Latest / In the Loop AI / Sam Altman Issues a "Code Red." The Tides Are Shifting In AI.

Transcript

Sam Altman Issues a "Code Red." The Tides Are Shifting In AI. · In the Loop AI

0:00
Josh: The golden child of AI, OpenAI, has issued a code red.
0:05
Josh: Just this Monday, Sam Altman, he sent out an internal memo declaring code red around ChatGPT.
0:10
Josh: What is code red? We're gonna get into it, but we understand that it's a direct
0:14
Josh: response to the increase in competition from people like Google and Anthropic
0:19
Josh: who are now releasing models that are superior to anything that ChatGPT has ever created.
0:24
Josh: Code red seems a little risky, a little dangerous for a company that is so levered
0:29
Josh: up on future costs and not able to make a lot of money so
0:33
Josh: i'm not really sure what's going on here ejaz you've lived through a code red
0:37
Josh: in the past through previous experiences can you walk us through what it looks
0:40
Josh: like when a company issues a code red because from what i understand the entire
0:44
Josh: company just gets put on pause and everyone focuses on a single
0:47
Josh: existential risk to the company and i assume that's around chat gpt right
0:52
Ejaaz: Yeah, Code Red can be described with one word, Josh, danger.
0:57
Ejaaz: So I experienced a Code Red when I worked at Coinbase, and it doesn't come about
1:02
Ejaaz: too often. And it basically means one simple thing.
1:06
Ejaaz: The core product or the thing that makes your company special and all the money
1:10
Ejaaz: that it makes is in danger.
1:12
Ejaaz: It is about to die and it is under the threat of major competitors.
1:15
Ejaaz: And you need to pile any and every resource that you can into protecting it
1:20
Ejaaz: and making it the number one leader again.
1:24
Ejaaz: And so my experience at Coinbase, when a code red was issued was literally I
1:28
Ejaaz: woke up one morning, Josh, there was a slack message across the entire company
1:32
Ejaaz: saying code red, here's a list of things that we need to do.
1:35
Ejaaz: And I lost half my team to a completely different part of the organization to
1:40
Ejaaz: focus on something else.
1:42
Ejaaz: And that pursued for like the next like quarter or so, which is just kind of crazy.
1:46
Ejaaz: And so in the context of Sam Altman, and with OpenAI, I just want to give a
1:50
Ejaaz: bit of a context as to like how this has come to be.
1:53
Ejaaz: Last week, I'm showing you this article here.
1:56
Ejaaz: It was leaked in an internal memo by Sam Altman that he gathered the company around and said, hey,
2:01
Ejaaz: in the aftermath of Google releasing their new Gemini 3 Pro model,
2:06
Ejaaz: which absolutely blasted them across every single benchmark by a wide, wide margin,
2:11
Ejaaz: we're probably going to face some economic headwinds because what he was basically
2:15
Ejaaz: saying is our models aren't good enough and we have to put in some more effort to catching up.
2:20
Ejaaz: And that kind of, that economic headwinds kind of like transgressed into a hurricane
2:25
Ejaaz: of sorts, if we wanted to keep the analogy going along,
2:28
Ejaaz: into a code red where he's issued an internal memo to his company that we need
2:34
Ejaaz: to gather all resources to make ChatGPT better.
2:36
Ejaaz: And there's three main buckets that they're compiling these resources into,
2:39
Ejaaz: Josh, which I just want to briefly go over. Number one,
2:43
Ejaaz: improve model performance. So that means improving the benchmarks to match and
2:47
Ejaaz: beat Gemini 3 Pro and other competitor models, and also to help on the pre-training
2:52
Ejaaz: and post-training front. So that's where the majority of the resources are going.
2:55
Ejaaz: The second bucket is going to be in image generation.
2:58
Ejaaz: Now, I found this really interesting because this is a direct response to Gemini
3:04
Ejaaz: 3's Nano Banana release. We've spoken about this on a previous episode.
3:08
Ejaaz: You should definitely go check that out. I think it was from two weeks ago.
3:11
Ejaaz: But this is their text-to-image editing model, which is actually really good
3:15
Ejaaz: and driving the most usage for Gemini 3 right now. Their downloads are off the
3:20
Ejaaz: charts right now because of this feature.
3:22
Ejaaz: Sam wants to combat that directly. And the third bucket is personalization, Josh.
3:27
Ejaaz: So you and I have said this before, but OpenAI or ChatGPT's moat is personalization.
3:33
Ejaaz: It's memory. It's the reason why you and I consistently use ChatGPT,
3:36
Ejaaz: despite there being a lot of these other models. they understand and they get us.
3:39
Ejaaz: He wants to double down on that. So you might then ask, well,
3:43
Ejaaz: what comes at a cost to all of this? Like what's the price that he's paying?
3:47
Ejaaz: Well, all the efforts and resources that he's been putting into advertising,
3:51
Ejaaz: into consumer products, in agents, into healthcare, all the stuff that he spoke
3:55
Ejaaz: about on his recent developer conference that everyone got super excited about is now void.
