 13:00
  advertising was the Internet's original sin, arguing that it was a necessity for a human
• 13:05
  centered web.
• 13:06
  Mark Andreessen shared the same sentiment from his reflections on early attempts at building
• 13:10
  browser native payment layers.
• 13:11
  This model works when the purpose of ads was to monetize eyeballs.
• 13:15
  But with the rise of AI agents, it represents a new challenge since they don't view ads, and it
• 13:20
  really threatens the incumbent ad supported model.
• 13:23
  While Microsoft's vision of this open agent web enables agent website interaction, we still
• 13:28
  need an AI native payment This would be a built in system that allows AI agents to seamlessly
• 13:34
  make, receive, or initiate payments without human intervention or legacy infrastructure,
• 13:39
  like credit cards or banking rails.
• 13:40
  And so Ben suggests that incorporating stablecoins or microtransactions between agents
• 13:45
  and creators is the solution, and that without this layer, the agentic web could be dominated
• 13:50
  by a few platforms instead of remaining in an open ecosystem.
• 13:54
  First, do you think advertising was the original sin of the Internet?
• 13:57
  I don't think it was the original sin.
• 14:00
  I largely agree with this article, which was if you think about advertising as a solution to a
• 14:05
  coordination problem, you have producers making content, you have consumers who want to be
• 14:09
  entertained, and you need a medium of exchange of value.
• 14:13
  Obviously, you could have had subscription, and I think the piece is a good job of articulating
• 14:17
  what does a subscription actually represent.
• 14:18
  Went down this rabbit hole a few maybe, like, eight years ago telling people, like, a
• 14:22
  membership and a subscription are actually different value exchanges.
• 14:25
  You can be a member of a club or a program, and your financial support makes you feel good and
• 14:30
  the value exchange is a tote bag.
• 14:31
  The tote bag isn't worth the hundred bucks, but it's a symbolic gift and it's a signal and I
• 14:35
  won't go down that rabbit hole.
• 14:36
  But I think advertising was probably the most effective way because it made the content free
• 14:41
  for the reader, which led to a great deal more consumption.
• 14:45
  If you think about it being a sin, well, what are the taxes we had to pay?
• 14:49
  Well, the taxes were occasional privacy blowups like Facebook beacon and all of this stuff as
• 14:54
  people started to see that, oh, the reason this stuff is free is because we're doing this
• 14:59
  stuff, which we now have kind of accepted as just like generalized ad targeting.
• 15:03
  And SEO and the stammification of the Internet is the other byproduct where because traffic
• 15:08
  and advertising were linked.
• 15:09
  But I think for the most part, Google has continued to fight the war on SEO spam, maybe
• 15:14
  not effectively as they did ten years ago.
• 15:16
  And I think the thing that was most interesting to me about this article is, you know,
• 15:20
  sometimes an idea gets in somebody's brain and it's really sticky, and this is, I think, why
• 15:25
  Marc Andreessen is so obsessed with crypto.
• 15:27
  You know, if you've heard him speak, he often says, we've never had Internet native money.
• 15:30
  And I'd be like, hey, if I had been working on building a browser, I would have said, you
• 15:34
  know, Internet native money would make a lot of these things a lot easier.
• 15:38
  And I've met some agentic payment companies, and I think a lot of them are looking into
• 15:44
  future.
• 15:44
  I fear, as I often do, that they're maybe a little too early because I think most people
• 15:49
  are not yet ready to deploy an agent on their behalf to go spend real money.
• 15:55
  But I would argue there are lots of other ways you could solve this problem.
• 15:57
  You could give your agent the equivalent of a secured credit card or a BREX card.
• 16:02
  Right.
• 16:02
  We have all these other questions.
• 16:03
  What are chargebacks, disputes, fraud?
• 16:05
  If I authorize my agent to go buy me a pair of shoes and at checkout, Zappos is like, hey, for
• 16:10
  5 more bucks, you could have a pair of socks.
• 16:13
  Do I want my agent making that decision?
• 16:14
  Like, I don't think we're gonna go from old school web to agentic web, and the first thing
• 16:19
  we're gonna do is go have agents buy first class plane tickets.
• 16:22
  Right.
• 16:22
  I think the first thing is gonna be, hey, agent, go book me a restaurant reservation.
• 16:26
  Hey, agent, go go do these, like, relatively low cost, low risk things on my behalf.
• 16:32
  And people are doing some of this now.
• 16:34
  And I think we'll build up the payments, and then we'll have to figure out what do you do
• 16:37
  with authentication, how much discretion do you give an agent, how do you settle disputes.
• 16:40
  You know, all the things that Visa, Mastercard, and, like, the payment rails have all had to
• 16:44
  figure out for credit cards.
