YT YouTube2Text

YouTube2Text → Transcripts → The AI Crash Is Over. Here's What Comes Next

The AI Crash Is Over. Here's What Comes Next — Transcript

by Jordi Visser · 8,861 words · 1,393 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00Making it through the summer.
0:02Uh
0:03rally continues into all-time highs. Val
0:06falls off sharply.
0:07Uh the biggest question I've been asked,
0:09and as I mentioned last week, I'm
0:11spending a lot more time with uh RIAs.
0:15And
0:16also with the institutions, and the
0:18biggest question
0:21I received post uh let's say situational
0:24awareness in the
0:25the
0:26breakdown in uh in factor val was how to
0:30hedge AI going forward. So, I'm going to
0:31go through that. I put together a prompt
0:34for you guys uh
0:35on something connected back to the uh
0:40thematic portfolio, so you can run that.
0:42I'm also going to start to hopefully
0:46give you guys some stuff on agents. Uh
0:49Gavin Baker interview, uh I'll basically
0:52make sure uh that I emphasize this is a
0:56must-listen to, regardless of your views
0:59on AI. Uh I've said a lot of these
1:02things,
1:04but I think, as I take you through it,
1:06um it's really important to make sure
1:08that you at least hear what he's saying
1:10based on a trip out to Silicon Valley.
1:12Uh
1:13I didn't get a chance because I record
1:16on Fridays during the summertime most of
1:18the time. That'll probably be the case
1:19going forward, but
1:21after I recorded um the US intervened in
1:25dollar-yen, and I will uh give you my
1:28thoughts on that because there was a
1:29major reaction in gold this week.
1:32And uh I think this is all going to lead
1:34into Bitcoin, uh and I'm going to go
1:37through why and how
1:41what my Substack that I wrote last week
1:43kind of connects everything back to it.
1:44So, um last week I talked about uh
1:48the bounce that had occurred in the
1:52final 2 days post the
1:55Citadel buying from situational
1:57awareness, and how it reminded me of
2:00prior panic situations where we finally
2:03had a cleansing event and that you can't
2:06especially in in the moves that we had
2:08seen that were historic with the
2:09volatility with the factor drop that
2:15you had to look for follow through days
2:18and basically what I wanted to see was
2:20two conditions one hold the the bounce
2:25pretty much across the board and then
2:27secondly I wanted to see factor
2:29volatility in the volatility stuff come
2:31down to indicate that people were either
2:34a hedged or they had taken their gross
2:35down. Now I had seen the data that the
2:37gross was down but
2:39we obviously got
2:41almost all of that. So
2:44on the queues I posted this in X but the
2:48four-day rally which included one day
2:51last week and the first two days of this
2:53week ended up getting close to 10% and
2:57you can see that over the course of the
2:58last 17 years since the great financial
3:01crisis ended you can see these days over
3:05time uh
3:08and a lot of these are obviously very
3:09famous bottoms in terms of liberation
3:11day and obviously COVID
3:13also this one here in 2022.
3:16This one here
3:18was a let's say a bear market rally in
3:21early 2022. The difference was then
3:25we were below the 200-day moving
3:27average. This is something I brought up
3:29before bull markets stay above the
3:31200-day moving average which is pointed
3:33upward. That's what we did in 80 plus
3:36percent of the AI thematic names which
3:38is why to me I was looking for where to
3:41put money to work. That's why we went
3:43through it that way. In terms of the S&P
3:45rate of change four-day rate of change
3:47the biggest one over the last year of 6%
3:52and just to give you an idea how vicious
3:54that rally was, we had been trading in a
3:57range. So, every week that I came in
3:59here, I kept saying the same thing,
4:01which is this is a pennant flag
4:03formation, whatever you want. And to get
4:05bearish charts like this is a very
4:06dangerous thing, but in particular when
4:09you're seeing Qs fall and you're seeing
4:11factor stuff collapse, you're looking
4:13for contagion on this, which is very
4:15difficult to happen when
4:17you have earnings growth at
4:20post-recession type times. Well, I'm not
4:22going to go through all of it you guys
4:23had, but the range that we were in
4:26basically from May
4:30to July. So, we're talking about 1 to
4:33almost 3 months, a 3-month pennant.
4:37And guess what? The rally we had was 6%
4:41bigger than the entire range that we saw
4:43for that time period.
4:45That is not something to fade. Um
4:48the S&P for the week up 3.6%.
4:52Qs up 5%.
4:57The thematic portfolio was up 6 and 1/2.
4:59It was the second biggest week of the
5:03year.
5:05In terms of where we stand year-to-date
5:07now in the S&P, again, I'll say this
5:09over and over again, this is a bull
5:10market.
5:12Every sector is up on the year. The S&P
5:14is up 13%.
5:18Pennant formation
5:20in there for a solid 3 months
5:23breakout higher than this. We not only
5:26hold after the Monday Tuesday, we make
5:28new all-time highs
5:30today.
5:32IWM made new all-time highs this week.
5:35New York Stock Exchange Composite made
5:37new all-time highs this week.
5:39The equal-weight S&P made new all-time
5:41highs this week.
5:43In terms of the KOSPI cuz I had a lot of
5:44people reaching out saying this is still
5:46a problem there. This is not a problem.
5:48The KOSPI is up 61% year-to-date. This
5:51is the KOSPI 200.
5:53>> [clears throat]
5:53>> The technology sector is up 116%.
5:56This is after a collapse, okay? And I'm
5:59not going to go through today, but on
6:02days now where SK Hynix is down
6:04slightly,
6:06the name the sectors that I showed, the
6:08machinery and the construction side,
6:10have
6:11been um leading on the way up and
6:14holding their gains. So, again, this is
6:16what long short is about. You get
6:18deleveraging on both sides and at some
6:20point you
6:22start to go the other direction.
6:25This is the DAX breaking out to new
6:27all-time highs.
6:30This is the FTSE breaking out to new
6:32all-time highs.
6:34Here's the big thing for the week. So, I
6:36showed you guys this. This is the
6:38market-neutral momentum. This is the uh
6:42So, this is the sector-neutralized
6:43market momentum. This is market
6:45momentum. This is technology momentum,
6:47and this is the industrials momentum.
6:50And if you look at all of these, the
6:52last 6 days,
6:55the drop-off in volatility says that
6:58funds are hedged.