4:00
Ejaaz: Sora 2, it's kind of like it's chilling. It's all been put on hold to focus
4:04
Ejaaz: on these three core buckets. He has gone into war mode, Josh.
4:09
Josh: Yeah, it's, um, I could understand why. I mean, OpenAI, they're reportedly valued
4:14
Josh: at what, half a trillion dollars, 500 billion.
4:17
Josh: They have greater than 800 million weekly active users, which has surprisingly
4:21
Josh: and suspiciously stayed the same over the last two months. Every time they mention
4:24
Josh: it, it's been pegged at 800 million per week. So I'm a little curious if that
4:29
Josh: means that they've kind of stalled out on growth.
4:31
Josh: They are still unprofitable, and they have greater than $1 trillion in cloud
4:35
Josh: and chip obligations to companies and partners like Microsoft,
4:39
Josh: Oracle, and NVIDIA. So they have this huge...
4:43
Josh: Like conundrum on their hands where they need to start making money.
4:46
Josh: And in order to do so, they need to have products that are superior to everybody else's.
4:50
Josh: And we have on screen here a few charts that I found fascinating that may have
4:53
Josh: spurred the code red, which is the relationship between Gemini downloads and ChatGPT downloads.
5:01
Josh: EJ, as we mentioned just on our episode earlier this week, how like Gemini is
5:04
Josh: now on the home screen of my phone.
5:06
Josh: And this is a new phenomenon because of how good it is. We talk about Nano Banana
5:09
Josh: all the time. We had an episode just about it because it was so good.
5:12
Josh: And what you're seeing here in the charts is that Gemini is starting to catch up.
5:15
Josh: And I suspect that it's okay if Gemini is a slightly superior model to ChatGPT's flagship model.
5:22
Josh: But if the consumer application, if that product is at threat,
5:27
Josh: which is the ChatGPT mobile app, the ChatGPT desktop app, if that's under threat
5:31
Josh: by a Gemini, then that's a serious problem.
5:33
Josh: And that's what we're seeing reflected in these charts is Gemini as a better
5:37
Josh: model, sure, but it's also now starting to impede on the actual user base of ChatGPT.
5:42
Josh: And when your company is contingent on this exponential line of growth and you
5:46
Josh: have a real threat to that growth, you must do everything in your power to fix that.
5:51
Josh: And that is what the Code Red is all about. It's like, okay,
5:53
Josh: how do we make this product better so we stop losing market share to these other competitors?
5:58
Ejaaz: Yeah, and it's not like this threat isn't very, very real.
6:01
Ejaaz: I mean, look at this chart, right? You've got not only Gemini users and downloads
6:06
Ejaaz: creeping up to match ChatGPT users, they're spending more time on the app,
6:11
Ejaaz: Josh, which has always been ChatGPT's moat.
6:14
Ejaaz: So, you know, we might be fear-mongering a bit here, but it is a very real threat.
6:19
Ejaaz: We've heard feedback from other people, such as this tweet that I have pulled
6:22
Ejaaz: up here, where, you know, this is just a fad.
6:25
Ejaaz: People are only using Gemini because of Nano Banana Josh. Josh,
6:28
Ejaaz: that's the main reason why you have it on your home screen right now.
6:30
Ejaaz: So there is an argument that this might die down in a few weeks.
6:34
Ejaaz: Time will tell, but it's obviously enough to kind of like push Sam into this
6:38
Ejaaz: hole where he needs to kind of like combat this, right?
6:41
Ejaaz: So then the next natural question I would ask is, well, what's he doing about
6:45
Ejaaz: this and how fast is he going to move?
6:47
Ejaaz: There's two main updates that you need to know about.
6:50
Ejaaz: Next week, he's rumored to release a new reasoning model, which is meant to
6:54
Ejaaz: match and even surpass Gemini 3.
6:57
Ejaaz: It's kind of crazy how he's just done that turnaround. We went from like no
7:01
Ejaaz: ChatGPT model for the rest of the year to, yeah, we're going to roll out one
7:05
Ejaaz: next week. So we already see the effects of this code red coming into place.
7:09
Ejaaz: But then, Josh, there is another secret model.
7:13
Ejaaz: Which is planned to be released early next year in January, called Codenamed
7:18
Ejaaz: Garlic, which is meant to be kind of like a GPT 5.5, which represents or implies
7:25
Ejaaz: a major jump in performance,
7:27
Ejaaz: and for them to retake the kind of number one spot convincingly.
7:30
Ejaaz: This is all hearsay right now. I have no confidence that this is going to happen,
7:35
Ejaaz: but these are the kind of like rumors being spread. And Google isn't the only
7:38
Ejaaz: threat that they're facing that is kind of instigating them with this code red.
7:42
Ejaaz: Anthropic has come out of the woodworks with a flashy new coding model,
7:47
Ejaaz: Claude Opus 4.5, which not only maintained their lead as the number one.
7:53
Ejaaz: Coding AI model, but kind of like boosted them even further than they already
7:57
Ejaaz: were. Like they've got like a marginal lead right now.