• 16:45
  We will have to figure these out for agents if we do it, whether it's with cryptocurrency,
• 16:50
  stablecoins, or with traditional like, we're gonna have to figure all this stuff out.
• 16:53
  There's simply no way.
• 16:54
  It's funny because New York I'm sure San Francisco is the same way.
• 16:57
  It's notorious for bots basically running
• 16:59
  their Totally.
• 17:00
  System, which runs.
• 17:01
  And so I tried to create my own to basically vibe everything.
• 17:04
  I used had GBT use Cursor and Twilio to basically try and get me an in, and I tried to
• 17:09
  construct a way to drop me the low hanging fruit reservation to pressure test how good is
• 17:12
  this.
• 17:12
  It got me halfway there, but then when it got to the points of checking, are you a bot?
• 17:17
  I kind of thought, if I could just give that security clearance to it, I probably would have
• 17:21
  to just let it go.
• 17:21
  Like, I trusted it at that point.
• 17:23
  But what are the chances that it gets me the wrong reservation?
• 17:25
  There's also credit card payments that need to be put down in order for you to get your
• 17:29
  reservations.
• 17:29
  What if it grabbed me the wrong reservation?
• 17:31
  I forgot about it.
• 17:32
  I'm now on the hook for 400, 5 hundred dollars.
• 17:34
  It's not frictionless yet.
• 17:36
  And then separately, just from thinking about SEO and digital marketing, if we do end up
• 17:41
  migrating to an open agentic web, it would completely upend everything we know about
• 17:46
  customer acquisition and digital marketing.
• 17:48
  So you give all the growth teams and growth shops that have been built to optimize for
• 17:52
  first click in digital marketing.
• 17:53
  I mean, you're trying to stay ahead of the curve as a marketer, what do you think you need
• 17:58
  to become really smart on in this environment?
• 18:00
  And I actually fear that the answer is that it might not be relevant because you also might be
• 18:04
  replaced by an agent.
• 18:06
  Well, think we've spent a lot of time figuring out how to influence humans.
• 18:09
  And the question is, will agents be subject to the same influence as humans?
• 18:13
  It's like photography, clever copy.
• 18:15
  So I think eventually, we're gonna have agentic marketers whose job is to influence agents.
• 18:20
  Yep.
• 18:20
  And the thing I think about a lot is, you know, there's a lot of stuff that's produced on the
• 18:25
  Internet that's designed to be consumed by humans.
• 18:27
  Visual content, written content.
• 18:28
  If agents do more of the work and spend more of the time online, like, where does all this
• 18:31
  stuff go?
• 18:32
  Yeah.
• 18:32
  Who finds it and how do they find it?
• 18:33
  And how do you replicate surprise and delight?
• 18:36
  Yeah.
• 18:36
  Which is typically how I find most of the things.
• 18:38
  I'm scrolling, I never even thought I needed this.
• 18:39
  And then because I saw it, now I do.
• 18:41
  I've talked to a bunch of people who are on the forefront of agentic commerce, and these are
• 18:46
  all the questions, and as far as I can tell, I've heard a lot of really clever hypotheses,
• 18:50
  but I'm not sure anybody's actually far enough along in solving this problem to have a true
• 18:54
  point of view.
• 18:55
  We live in interesting times.
• 18:56
  Charles, you've also talked about how AI companies are still trying to crack
• 18:59
  monetization.
• 19:00
  There seems to be this rekindled interest in fintech, increased access to stablecoins, shift
• 19:04
  away from traditional SaaS models.
• 19:06
  Does this point to the next generation or evolution of AI companies?
• 19:10
  Most people
• 19:11
  I've talked to seem to think that this usage tokens model is a bridge to wherever we're
• 19:16
  headed.
• 19:16
  It's not, as they say, it's not the AdWords moment.
• 19:19
  It's not the big unlock for AI.
• 19:21
  And I think that's probably true.
• 19:23
  I also have found that these business models, once they stick, tend to stick around longer
• 19:27
  than people like.
• 19:28
  Ah, proceed is a hack.
• 19:29
  It's a hacky way to, like, get at.
• 19:31
  And it kinda stuck around for a really long time.
• 19:33
  On prem stuck around.
• 19:34
  It's still around for some software products.
• 19:36
  So I'm not sure what the end state business model will end up being for a lot of these AI
• 19:41
  products, but I suspect that token consumption and token usage and sort of usage based pricing
• 19:45
  will be around longer than people think until and unless we find something that we think
• 19:49
  works better.
• 19:50
  And what that something is is unclear to me at the moment.
• 19:52
  More to come.
• 19:54
  We are all out of time for today.
• 19:55
  Thank you, Charles and Leon, for stopping by.
• 19:58
  To our listeners, please like and subscribe, and we will talk with you all next week.
• 20:02
  Thank you.
• 20:03
  Bye.