7:01Here is the
7:037-day realized vol of the tech
7:06momentum [clears throat] vol.
7:08Uh basically collapsing,
7:12which is indicative of what I just
7:13showed. But, the what the main point is
7:16that we've seen factor volatility drop
7:18off. That is usually what happens after
7:21this. And as I've said, it takes a long
7:23time, particularly when you're below the
7:24200-day moving average. You've knocked
7:26people out. Um anyone that's vol
7:28targeted just cannot lever back up. Any
7:31of the pods that were hurt, that had
7:33their risk management either reduced,
7:36it's the summertime. This is why we're
7:39going to rally on lower volume,
7:41and everyone is going to be put in a
7:43position of chasing. They don't want to
7:45chase, they're going to read about
7:46seasonality, they're going to read about
7:47September, they're going to read about
7:48the midterms and all of this stuff.
7:51And in my opinion, I've heard enough
7:52podcasts this week from quote-unquote
7:56market people who
7:59are in the game, I think, of being
8:01bearish. Uh
8:04they remain bearish in the rally and
8:07they're looking for a false breakout.
8:08So, we'll see if we get that, but
8:11this is really hard to be bearish on
8:13this stuff when this is going on.
8:15Um the VIXY Q for single name ball
8:20has fallen sharply, too, and it's
8:21basically in this period now all the way
8:23back to where it was before this whole
8:27thing
8:28rallied. Not where the panic started,
8:30but where it goes. So, we've seen mean
8:32reversion now on the ball side. In terms
8:34of the VIX,
8:35you're at the lows since January. So,
8:38for everyone who reached out to me
8:39saying the contagion was coming, they
8:41were buying S&P puts,
8:43uh
8:45you not only lost on the direction, you
8:46lost on the ball and the ball never
8:49really in was indicative of the panic
8:51that we saw with factor volatility at at
8:5345-year highs on some measures.
8:56Uh we never got it in the S&P.
8:58CDX, um
9:01I talked last week about the silliness
9:03in the hyperscalers and how the
9:06companies with the best balance sheets,
9:07people are
9:08looking at their CDS uh and especially
9:11when they go into X. Here's what the CDX
9:15index did, basically the same as the
9:17VIX.
9:19There's no credit issues at all.
9:21Earning season, uh I'm not going to show
9:23you this number, but you know it's up
9:25near 50% in terms of earnings growth.
9:27Sales growth, though,
9:29uh
9:30we're just killing it.
9:32Killing it.
9:33And as you'll notice, this is across
9:35sectors. Um the AI trade from an
9:38earnings basis is just taking everyone
9:40up with it. Here's the revenue growth
9:42now up at 15% and I just want you to
9:45look that since Q1 of 25
9:48basically this has gone on. Now, what
9:50happened in Q1 of 25?
9:52Donald Trump took over as president and
9:55sentiment changed on the business side.
9:57He put tariffs
9:59in place. Everyone got worried, but the
10:02AI trade was now in full force.
10:04Uh and that's what is driving this. This
10:07is the certainty over the AI CapEx. So,
10:09these are the CapEx numbers basically
10:11driving this. It's having a huge impact.
10:13Margins
10:15last year at this point 13%. This year
10:1717%. And again, not just technology, it
10:21is
10:22margins are increasing and as AI
10:24adoption and AI agents go, you're going
10:26to see more of this.
10:27Uh the breadth
10:29again, I highlighted last week that it
10:32was a good sign that we basically had
10:34two prints higher than these two
10:36and we followed through this week.
10:39The 50-day is up to the highest level
10:41since July 14th and most importantly
10:44the 200-day slope
10:4780 plus percent pointed upward.
10:51Uh this has been the best quarter for
10:53years
10:55in terms of the earnings
10:58estimates beats.
11:02If you don't believe um
11:05well, when I If you don't believe the
11:06earnings
11:07then look at the revenues. They continue
11:10to move higher and this is the PMI which
11:13soared higher. I can remember it was a
11:14year ago PMI was below 50. I was writing
11:17papers on the
11:20PMI heading up towards 60 and what that
11:23would mean all on the back of the AI
11:24trade. Now, it's fairly consensus, but
11:28we're almost at 60 now and more
11:30importantly
11:31uh
11:33this implies that we're going to get
11:34closer to it. So, we continue to move
11:37higher, revs keep going up, and we got
11:39durable goods this week, non-defense ex-
11:42air,
11:43again, up at a level that
11:46is very close to 60. So,
11:48again, for anyone who has faded this,
11:50it's not just the earnings reports now,
11:53you're seeing it at the PMI level,
11:55you're seeing it global stock markets
11:56around the world, and you're still
11:58seeing revisions go higher.
12:00Here is the AI thematic portfolio with
12:03all four parts. Remember, you've got the
12:05Morgan Stanley one here overlaid with my
12:08100 name one. The Morgan Stanley one was
12:10just created later. For those of you
12:12want to trade it, you can call up Morgan
12:14Stanley desk, they have it. You've got
12:16the concentrated portfolio, which has 10
12:19names, and then the more global
12:21concentrated portfolio, which has 25.
12:23Again, these were put together so that
12:25you can have an index.
12:27I think if everyone learned anything,
12:29the indexes are really important when
12:31you have this much volatility. So,
12:34I want to get into the hedging side down
12:36some of the conversation. So, I showed
12:37this slide last week. Uh I believe the
12:39market is changing structurally, and
12:41this is never going back to the way it
12:43was before, because AI is compressing
12:46economic time.
12:47Um I think you're going to see that with
12:49the rest of the year. I'm going to go
12:50through the payroll numbers from
12:52yesterday, and I want you to reach out
12:54to every one of your favorite economists
12:55and ask them
12:57why they're not seeing the weakness.