8:00
Ejaaz: I'm showing up the quick stats on Opus 4.5 right here.
8:04
Ejaaz: You've got stuff as 80% on agentic coding, terminal coding, 60%, agent tool use, 90%.
8:12
Ejaaz: If you compare these to any of their previous models, as well as Gemini 3 Pro,
8:17
Ejaaz: they're absolutely smashing it, right?
8:19
Ejaaz: And so the fact is, despite Google's kind of like advantage with TPUs or whatever
8:24
Ejaaz: stuff, we've spoken about this on a dedicated episode actually earlier this
8:27
Ejaaz: week or last week, definitely go check that out.
8:29
Ejaaz: They've been able to find a secret source to kind of like boosting their models,
8:33
Ejaaz: Josh. And this is one of many good things about this new model release and about Anthropic this week.
8:39
Ejaaz: Josh, do you have any thoughts on this model? The reason why I'm impressed by
8:43
Ejaaz: this is I think coding is such an important kind of characteristic to have or
8:50
Ejaaz: advantage to have for a model.
8:52
Ejaaz: Why? Because if you have the best coding model, you can practically create whatever
8:56
Ejaaz: the hell you want. If there's a competitor that creates a cool new feature,
8:59
Ejaaz: well, hey, Anthropic can just spin up an exact replica and hearsay. Do you feel the same?
9:04
Josh: Yeah. So here's the getting back to the ChatGPT point. Gemini is better at math and science.
9:09
Josh: Anthropic is better at code. That leaves a narrower and narrower sliver for
9:14
Josh: ChatGPT to actually be great at.
9:16
Josh: And this Anthropic model, it's amazing. I think one of the most important things
9:19
Josh: that I want to highlight here is the scores are great but the cost
9:23
Josh: per token is the most impressive because not only is it a huge intelligence jump
9:29
Josh: it is significantly cheaper per token yeah you could see this chart here how
9:32
Josh: much cheaper it is relative to previous models and how we're kind of getting
9:35
Josh: this this exponential increase where it seemed to previously be flattening flattening
9:39
Josh: out we're getting a new increase and
9:42
Josh: anthropic kind of exists in this strange
9:44
Josh: bubble of code but my god is it really good at code and it comes equipped with
9:49
Josh: this new agentic feature where it will actually ask you a series of questions
9:52
Josh: after you submit a prompt that will kind of go through all of the edge cases that you might want
9:57
Josh: and then it will one-shot things that are really impressive we have an example
10:01
Josh: of a designing a front end right
10:03
Josh: yeah this is crazy yeah i mean please look at this
10:06
Ejaaz: Yeah i mean literally like look okay like i don't need to demonstrate the entire
10:10
Ejaaz: video but just look at the prompt on the left here that i'm circling with my
10:13
Ejaaz: cursor and then just look at the result.
10:16
Ejaaz: This happens, by the way, over a matter of like a few minutes,
10:19
Ejaaz: you have like a fancy fintech background,
10:22
Ejaaz: fintech front end, which is what the prompt asked it to build,
10:25
Ejaaz: in a matter of minutes. This would normally take like a Fortune 500 company
10:30
Ejaaz: a couple months to produce about two years ago. And now it's happening in under
10:35
Ejaaz: four minutes, which is just insane to see.
10:37
Ejaaz: But Josh, there's also this really cool blog post I saw about the Red Team project.
10:42
Ejaaz: Can you tell me about this?
10:44
Josh: I love that. This week we have OpenAI with CodeRed and Anthropic with Red Team.
10:48
Josh: And it kind of shows you it's a testament to where both companies are at right
10:51
Josh: now one is very clearly on the defense one seems to be on the offense so part
10:55
Josh: of this new anthropic rollout is they published this blog post
10:58
Josh: about whether ais can actually exploit blockchains and smart contracts so for
11:02
Josh: the crypto fans this is for you um in simulated testing they found that
11:07
Josh: It could gain access to 4.6 million dollars in exploits and again i mean anthropic
11:12
Josh: is the best coding model in the world if you ever wanted to penetration test
11:16
Josh: any software, this would be the model to do it with.
11:19
Josh: And with every new incremental increase and improvement in this model,
11:22
Josh: it unlocks new exploits that we can actually find. So Nick Carter has this amazing
11:27
Josh: take that I love. I'm just going to read it word for word because it's exactly how I was feeling.
11:30
Josh: And he says, within a year, when people tell their AI agent to go make money,
11:34
Josh: the AI will interpret that as go and steal crypto by phishing our exploits.
11:38
Josh: We're going to have to take OPSEC and smart contract security a lot more seriously
11:42
Josh: on a go forward basis and that's that's true it's like you get into the alignment question where
11:48
Josh: like is this ai good or is this ai trying to answer
11:51
Josh: you solve your request and if your request is go make me some money
11:55
Josh: well it'll just go hack from someone else's because a lot of these systems are
11:59
Josh: not secured against the most cutting edge ai just yet so it unlocks this new
12:04
Josh: interesting paradigm that
12:06
Josh: i find is unique uniquely for anthropic in the case because they are just the best at coding
12:11
Josh: and code is what secures all of our financial assets today so interesting thought experiment
12:15
Ejaaz: Yeah. And the fact of the matter is like, you know, this, the coding frontier
12:20
Ejaaz: model is an incredibly valuable thing to have.