12:59This is not about people um necessarily
13:02being replaced by jobs. This is not
13:03about jobs being lost,
13:06but this is about wages. This is about
13:09the fact that digital employees are
13:10here, and they're coming
13:12incredibly quickly, because the adoption
13:14pace is picking up. I'm going to take
13:16you through Palantir, which is um
13:19a name that I own personally, but also a
13:21name in the model portfolio,
13:23which just showed their corporate
13:24numbers. Since we can't get Anthropic
13:26and OpenAI in terms of numbers that
13:28we're getting released, you can look at
13:29Palantir's. These numbers are insane in
13:32terms of the adoption. When the adoption
13:33picks up, the AI agents are going to
13:35come and the AI agents come, not only
13:37are you going to have less hiring, you
13:38will start to see more job losses in
13:40public companies. You'll have job
13:42creation in private AI native companies,
13:45which I'll go through. Terminal value
13:47gets harder to ask to estimate, which
13:49makes it more difficult that companies
13:51all of a sudden can go from trading at a
13:5430 PE to a 20 PE in a very short amount
13:56of time. We've seen that in many many
13:58places and I think you're going to
13:59continue to see it. I think Google's
14:01kind of going through that right now. Um
14:04I won't read all of these, but all of
14:05these things, including this one, which
14:07Gavin Baker brought up, too.
14:10The the fact that everyone
14:13has access to it and everyone can create
14:15a backtest, everyone can create an
14:17optimized portfolio, everyone can find
14:19the names that are working, leads to
14:21more momentum, it leads to crowding. And
14:23again, it leads to bubbles, parabolas,
14:25and speed crashes. Exactly what we had
14:27this year, what I wrote about. The speed
14:29crash has occurred, it's over. The
14:31bottom is in as far as I'm concerned on
14:33a probability basis. If we take out the
14:35lows, I'd be surprised, uh mainly
14:38because I continue to believe people are
14:41way too bearish relative to the reality
14:43of the earnings growth that's coming. I
14:45also don't believe you're going to ever
14:47see that type of move again uh that we
14:49saw in the first quarter, but you never
14:51know. So, with that hedging side, how do
14:53you deal with bubble bubbles, parabolas,
14:55and speed crashes?
14:57Well, those are all based on
14:59price and time, which I started to show
15:01you last week, but I'm going to go
15:02through one of the demos in terms of how
15:04I how I'm handling this. And for the
15:06subscribers, this will become part of
15:09the way that I'm going to identify
15:10things. I built turbulence models. The
15:12turbulence models because everything was
15:14so correlated, they never really showed
15:16leading problems. The breadth did, and
15:18that's how I'm going to kind of go
15:19through this. So, I try to learn from
15:22what happens and look for the signals
15:23that will warn you. This is one signal
15:25that will warn you when factor vol
15:28continues to move higher and break away
15:31from index vol. The reason factor vol is
15:34is so much different than say single
15:36name equity vol, which also broke away
15:38and I showed last week,
15:40this has got a long short component of
15:42it and this is directly related to gross
15:44leverage. So, the great thing about
15:45factors are as gross leverage is going
15:48to all-time highs and I'm making the
15:50call that we are at peak gross leverage
15:53and that we will see leverage come down
15:55over the course of the next few years
15:57because of what I showed before that AI
15:59is compressing time, it's making it more
16:01crowded and it means the sharp ratio of
16:03this stuff is going to come down. I
16:04think factor vol is structurally going
16:06to be higher. I don't think we're going
16:08back down here. Maybe we come back down
16:10into 25, but if we spend now the rest of
16:12this year in the 30 to 25. And again,
16:15this is an equal weight of four separate
16:17factors. It's not just momentum because
16:19I have value in there, I have quality
16:21and I have beta. And so, I'm averaging
16:23them all in there. You can see the
16:24difference in the breakaway from the
16:26S&P, which means using index
16:29puts is not going to work the way that
16:30it used to.
16:31So, here's what I am using and this is
16:33the price in time side. So, right here
16:36when we when we got to 50% or almost
16:3950%. This is a 50-day rate of change on
16:42my thematic portfolio. So, again, for
16:45those of you on the AI side that are
16:47more tech focused or you're more
16:48industrial focused or you're more
16:49utility,
16:51you need this tool. You need to have
16:53access to this data and the reason is
16:56because there are no indices that are
16:58diversified to this level. This is
17:00global. This is across sectors. This
17:03involves 100 names that are involved in
17:05there. There are chemicals, power
17:07companies, industrial companies. So, it
17:09is diversified.
17:1150% move
17:13over 50 days. So, 50 trading days, which
17:16means slightly more than 2 months. To be
17:19up 50% you're going to fall. Now, here
17:22is the historical levels. At 20% I think
17:25the risk grows and I think that's when
17:27you need to combine it with
17:30this.
17:31These happen to peak at the exact same
17:33time.
17:34So, rate of change does not mean that
17:37the thematic trade goes down. It just
17:39means you start to run in. So, I've
17:40always used rate of change. In the past,
17:43before we had exponential
17:46uh innovation, in 2007 my model gave a
17:49recession
17:51uh warning or not it gave a recession
17:53signal in September of '07.
17:56So,
17:57in September of '07 is when the warning
17:59went out that we were going to have a
18:01recession. And the reason was because
18:03year-over-year S&P, the 252-day rate of
18:07change of the S&P,
18:09had basically started to turn negative.
18:12Once that happens, historically, that
18:14means we're at a bad point. When you add
18:17in corporate profits, you add in jobless
18:18claims, again, this is before
18:20exponential AI or before exponential
18:23innovation with software. So, for me,
18:26this is the way that I'm replacing this.
18:28I'm looking for
18:30the rate of change to be up at a level
18:32which doesn't make sense, which shows
18:34the crowding, it shows the AI agents, it
18:36shows everyone pushing everything. And
18:38then I'm looking for breath to be
18:39breaking down within inside the 100
18:41names. You need a diversified portfolio.
18:43So, this is the way I'm going to measure
18:44it. I gave you guys the warnings back
18:46here. We didn't make a high in the index
18:49until right around here.
18:51Then we didn't violently fall on the
18:53memory names and stuff here. So, I
18:54talked about selling Micron in here.
18:56This is when I got rid of all of it. It
18:58was early.
19:00I rotated to silver, to Bitcoin, and to
19:04Eli Lilly, and to some other names.
19:07The main point was those names in the
19:09month of July, go look what they did.
19:12Silver and Bitcoin
19:14Well, silver was flatish. Bitcoin was up
19:17a good amount. Lilly was up. You're
19:19trying to navigate where you're still in
19:21a position to buy when you get the puke.
19:23So, this is the way it looks and this is
19:24the way I'm going to go through it. This
19:26is the whole year actually.
19:28It's the It's the rate of change. So,
19:30once you break above
19:32this level, then the warning sign
19:34starts. So, we broke above this level.
19:36We started peaking in terms of the
19:40the breadth, we were up at 90% believe
19:44it or not. We're above the 20-day moving
19:46average and then gradually we broke
19:47down. You can see how this goes. Now,
19:50the four-week average peaked here and
19:52rolled over. So, as time goes on, I
19:55believe this is going to happen now more
19:57often and I think this is the way you're
19:59going to have to deal with it.
20:00I've had a lot of subscribers reach out
20:03about a bunch of names that just haven't
20:05bounced yet. Things like Camurus, I own
20:08this one. Fluence, I own this one.
20:11Entegris did go higher. I own this one
20:13and Palantir, I didn't include the price
20:15today. It's all the way up here now.
20:18For all the Palantir haters and all the
20:20short people and the people that go,
20:22you're missing something on this one.
20:23It's the only software name I own
20:27personally and then you've got Cadence
20:29and Synopsys that are part of the
20:32the thematic portfolio that are software
20:34design. I'll probably be talking about
20:36them soon
20:37because there's been some stuff. So,
20:39that's the way that I'm kind of working
20:41through with people to automate things
20:43on the hedging side, to use my index for
20:46that, but also to use the breadth and to
20:48use the different parts. This is the
20:50thing I was saying on an earnings
20:52mosaic. So, what I did create and the
20:54prompt will go up over the weekend.
20:57I created a prompt for you guys that
20:59will basically spit out something like
21:00this. You can use it in Gemini. You can
21:02use it in ChatGPT,
21:04but basically this will use and this is
21:06to give you a sense where if you take my
21:07technical sheet and you update you put
21:11the technical sheet into chat GPT
21:14and then or Claude or Gemini or
21:16whichever one you want Grok and then you
21:19just run the prompt it will go out and
21:21get the earnings commentary from the
21:22week and put it together with a bunch of
21:25different things including kind of a
21:26look at a mosaic side. So this was from
21:29Eaton you had Eaton report this week.
21:32You also had Schneider report. They're
21:34both on the electric side. They both
21:35happen to be participants in Vera Rubin.
21:38So it will go through and it will look
21:40at and give you basically a recap of
21:42that with some of the most important
21:44commentary that was in there. I'm trying
21:46to help you guys especially as times
21:48were goes so this is what it does. This
21:50is what you'll see up there.
21:52I'm trying to keep you guys going with
21:53the AI side. Again, if you guys want to
21:56access to the tools, if you want access
21:58reach out to Mark Wailing, go to the
21:59website.
22:00We're doing more and more of these but
22:02in particular I'm doing more and more
22:03presentations for RIAs where I'm also
22:06doing regular calls with them. They're
22:08setting things up that way so you can
22:09reach out to Mark. I did want to
22:11reference this because this is where we
22:13we are in the agentic side. This is a a
22:16a text
22:17exchange with my son. Uh
22:20he's a entering his junior year at
22:23college. He is not a computer science
22:26major. He I put him through
22:29Python boot camp when he was a junior in
22:31high school.
22:33I have my AI chief of staff spitting out
22:35new sub agents on the regular. This is
22:37what he's doing at his internship.
22:39He even said I prefer chat GPT Gemini
22:41coding is insane though. I got the
22:43information from YouTube and AI.
22:46Your kids have to be doing this guys. He
22:48is guaranteed to have a job at whatever
22:50firm he works because he has already
22:52built a chief of staff. He's spitting
22:54out those new sub agents. There's
22:55absolutely no reason
22:57why your kids can't do this. So remember
23:00the knowledge brain thing that I put in
23:01there, I will be doing AI agent stuff
23:03for the subscribers. I will probably be
23:06going through some videos
23:07as we get into this year. I got a lot
23:10going on with the crypto side, but my
23:12goal is to really make sure that you
23:13guys realize that the agentic side is
23:15here and if you're not using it, you're
23:17falling way behind. So, let's go through
23:19the Gavin Baker side. Again, this is a
23:21must listen. The reason it's a must
23:23listen is cuz very seldom do you find
23:25someone who is this knowledgeable about
23:27the AI trade from a very high level,
23:30which Gavin is. He speaks about it
23:31regularly. He's asked questions on it
23:33regularly. He invests in it. It's very
23:36hard to find people that know this from
23:38the industrial side all the way out.
23:41I'm not going to read all these up. I'm
23:43just going to go through the speaker
23:44argues the AI sell-off is disconnected
23:46from the quantitative data. He says he
23:48has not found a real negative demand
23:50metric.
23:52So, I just want to make sure that you
23:54guys, if you listen to this,
23:57if you just listen to the first 5
23:59minutes of this,
24:00these are all quotes from the first 5
24:02minutes.
24:04Nvidia actually, as we record this, is
24:06at the lowest forward PE of the last 10
24:08years.
24:09I've bought a decent amount of Nvidia
24:11over the course of the last 3 weeks. It
24:14looked good for a while, broke back down
24:16again, now it's looking better.
24:18I believe with Vera Rubin coming out and
24:20with all the chatter on this and the
24:22fact that it's this cheap, but also
24:23going through the token per watt needs
24:26and realizing how quickly they're
24:28they're coming out with this stuff, I
24:29just can't see a sign with what GPUs are
24:32doing, the price of them which he goes
24:34through,
24:35that they're not going to benefit from
24:36this at this level. So, they've done
24:38their multiple compression, it's been
24:40significant.
24:42I just it's it's not it's not going to
24:44double for you, I'm sure, in in in a
24:45short amount of time, but I think when
24:48you're looking at kind of safer names
24:50that already have seen their multiple
24:51compressed, I still think Nvidia is one
24:53of the best ones. Um
24:56not a single instance of deceleration,
24:58nothing. In fact, every metric rate is
25:00accelerating. Whether you cut GPU
25:02availability, whether you cut GPU rental
25:04pricing, whether you cut the spot price
25:06of DRAM this month to grow, everything
25:07is actually accelerated.
25:09So, when you go through all of these and
25:12you hear what he says about Anthropic
25:14and you hear what he says about deep
25:16compute prices and the anecdotal stuff
25:18that he gets out there.
25:21I He's basically saying that these
25:24companies
25:26Microsoft, Meta, Amazon, the
25:27hyperscalers, the critical part of this
25:29because if you've been bearish on them,
25:30which most people have, they've called
25:32this a bubble, they've been wrong.