12:22
Ejaaz: And the numbers don't lie, Josh. So if you remember, in January of this year,
12:28
Ejaaz: Dario Amode, the CEO and founder, projected that they would probably reach around $1 billion worth of AR.
12:35
Ejaaz: That's currently what that was what the revenue they were earning at the time.
12:40
Ejaaz: He went on stage yesterday and announced convincingly that they are on track
12:44
Ejaaz: for $10 billion of annual recurring revenue at the end of this year.
12:49
Ejaaz: That is a 10x. That is insane.
12:52
Ejaaz: And so it's going really well. He's on track to hit that $70 billion projected goal,
12:59
Ejaaz: ARR in 2028, which we spoke about a few episodes ago, which will make them the
13:04
Ejaaz: first profitable major AI lab in 2028, beating OpenAI, who's projected to get there in 2031, 2032.
13:12
Ejaaz: So all things are in favor of Anthropic to kind of like do something pretty
13:16
Ejaaz: dramatic, such as a IPO, maybe.
13:20
Ejaaz: Rumors broke, maybe, well, maybe convincingly from an insider.
13:24
Ejaaz: News broke that Anthropic has engaged a firm to start prepping them for an IPO
13:30
Ejaaz: in 2026. Of course, there hasn't been any official confirmations yet,
13:34
Ejaaz: but the regulations and the docs do not lie. um.
13:38
Ejaaz: It also says that Dario is in the kind of midst of closing a massive round,
13:44
Ejaaz: which would value Anthropic at $300 to $350 billion, which will place them up
13:50
Ejaaz: there with OpenAI, who's currently valued at $500 billion.
13:53
Ejaaz: And with OpenAI going into Code Red, with Anthropic entering Red Team and this
13:57
Ejaaz: new Frontier model, maybe they might be even worth the same amount, if not more.
14:02
Ejaaz: So that's just a very interesting kind of like run for Anthropic in general.
14:07
Ejaaz: I saw this tweet, I want to mention it very quickly, that states that Anthropic,
14:11
Ejaaz: might even value the IPO at 300 to 350, which would mean that they don't make
14:16
Ejaaz: any money on it. And the argument for that would be they've already raised so
14:20
Ejaaz: much money that they don't need it.
14:21
Ejaaz: And it is more beneficial for investors to purchase or buy Anthropic stock in the future.
14:27
Ejaaz: I don't know this is all hearsay, but it's something interesting to kind of check out.
14:31
Ejaaz: And the final thing I want to make a point of is these IPO valuations,
14:35
Ejaaz: Josh, are getting absolutely insane.
14:38
Ejaaz: I have this really fun anecdote where a Chinese AI company, which is meant to
14:43
Ejaaz: try to compete with NVIDIA, IPO'd on the Chinese Stock Exchange this week.
14:48
Ejaaz: And Josh, it IPO'd for 4,000 times oversubscribed, which placed it at a 4.5 billion.
14:57
Ejaaz: A trillion dollar market cap beating nvidia's market cap now that was for a
15:04
Ejaaz: brief moment it is now regulated down but it just shows the insatiable appetite
15:08
Ejaaz: and demand for ai stuff right now.
15:11
Josh: Yeah the the four and a half trillion dollar number it's outrageous it's not
15:14
Josh: the only company with the valuation close to that high that we're going to talk
15:16
Josh: about this week we have updates from amazon amazon's been moving kind of slow
15:20
Josh: out of all the mag 7 companies they are the
15:23
Josh: the one showing the least amount of growth this year so it's it's fitting that
15:27
Josh: it's their time is coming if not already here ijaz and you have some updates
15:31
Josh: for us about what they actually announced this week at their aws event so can
15:34
Josh: you walk us through where does amazon stand why are they losing and why is that
15:38
Josh: hopefully not indicative of the future growth that they're going to see from this ai
15:43
Ejaaz: Advancements you can see i'm rubbing my hands josh for those of you who are
15:47
Ejaaz: listening i'm rubbing my hands uh the reason why is uh i've been purchasing
15:50
Ejaaz: amazon stock recently um right before.
15:53
Josh: We recorded actually you guys were saying i'm dcaing in right now
15:56
Ejaaz: Oh yeah oh yeah um and it's because i have an amazon thesis which is it is a sleeper ai stock hit um
16:03
Ejaaz: and this uh set of releases the slew of ai releases that aws um launched this week is proof of that
16:11
Ejaaz: um so i'm going to quickly run you through it um they released a new uh ai chip
16:15
Ejaaz: or ai gpu which competes directly with nvidia's gpus and Google's TPUs.
16:21
Ejaaz: And it's cool for a few different reasons. Number one, it's four and a half
16:25
Ejaaz: times more performant than the previous Tranium chip that they launched.