25:33They've said ROIC will not come through,
25:35they've been wrong.
25:37Maybe they'll be right. The CDS is wider
25:40for justified reasons cuz they're still
25:41leveraging up more and more and I'll go
25:43through the reason why it does make
25:45sense in a second, but
25:47you can't argue the fact that at this
25:49point there's a chance that they could
25:50see significant revenue very short-term
25:54if for no other reason because they are
25:56building so much compute and the value
25:59of that compute is going up and he talks
26:01about it here.
26:02The same type of B200 cluster that
26:04rented around mid $2 per hour 7 months
26:07ago is now expected to rent for just
26:10under $4 per
26:11GPU hour. So, remember all the stuff
26:15about these chips won't survive, their
26:17lifespan. We have a bottleneck in
26:21compute.
26:22So, you have to go through all compute
26:24has value right now regardless of where
26:27it is. Open source would love to be
26:29using this stuff. So, let's just
26:32forget all the stuff that you've heard
26:34on this.
26:35Don't get all bearish on this because of
26:37something that isn't true by the data.
26:40If you see this data start not to happen
26:42and that's what he goes through,
26:44it's like the breath that I have on the
26:46index. If one thing is going down,
26:48that's fine, but when people look at the
26:49token index and that's their thing and I
26:51say on this video, even the people at
26:54the place that provided say it's not
26:56bearish on
26:58on market. It just means more and more
27:01of the usage is going on open source,
27:03which makes sense.
27:05It doesn't mean it's bearish. It doesn't
27:07mean that Anthropic's revenues aren't
27:08going to grow, but very smart people
27:10that I've known in the business for a
27:11long time were saying that to me. So,
27:14we've had a point that if you can listen
27:16to someone like him and go through it,
27:18it's there.
27:20It's rewiring the market structure. So,
27:22he got into the same thing that I talked
27:25about.
27:26He even brought up the probabilistic
27:28Bayesian interpretation of the future,
27:29which I write about all the time and I
27:31talk about all the time.
27:32AI compresses economic time, but it it
27:35may also compress market time.
27:38The month of July
27:40for momentum was worse than any year
27:44over the data that I have.
27:46So, that's how we're compressing market
27:49time. Bubbles, parabolas, and speed
27:52crashes.
27:54Here are the key quotes from the
27:55transcript. You guys
27:59can pause it and look at it on your own,
28:01but here's the most important one. We
28:02are in a cute compute shortage.
28:05And at the same time, 500,000 people,
28:08maybe 250,000 are using a generative AI.
28:11So, my son is one of them. He's spinning
28:13out sub-agents.
28:15But only 250,000 of 8 billion people.
28:18What happens when we go from 500,000 to
28:211%? To 100 million? To 500 million?
28:24Guys, I say it all the time.
28:27There will never be enough compute,
28:29ever. We will continually be building
28:31compute. We'll be doing it in different
28:33ways. We'll continue to get more
28:35efficiency. Everything will keep
28:36happening, but nothing is on the horizon
28:38for that to go away. So, when we see the
28:41prices of GPUs on older stuff start to
28:44head down, at least we're at a point
28:46where the newer stuff is saturating some
28:48of the demand ahead of adoption, but the
28:50adoption is happening because of agents.
28:52And once you set them off, they're
28:55tireless workers. So, it is exponential
28:57against
28:59the physical side. No amount will ever
29:01be enough. The math behind hyperscaler
29:03CapEx. This is my post earnings paper
29:07that went out this week
29:08on the hyperscalers and the reason why
29:12you should not be playing for them to
29:14break apart.
29:15Now,
29:16I got a lot of questions saying does
29:18this mean you're bullish on
29:19hyperscalers? No, not relative to the
29:21infrastructure trade. Nothing changes in
29:24that front. I still think they're going
29:25to have heavy competition, and here's
29:27the other reason why.
29:29The hyperscalers quarterly free cash
29:31flow, when we think they're going to go
29:33positive as a group by Goldman Sachs.
29:36Look how long this is out. So, here's
29:38the problem.
29:40They're going to keep spending money on
29:42hoping they get the revenues in the
29:43door.
29:44I I understand they have the cloud
29:46revenues, but the cloud revenues are not
29:48going to be enough in my opinion. You're
29:50going to need to have products, you're
29:51going to have need to have other things,
29:53or you're going to get priced like a
29:54utility or real estate company, cuz
29:56eventually, when the compute isn't
29:59as
30:01outsized in terms of supply versus
30:03demand as it is now, and by then we
30:05could have a lot of innovations. We
30:08could have a lot of things on the
30:09algorithmic side. Things could change.
30:12There could be breakthroughs in hardware
30:13that when we go to uh humanoids and we
30:16go to cars and we go to all this,
30:17everything could change. Now,
30:19I still believe we're going to need more
30:21compute, we're going to need more
30:22hardware, but
30:24whether their businesses will be there
30:26is a question, and that's why their
30:27multiples should be compressed, and
30:29their CDS should be
30:32wider. It There's a risk there. So,
30:35it's [clears throat] not a
30:36a a risk that I want to
30:38go out there and trade, but if you're
30:39buying their paper
30:41and you want to
30:42buy some CD I of course you do. If
30:44you're trying to short the AI trade and
30:46you want some stuff out there, of course
30:47you go buy CDS in this, especially when
30:49it just keeps moving. Um I just want to
30:51bring this up quickly cuz I think
30:53[clears throat] this is a sign about the
30:55problem inside these companies. So, I
30:58also believe that this is a big deal
31:00what happened this week with Jeff Dean
31:02leaving.
31:03And also, Demis Hassabis kind of moving
31:06into another area.
31:08This is a big deal to me and this is
31:10because Google is the perfect example of
31:13a company where
31:15they released AlphaFold,
31:18which makes them no money.
31:20If it makes them some money, it's coming
31:22through something, but for the amount of
31:24people using it to release it free,
31:28it's probably created some conflicts
31:30inside cuz they are spending tons of
31:31money, they're taking tons of debt. They
31:33need actual consumer products,
31:35corporation, they need something to come
31:38through and I think this is where the
31:39people who are trying to leave a legacy
31:42on this technology, the talent, are
31:44getting frustrated with the bureaucracy
31:47at big companies. And I think you're
31:49going to see this over and over again.
31:51It's reason why I'm negative on all
31:53public companies by the time we get to
31:552030, the bureaucracy is going to be a
31:59killer. At the same time, you're going
32:01to have AI native companies growing
32:03rapidly.
32:04Okay, I just want to remind you on the
32:06memory side. I wrote this paper
32:09is in June, AI equals memory. This is
32:11after I had sold my Micron. This was not
32:14to buy anything. This was just to let
32:15people know that memory will remain in a
32:17shortage and it is something to watch
32:19when it comes to the trade, but it's
32:21also something to watch closely to see
32:23if we ever get to the point where
32:25catches up to compute. So, on the SpaceX
32:27earnings call,
32:29Elon Musk
32:31basically reiterated what I wrote in
32:32that paper. The limiting factor
32:33currently is memory. The memory output
32:35is increasing by around 20% a year. So,
32:38supply is increasing by about 20% a
32:40year. The demand is increasing by 200% a
32:43year, maybe higher.
32:45If you've got demand increasing faster
32:47than supply, economics 101 would suggest
32:48the price increases.
32:50Uh again, I'm bringing this up because
32:56watch memory.
32:58And this goes through more compute does
33:01not solve the bottleneck without more
33:02memory. So, the memory side is going to
33:04be needed and that's what I wrote about
33:05is that memory is needed for all parts
33:07of this when you start thinking about
33:09what AI agents need, but especially when
33:11you get into consumer agents, we need
33:13lots of memory on everything that we do.
33:16Palantir shares jump on uh
33:19otherworldly sales improved forecast.
33:23This is what I was saying. This is
33:24commercial revenue. This is the adoption
33:26side of companies.
33:28This is the growth rate. 54%, 64%. These
33:31are year-over-year. They're not only
33:33growing by this much,
33:34they're still growing faster.
33:37So, 149%.
33:39So, in Q2 last year, they had 306
33:41million of revenue. This year, it's 764.
33:43So, whether it's quarter-over-quarter
33:46growing at 20-plus percent or whether
33:48it's
33:49it doesn't matter. These guys are
33:51growing rapidly. And everyone is focused
33:54on them and commercial was not their
33:56business. So, it's starting from a small
33:58pace. This just shows the adoption pace
34:00that's going. They closed on 220 deals
34:02of at least 1 million, 98 deals of at
34:04least 5 million, and 78 deals of at
34:06least
34:0710 million.
34:11Only reason I brought this up is to just
34:13show you that the bear porn just remains
34:16out there.
34:17Here are new business formations.
34:20This is what the story is, guys. This is
34:22AI native companies. They're starting
34:24their own businesses. That's what I did.
34:27This is how This is what's happening.
34:29It's going parabolic with AI adoption.
34:32All started with Opus 4.5.
34:34The rise of million-dollar companies
34:36with just one employee. So, this is
34:39what's happening to disrupt the profits
34:41of the Fortune 500 companies eventually.
34:43It will take some time because these are
34:45small, but it's coming.
34:49The age of abundant intelligence and
34:50scarce Bitcoin. Two printing presses
34:53running at once. One dilutes the unit
34:55you measure wealth in.
34:58That is the debasement of your money.
35:00The other dilutes the durability of the
35:02corporate claims you bought into to
35:03protect that wealth. So, this one is
35:05from the government and the printing
35:06press.
35:07This is the AI printing press, which
35:08will disrupt public companies.
35:12That's in my Substack.
35:13Okay, 30-year yields. So, I put this
35:16line here cuz basically we took it out.
35:19This was a big story.
35:20Um
35:22I'm writing a paper right now on the
35:25JGBs
35:26as the widow maker from the 1990s.
35:30And how every time
35:32I get a technical chart or a person
35:35saying we just broke out, this is going
35:37to collapse,
35:38rates are moving higher,
35:40stocks pretty much almost always rally.
35:43Now,
35:44the reason is
35:46cuz the government doesn't really have a
35:48choice. So, if you're going to fight
35:50something
35:51and you want to be bearish on the
35:52equities because of bonds, the
35:54government, if the bond yields go too
35:55high,
35:57they clearly are going to do something.
35:59They've done it repeatedly. We know
36:01that. They did that on liberation day.
36:03Okay, we don't care if equities go down.
36:05Then when the bond market went down,
36:06they cared that equities went down and
36:07they had to do something. Here is the
36:09relationship between corporate tax
36:10receipts
36:13and household
36:14equities
36:16asset values.
36:19If you have asset values go down in
36:20stock market, you're going to have tax
36:22receipts go down at a time when we have
36:235 to 6% deficits. That's why Lyn Alden
36:27wrote about why nothing stops this
36:29fiscal train. It is still in place, guy,
36:31and we learn more about it this week
36:33because it's not just the US, it's the
36:35whole globe. So, when you see these
36:36charts and you got the US and you got
36:38the UK and you got JGBs and you even got
36:41bonds.
36:43They're all 30-year yields around highs
36:46of some level and what did happen so on
36:48Friday night Scott Bessent has to
36:50explain why he helped Japan strengthen
36:53the yen.
36:54So, I believe this is more a signal.
36:58Um Scott's obviously a savvy market
37:01person and he's trying to find a way to
37:03help the BOJ for a variety of reasons
37:06and one of the reasons is because they
37:08own a lot of treasuries and he doesn't
37:09want them to have to go in the market to
37:11sell the treasuries to be able to go out
37:13and
37:15buy the yen.
37:17So,
37:18first coordinated move since 1998.
37:22Long view, US props up the yen to
37:24support treasuries.
37:27Then Bessent is urging the Fed to help
37:29defend the Japanese yen using
37:32FEMA and hoping to expand the capacity
37:35of it. So, then you got the Wall Street
37:37Journal and Nick
37:39Timiraos
37:41Why Bessent is leaning on the Fed to
37:43help prop up Japan's currency. I want
37:45you to remember that cuz last week Kevin
37:46Warsh spoke. So, last week you had
37:48situational awareness. You had Kevin
37:50Warsh. Market got angry at Kevin Warsh
37:53cuz he didn't either raise rates or he
37:54didn't say he was going to raise rates,
37:56he gave no guidance. Nobody knows what
37:57to do. He didn't even say
38:00I think we're kind of getting the sense
38:03here that uh it's going to be really
38:04hard for Warsh to raise rates if the
38:07Fed's going in to help Japan and we need
38:09interest rate differentials there and
38:10he's actually going to the Fed and
38:12trying to see if they can raise the
38:13borrowing limit. Seems like the Treasury
38:15and the Fed might be a little closer
38:17than um they used to be.