16:28
Ejaaz: This is Tranium 3. I'm talking about Tranium 2.
16:31
Ejaaz: It is 4x more cost efficient. So it's cheaper.
16:36
Ejaaz: And it also has higher memory bandwidth, which makes it just all around like
16:39
Ejaaz: a really good infrastructure product to use for people to train their models.
16:44
Ejaaz: So that's super cool. And I think it kind of competes directly with NVIDIA in
16:49
Ejaaz: the sense that it can get you about 80% as good as NVIDIA GPUs, but 50% cheaper.
16:56
Ejaaz: So for anyone that's kind of like running on NVIDIA GPUs and they're like spending
16:59
Ejaaz: billions of dollars, they could slash that bill by half, which is really cool.
17:04
Ejaaz: The other really cool thing that they released is something called Nova Forge.
17:08
Ejaaz: In a single sentence, it allows any enterprise to train their own AI model using their own company data.
17:15
Ejaaz: The reason why that's cool is, imagine, like ChatGPT is good,
17:19
Ejaaz: right? But imagine if you had ChatGPT for your own company, imagine how productive
17:23
Ejaaz: you'd be. Imagine how productive each of your employees would be.
17:26
Ejaaz: That's a really unique thing. And it just tells me that like Amazon is going
17:30
Ejaaz: down the enterprise route, similar to the company that they have a 19% stake
17:35
Ejaaz: in, Anthropic, right? So I kind of see the alignment here.
17:38
Ejaaz: And the third point is around these three agents that they released,
17:43
Ejaaz: which basically helps them automate a bunch of things using AI.
17:47
Ejaaz: The way I think about it is Amazon's kind of like the automation AI giant.
17:52
Ejaaz: And I think they're going to do really well, Josh.
17:54
Josh: Yeah, Amazon. I mean, we could lay out a bit of a brief bull case here because
17:58
Josh: I'm super excited about Amazon, too.
18:00
Josh: I was looking at the Mag7 leaders, like who in the Mag7 moves the most amount of atoms?
18:08
Josh: Because when you apply this AI leverage to these companies, the ones that move
18:12
Josh: the most amount of atoms stand to benefit the most.
18:15
Josh: And we have Tesla, we have Apple, we have Microsoft, we have Google.
18:18
Josh: But Amazon moves the most amount of physical stuff in the world by a large margin.
18:24
Josh: I think they're estimated, what is it, 6 to 7 billion orders per year,
18:28
Josh: containing 12 to 15 billion items shipped in 8 to 10 billion packages globally.
18:34
Josh: So in the US alone, that's like 6.3 billion partials in 2024.
18:39
Josh: They're on track for 8.4 billion in 2028. It's a huge amount of stuff just moving around.
18:44
Josh: And when you apply automation to that stuff, you can cut your cost and increase
18:50
Josh: your margin at such a significant rate that amazon stands to benefit hugely from this automation
18:57
Josh: and we're seeing this with the things that they announced this week with in
19:00
Josh: terms of like building their own
19:01
Josh: vertically integrated ai systems we have some interesting takes on chips that
19:05
Josh: they're making so to me amazon is just like poised to explode just because the
19:10
Josh: sheer amount of volume that they move throughout the physical world
19:14
Ejaaz: I i that's such a good point josh did you see the news uh that got released
19:17
Ejaaz: this morning about the USPS stuff?
19:20
Josh: No, what was that?
19:22
Ejaaz: Okay, so get this. Amazon currently pays USPS six to $7 billion a year for them
19:30
Ejaaz: to ship packages for them, right?
19:32
Ejaaz: Because they don't want to like, they can't claim everything because like,
19:36
Ejaaz: you know, you can use the existing postal network.
19:39
Ejaaz: They announced plans, they being Amazon, that they're going to create their
19:42
Ejaaz: own postal service equivalent this morning, which will basically wipe out $6 to $7 billion.
19:49
Ejaaz: No other company can do this to the scale that Amazon is doing it. It's just insane.
19:54
Ejaaz: The other thing I wanted to address, because I know you listeners,
19:57
Ejaaz: I've read all your comments. I know exactly the criticism you're about to give
20:00
Ejaaz: to this segment is, oh, but NVIDIA is still good because everyone is using their
20:05
Ejaaz: software CUDA, C-U-D-A.
20:08
Ejaaz: Amazon has a special source in response to that, which is with their new Tranium
20:12
Ejaaz: chip, they have software that you can just simply port from NVIDIA's to Amazon's,
20:18
Ejaaz: which makes the switching cost not just a hardware consideration,
20:21
Ejaaz: but an easy software consideration as well.
20:23
Ejaaz: Now it becomes really easy if you're running a NVIDIA GPU cluster on AWS to
20:28
Ejaaz: hop onto Amazon. So it's actually a really smart strategy. And I'm pretty,
20:32
Ejaaz: I'm bullish, Josh. Can you tell? Yeah, I can.