38:20Uh then we go through and we're starting
38:23to talk about the unintended
38:24consequences and I think
38:26I think the joint US
38:29Japan support at the end is unusual. The
38:31way it is being financed is
38:32unprecedented and adds liquidity when
38:33the punch bowl of the US economy and
38:35markets is already overflowing.
38:39And he pledged to do whatever it takes
38:42to help.
38:45Again, I think the bazooka he doesn't
38:48want to do anything, but I think he's
38:49hoping we can stay away from 164 and
38:53maybe they'll raise rates in September,
38:55but if Warsh raises rates in September,
38:57then what does it really accomplish,
38:59which is one of the reasons why
39:01I'm shocked after this last 2 weeks that
39:04we still have a 45% chance of a rate
39:07hike in September, which is why I like
39:09all the debasement trades, as I talked
39:12about a few weeks ago, thinking that
39:14people were getting way too over their
39:16skis on rate hikes.
39:19Remember, a lot of places were talking
39:20about three, at least two, before the
39:23end of the year and now I'm now we still
39:26have one in change in there. We'll see
39:27where it goes. Uh, the policy fails to
39:31tackle the underlying problem can easily
39:32backfire. Not least drawing the Federal
39:35Reserve into exchange rate policy at an
39:36especially awkward moment. That's the
39:38awkward moment what I'm talking about.
39:40Um,
39:42Treasury has now spent all the ammo in
39:43their exchange stabilization fund.
39:46They can probably find other ways, but
39:47unless interest rate differentials
39:49change, it's all waste. So, we're back
39:50to the interest rate differentials.
39:52Then a former ex former Japan PM warns
39:56FX action won't save yen without 2.3
39:59trillion growth strategy. This is the
40:01run it hot into a scarcity of tools that
40:05the central banks can use. Can the US
40:08Treasury save the yen?
40:11Kevin Warsh is facing more than a
40:12communication problem. Um,
40:16I wanted to bring this up again because
40:19I think last week
40:20was a big week. Situational awareness,
40:24Kevin Warsh not raising rates,
40:26and then the Fed doing what they did
40:28with the BOJ, the first coordinated move
40:31since 1998.
40:33I think with all of that at the end of a
40:36historic month where we had deleveraging
40:38of significant size, there's a lot of
40:40similarities between October of '98 in
40:43there, a lot of them. And I believe when
40:46we look back as long as those lows are
40:47in place that that will be a time that
40:50we started to see a reality check
40:53that when the US is intervening in the
40:56Japanese currency market to help their
40:59treasury market, the fiat system might
41:02be in trouble. Uh
41:05especially when I go through some of the
41:07charts. So, we're 45% as of Friday for
41:10the Fed to raise
41:12rates, and and that's after the payroll
41:14number. And we're at 63% for the BOJ.
41:18Uh Darius Dale put this out on the
41:20quarterly refunding,
41:22the Fed has little choice but to
41:24continue monetizing US sovereign debt
41:26over time. The Basically, the refunding
41:29announcement, there's
41:31whatever Kevin Warsh people thought he
41:33was,
41:34again, this is why why nothing stops
41:36this fiscal train. It is we just have
41:39too big a debt, too big a deficit.
41:41Remember when Paul Tudor Jones went on
41:42and did the math and said there's just
41:44no way to solve this. They're running it
41:45hot, but they're not making any impact.
41:47So, here we go.
41:49During Donald Trump's tenure as
41:51president,
41:53July was the biggest 30-year
41:56yield rise in a month.
41:59So, back here is when he drew the line
42:01in the sand for the equity market. So,
42:03here's what happens to the S&P after the
42:05bond market ends up going higher.
42:08Bond market goes higher, guess what?
42:10Oh, you're starting to get the picture
42:12here. So, the S&P has rallied when rates
42:14go higher because everyone starts
42:16getting bearish on it and the Fed or the
42:18Treasury needs to step in. Nothing stops
42:20this train. We keep doing the same
42:22thing. So,
42:23gold is waking up that maybe Warsh is
42:25not going to be bad for gold. Maybe he's
42:28not going to raise rates. So, gold
42:31over the course of the last 16 years, it
42:34was the third biggest week for gold.
42:36Now,
42:37this was
42:39one of the largest printing press weeks
42:41of all time.
42:44This is when gold was out of control.
42:46This is a big week in gold, up 7.2%.
42:50Here's the chart. Broke through the
42:5250-day. Look how long we were below the
42:5450-day. We broke below the 50-day
42:56over here. We hadn't been back above
42:59since March. In fact,
43:01it was the third longest time that we
43:03had been below the 50-day moving
43:05average. Now,
43:06here are the returns 3 months out from
43:08the time of breaking back above. We'll
43:10see what this one is.
43:11But, year-over-year at the break point,
43:16most of the time these things were down
43:17big.
43:18Gold's in a bull market. So, again, I go
43:20back to what I said before.
43:22When you're in a bull market and you
43:25correct and then you consolidate and
43:27then you break back above the 50-day,
43:29things can run pretty hard. I think
43:31there is an absolute reason
43:33to be thinking that gold, silver,
43:36platinum, Bitcoin, all of them, assuming
43:39that the Fed is not going to cut rates
43:41in September,
43:43which again, I'm surprised we have 45%.
43:46We'll see what happens with the
43:47inflation data next week, but I think
43:50we're getting closer and closer to this
43:52and the market is telling you the
43:53signals where people like, well, why
43:55don't I just short Treasuries and go buy
43:56gold and go buy this? I think either way
43:59right now,
44:01I'm going to win because they're going
44:02to come in and stop it, but I don't
44:04think it can go down. So, here's what we
44:06got. Platinum breaks above the 50-day.
44:09And again,
44:11nice rounding chart. We'll see if
44:13there's follow-through. Silver finally
44:15closed above the 50-day today. I have a
44:18good-sized position in silver because of
44:20the AI side, but I like it now because
44:23of the the basement side. I showed this
44:25last week.