20:36
Josh: The switching cost being as close to zero as possible is huge,
20:39
Josh: right? Because like as much,
20:41
Josh: I'd say what CUDA is to NVIDIA memory is to OpenAI, where it is a very strong
20:46
Josh: moat. And if you can, it's basically saying we can make your memory portable in a commercial way.
20:51
Josh: And that's a huge unlock for anyone who is interested in using these AWS chips.
20:55
Josh: Now, when they make it, I don't know.
20:57
Josh: There are some problems in the world of computers, particularly around RAM right
21:01
Josh: now. If you have ever built a computer before, you know about RAM.
21:04
Josh: RAM is this, it's the very lightning fast memory that holds temporary memory, but it's very fast.
21:10
Josh: And it's required for training AI data centers, so much so that there has become
21:15
Josh: this egregious shortage of RAM.
21:18
Josh: And for people who haven't lived through the previous GPU cycle,
21:22
Josh: there was this huge problem where you couldn't buy an NVIDIA GPU because they
21:25
Josh: were being used for crypto miners.
21:26
Josh: And if you're a hobbyist at home, now you can't buy RAM for your computer because,
21:31
Josh: well, the prices have tripled in the last 12 months.
21:34
Josh: And this comes on the back of an announcement from a company that makes the crucial RAM.
21:39
Josh: They just stopped making it for commercial purposes. They're using it only for
21:42
Josh: institutional entities, for private entities who are building AI.
21:47
Josh: So now one of the three major manufacturers of this RAM is stepping away.
21:51
Josh: If you are building a computer, I am sorry, it is about to get a lot more expensive.
21:55
Josh: If you are building an AI data center,
21:57
Josh: It's also expensive, but now you at least have one of these three major suppliers
22:00
Josh: explicitly making stuff just for you.
22:03
Josh: So just something to note that there are these supply chain crises happening.
22:08
Josh: This most recent one is in RAM. Previously, it was GPUs. I guess it's still GPUs.
22:12
Josh: But as we go along, it's going to be interesting to see where these kind of bottlenecks form.
22:16
Josh: And just in terms of, you know, bubble watch, we just want to see where everything's landing.
22:20
Josh: This is the RAM segment. Here's where it is. It was getting very high very quickly.
22:24
Josh: One of the companies bowed out. And that's just kind of like the nerdy hobbyist enthusiast,
22:28
Ejaaz: Like state of the union for it. Sorry, Josh. Sorry. I can't let you get off
22:32
Ejaaz: the hook that easy. You sent me this website and it looks intense.
22:38
Ejaaz: What am I looking at?
22:40
Josh: I can't believe you've never used PC Part Picker. If anyone has ever built a custom PC before.
22:45
Josh: There's a lot of compatibility issues between parts because sometimes a specific
22:50
Josh: manufacturer's piece does not work with another manufacturer's piece.
22:53
Josh: So what PC Part Picker does is it allows you to place all of these pieces into
22:56
Josh: one place and it'll tell you which is compatible with which.
22:59
Josh: It also does a great job of tracking prices of these items. And what we're seeing
23:03
Josh: here in these charts that are all going up and to the right is the price for
23:07
Josh: the average person to get their hands on a critical piece of compute hardware, which is the RAM.
23:12
Josh: And there's another nerdy joke she made here where now developers are going
23:15
Josh: to have to actually optimize their websites because a lot of the times these
23:18
Josh: big clunky websites and the big clunky code, it all gets stored in your RAM
23:21
Josh: because they just throw it in there's no budget.
23:24
Josh: Well, maybe if there's a budget, they'll start making more efficient code.
23:26
Josh: But that's kind of a deep cut.
23:27
Josh: I don't want to spend too much more time on here. We have other interesting
23:29
Josh: things to talk about, mostly in the world of Apple.
23:32
Josh: There is a big switch up this week in Apple world. Please, Ejaz, tell us what happened.
23:36
Ejaaz: Well, I want you to tell me about this, Josh. This is your favorite company
23:40
Ejaaz: ever, Apple, and they just lost their chief of AI.
23:45
Ejaaz: John is leaving and he's being replaced by this guy called Amar from Microsoft,
23:51
Ejaaz: and is joining to lead AI under Craig Federini.
23:53
Ejaaz: Can you give me the law of this?
23:56
Ejaaz: Is this kind of bearish,
23:58
Ejaaz: eventually bullish or should I just be bearish?
24:01
Josh: I don't really, I don't know these guys. It's like, I mean, Apple AI sucks.
24:07
Josh: So whoever's in charge of it, like, sure, get rid of them. Whoever tried,
24:10
Josh: they failed clearly because like, I still don't have Siri enabled and they're
24:14
Josh: now hopefully going to outsource to Gemini.
24:16
Josh: I think this is probably a good thing. If you're in charge of AI at Apple,
24:19
Josh: you probably should be fired.
24:20
Josh: I find it funny how everyone kind of reports to Craig Frederici.
24:24
Josh: So everyone comes in and out, but Craig is always that guy. He's always sitting
24:27
Josh: there he never takes the blame he never takes the fall um i'd be curious to know
24:31
Josh: how much of this is his responsibility versus people who are beneath him um
24:37
Josh: but yeah it's funny actually the first comment says under craig federighi that's
24:40
Josh: the problem right there um so
24:43
Josh: I don't know where to stand on this i know that apple's ai sucks i hope this
24:47
Josh: new guy i mean god speed brother amar whatever you're like whatever you're doing
24:51
Josh: i wish you all the best i hope you figure it out um that's kind of what i got for this
24:56
Josh: there is another guy as well i was gonna say this is the more interesting conversation
25:01
Josh: to me at least i think uh which is this guy who was previously in charge of
25:06
Josh: a lot of things at apple his name is alan dye and he's
25:10
Josh: responsible for this whole slew of things he was working on packaging then he
25:13
Josh: helped with the ios 7 redesign he created the dynamic island he led the ui for
25:17
Josh: vision pro he helped with the vision pro launch he did liquid Glass, he did all of
25:22
Josh: the human interface spatial computing things.
25:25
Josh: And he just, one of the things that we love and one of the things that I'm personally
25:28
Josh: excited about in the future is to switch to spatial computing.
25:31
Josh: And this guy did a great job. Like Vision OS looks great. Liquid Glass,
25:35
Josh: we've had our debates on the show. I think it looks pretty good.
25:38
Josh: This is the guy kind of behind and in charge of all of that stuff.
25:41
Josh: And Meta poached him for a presumably ton of money. Meta said,
25:45
Josh: hey, we'll pay you whatever you want. Come over here, do it for our products.
25:49
Josh: That's presumably going to be for the meta ray band displays that we had a few
25:54
Josh: episodes on earlier on and
25:56
Josh: it made me a little bummed because this guy actually did do a great job on these
26:00
Josh: apple products and now he's going to be going to develop for meta who makes
26:04
Josh: really crappy hardware products in really siloed software suites um so
26:08
Ejaaz: To have a little bit of sweet zuck is really on a on a rampage isn't he to just
26:13
Ejaaz: kind of like hire the talent from everywhere but like will he actually prove
26:17
Ejaaz: it like listen he spent what 25 billion dollars in hiring like a hundred people,
26:23
Ejaaz: less 100 people actually at 30 35,
26:27
Ejaaz: to 40 people and like i haven't seen the fruits of that labor so like is he
26:29
Ejaaz: going to do the same thing with design and his glasses probably not it was an
26:32
Ejaaz: absolute shambles of a launch so we'll see uh josh do you know anything about
26:36
Ejaaz: the guy that's replacing this apple design guy steve lemay.
26:40
Josh: New guy uh no other than the fact that he's been around for a very long time
26:44
Josh: and apparently has been here since 1999 he helped ship the original iphone uh
26:49
Josh: mac os ios iPadOS, watchOS, and most recently, visionOS.
26:53
Josh: So somebody who was around during the Steve Jobs, Johnny Ive era seems bullish.
26:59
Josh: That's about all of the context clues that I have. But hey, again, I wish you the best.
27:05
Josh: I hope you could help Apple figure it out. They have a lot of figuring out to do.
27:09
Josh: And that's just, we're going to take it from there. But there is another interesting
27:12
Josh: topic that I want to cover here, some science corner stuff, which is pretty cool.
27:16
Josh: We found sugars in outer space, essential to life.
27:21
Josh: This is kind of cool. It was found on some asteroid that was in deep space.
27:27
Josh: And it is one of the first times that we actually discovered some sort of building
27:32
Josh: blocks for life that came from outside of our galaxy.
27:36
Josh: So I wanted to just briefly mention this because it creates this interesting
27:40
Josh: conversation around the Fermi Paradox.
27:42
Josh: For people who don't know the Fermi Paradox, it's the apparent contradiction between
27:47
Josh: The high probability that intelligent life should exist elsewhere in the universe
27:51
Josh: and the total lack of evidence that it actually does.
27:55
Josh: So from a Fermi paradox perspective, this is great and terrible news.
27:59
Josh: Terrible as the paradox is stronger than ever, but great because whatever filter
28:03
Josh: causes the paradox is more likely to be early in our evolutionary history rather than in the future.
28:09
Josh: You can at least make the case that this life was kind of randomly placed
28:15
Josh: and there is a high probability that it can be placed randomly elsewhere because
28:20
Josh: there are the basic building blocks coming from all these external places. What do you think?
28:27
Ejaaz: Listen, my pastime in a very, very long ago life was a biologist.
28:34
Ejaaz: That's what I kind of like studied at university.
28:37
Ejaaz: And all I could help but think when I learned about all these different amino
28:40
Ejaaz: acids, building blocks, DNA, blah, blah, is like,
28:43
Ejaaz: there is so much chance in this that it is almost certain that there exists
28:47
Ejaaz: another universe where there are other sorts of life forms because all the building
28:50
Ejaaz: blocks are just out there, right? So why should we be the only favored ones?
28:55
Ejaaz: Josh, I pulled up the sugar futures chart and I have to say,
29:03
Ejaaz: I have to say, no one seems to be reacting as violently as I had expected here.