44:26We're at the highest dollar long
44:28position. So, the dollar has already
44:30broken down, but this is just fuel for
44:32the fire for gold and
44:35and silver and platinum and Bitcoin. And
44:37again, just a reminder, still 44%, not
44:39just that, we still have more
44:41tightenings in.
44:42I I think there's I mean, I think all of
44:44these should be coming out, personally.
44:46That's just me. And the reason is
44:49this inflation thing is driving me nuts.
44:51We have one inflation data point, which
44:53is upward. Um
44:55this is the Truflation core number
44:59relative to the CPI core number. You
45:01cannot get away from the fact that it
45:03has led and it has gone.
45:05This is a violent move lower over the
45:07course of the last 4 weeks. So, the
45:09year-over-year Truflation number just
45:11continues to fall. This is on the same
45:13scale, people. So, where this was around
45:16the same at the beginning of the year,
45:18you've had a slight uptick here, and
45:19you've had this collapse. Go read how
45:21they do their work.
45:23If Kevin Warsh is looking for better
45:25data points than just taking the way
45:27something was calculated 50 years ago,
45:30at least give Truflation a chance, along
45:33with trim mean, along with the other
45:34ones that I've shown. I think the
45:36inflation thing is a disaster. So, let's
45:38go the other way.
45:40I'm sick of economists and the payroll
45:42side.
45:43The job situation is not good.
45:46Yes, there are low claims. And the
45:49reason there's low claims, part of this
45:51is the labor force is coming down, but
45:52they never talk about that. So, the
45:54supply of labor is going down because
45:56the labor force participation rate is
45:58coming down sharply, has come down
46:00sharply this year. Well, if there's less
46:02labor supply, shouldn't that mean wages
46:03are going higher?
46:05Here are the wages.
46:06So, this is the quits rate and the
46:08Atlanta Fed wage side. On both measures,
46:12we are back to 2019 levels. You can't
46:16tell me the labor market is good with
46:18those metrics. Let's just take the
46:20average hourly earnings that we just
46:22got, which is now down to 3.2%. It's
46:24also the white line is also where it was
46:26before 2020.
46:28Overlay it again with the Atlanta Fed
46:30median wage growth, right here.
46:32You can't tell me the labor market is
46:34good.
46:35Let's go a different route. Let's now
46:37take the aggregate payrolls.
46:40This is the aggregate payrolls,
46:42which is taking weekly hours, average
46:45hourly earnings, and the number of job
46:47creation. I've talked about the fact
46:48that the only jobs that have been
46:50created over the last 18 months are in
46:51healthcare. When you strip out
46:53healthcare, there is no job creation.
46:54That is not a good jobs market. AI
46:57agents are have or AI is having an
46:58impact on hiring. It's not yet had an
47:00impact on firing, but again, if we're
47:02having job creation with entrepreneurs,
47:04and we're gradually losing people on the
47:06other side, which is the way I think it
47:07goes,
47:09I I don't think we're ever going to see
47:10big job losses. I think the demographics
47:12will kick in, we'll see more people
47:13retire, the stock market as it goes
47:15higher, more people say, "I don't need
47:16this anymore. I've made enough money."
47:18So, here is the last 6 months. Just look
47:21at these 6 months, guys,
47:23and take it back over time. This is a
47:25very weak period, and if you want to see
47:27really how weak it is, and again,
47:29this is the weakest 6-month rate of
47:32change
47:33on the aggregate weekly payrolls
47:37going all the way back to 2012
47:40when you throw out the COVID time
47:41period. I don't care how you measure
47:43this. Maybe it'll change, but right now
47:45the data is a fact.
47:48The number of people looking for work
47:49and the number of people counted as
47:51unemployed both fell.
47:54We're losing people in the labor force.
47:56This is 10-year yields, the 20-month
47:58moving average,
48:01which is at 4.29,
48:03and we're currently up here.
48:05Here's the labor force participation
48:07rate.
48:09As more people leave the labor force,
48:11rates should be coming down.
48:15That's it. So, we'll see what happens,
48:17but again, we've got the rise of the
48:19million-dollar companies happening now,
48:21so I think the
48:23future of the job situation will remain
48:26weaker because AI agents are coming.
48:28It's not that they're going to collapse,
48:30but I think people are making a huge
48:32mistake not understanding that these
48:34little tiny companies that have no
48:35employees will be replacing bigger
48:37companies
48:38with maybe 50 employees, 100 employees.
48:41And when those when that company closes,
48:43those people have to recycle in. I just
48:45think the quits rate, all of that stuff
48:47is showing what's happening much better
48:49than what your economists are telling
48:51you. And so, at some point here, I do
48:54think we're going to see weakness either
48:55A in inflation or B inside the job
48:59market driven by AI, which surprises
49:01people.
49:05So, here we are. Bitcoin is still in a
49:06bear market until we get above there.
49:08Anything from my perspective is a
49:10trading side. So, I'm going to I I think
49:14this is going to hold, and I think we're
49:16going to break out. But as of right now,
49:17we haven't. But there is some good news.
49:20I want you to go read this because I do
49:22believe that this is the most important
49:24time because of the agent to kick in.
49:26But what I really like is as the agent
49:28and my crypto 40 name index,
49:31the ecosystem is doing well.
49:34We just saw the Clarity Act basically
49:37over the course of the last 2 weeks go
49:39from about 33 to 40 uh let's say 3 weeks
49:44ago, we were getting close to 50. We
49:46collapsed to 17%
49:50chance that it gets signed into law this
49:52year.
49:53We also had strategy Michael Saylor sell
49:55more.
49:57And we also had
50:00cold card exploit. And if you guys
50:02didn't see this, there was even a, you
50:05know, a pretty big story on cold storage
50:08problem with one company
50:11uh and people's wallets. And yet, all
50:13that going on mm
50:15Bitcoin was okay this week.
50:18That's it for me this week, guys. Uh
50:20keep enjoying your summer. Reach out to
50:22Mark Whaling. Reach out to me.
50:25Uh and we'll keep going through it. But
50:27uh let's make some money
50:28uh starting now for the rest of the
50:30year. See you.

About this transcript

This page contains the full transcript of The AI Crash Is Over. Here's What Comes Next by Jordi Visser, generated from the public captions YouTube serves with the video. The transcript has 8,861 words across 1,393 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.