29:08
Ejaaz: Listen, I'm a markets man. I'm invested in this stuff.
29:11
Ejaaz: I want to see how I can make returns on this. And I don't know,
29:15
Ejaaz: maybe there's a delay in this information, Josh.
29:17
Ejaaz: Maybe I should be shorting the sugar stocks as much as I can before they kind of catch on.
29:23
Josh: That's so funny. You're the type of guy where like a meteor will hit earth and
29:27
Josh: it will be made of solid gold.
29:29
Ejaaz: And the first.
29:29
Josh: Thing you will do is hop on your computer and figure out how do I short gold?
29:34
Josh: Forget about my bug out bag. Forget about how do I short gold?
29:37
Josh: We just got hit by tons of it on earth.
29:39
Josh: That is so funny. Well, that's, it's just an interesting thing.
29:41
Josh: It's like, it makes you think it's like, okay, we found a critical building
29:44
Josh: block for life that came from another universe, yada, yada, yada,
29:47
Josh: whatever that science corner.
29:48
Josh: We have something important to celebrate as we wrap up this episode which is
29:52
Josh: our spotify wrapped which just came out and we have some really amazing stats
29:57
Josh: and some really gracious thank yous to hand out to everybody because
30:01
Josh: wow i think uh i share this thought in the fact that we are both blown away
30:05
Josh: by the support that we've had over the past year and these statistics that we're
30:08
Josh: showing on screen are like pretty outrageous you guys right like top one percent across the board
30:13
Ejaaz: Top 1% across every single category. So that is a 2025 instant hit show,
30:18
Ejaaz: 2025 most shared show. That's you guys listening to this right now,
30:22
Ejaaz: watching this right now.
30:23
Ejaaz: Thank you for sharing it with all your friends. It turns out you guys are listening.
30:26
Ejaaz: Someone's listening to us, Josh, which is great. I wonder if it's the aliens.
30:31
Ejaaz: Exactly. I also do want to point out that our show has only been going for six months.
30:38
Ejaaz: So we got like half a year delayed on the headstart for this.
30:42
Ejaaz: And we still managed to pull the top 1%.
30:45
Ejaaz: No easy feat, Josh. But the one I'm probably most proud of right now is we are
30:49
Ejaaz: ranked as one of the top technology podcasts in the world, ranking at 21, Josh.
30:55
Ejaaz: I can, I can't, I'm not looking at your face right now because I'm staring at
30:58
Ejaaz: the camera, but I know part of you is seething because you want that top 10 spot. And so do I.
31:03
Ejaaz: So we are going to be gutting for this. But what a crazy start.
31:07
Josh: Seething but grateful. Like man, all the people who are listening here,
31:10
Josh: normally it's the real ones that make it to the end of the episodes.
31:12
Josh: So if you're watching still, chances are you're one of the real ones that made this possible.
31:16
Josh: So thank you from like the bottom of my heart, seriously. Like the support and
31:21
Josh: the sharing with your friends and the liking and the subscribing and the rating.
31:25
Josh: Like it really makes a meaningful difference.
31:27
Josh: And we're seeing that support reflected in the numbers for the first time ever
31:30
Josh: really through the Spotify wrapped. And it is so, so awesome. Like it really, it's...
31:35
Josh: It's one thing to see the numbers on the screen but it's a second to actually
31:39
Josh: see the the impact that it has through all of the comments through everyone
31:42
Josh: sharing through just like the nice messages we receive on a daily basis so seriously
31:46
Josh: like thank you so much for the support
31:48
Josh: i think based on this we can give limitless rookie of the year podcast award
31:52
Josh: because like man i think so that was uh it's a pretty crazy run
31:56
Josh: it's a really crazy one we've published 93
31:58
Josh: episodes this is number 94 that will be coming out in a six-month span and if
32:02
Josh: you've listened from number one or if you listen from number 92 um
32:06
Josh: thank you i hope that you have found value in the show i hope that you will
32:10
Josh: continue to find value in the show
32:12
Josh: the frontier is like expanding faster than ever it has never been more exciting
32:16
Josh: to be right on the bleeding edge of it and that's exactly where the show lies so we're going to keep
32:22
Josh: doing more of this i hope you still continue to find it exciting exhilarating
32:27
Josh: we're going to keep working to get better hopefully you can keep working to
32:30
Josh: share this with every single person you know who could be remotely interested in this
32:34
Ejaaz: And we will just continue.
32:35
Josh: Chugging along. Yeah, so who knows? Maybe by next year's Spotify wrapped,
32:39
Josh: we can trim a two from the front of that 21 and just leave it at number one.
32:44
Ejaaz: Oh, that's good. I like that. And the final thing I'll say is,
32:48
Ejaaz: We don't just do podcasts, Josh. We also do newsletters.
32:51
Ejaaz: And I know you guys are reading a bunch of newsletters, but if you'd read our
32:55
Ejaaz: newsletter about a month ago, you would have bought Google stock and you would
32:59
Ejaaz: have been up about 15 to 30% on the stock right now, which is nuts. So I have a thesis.
33:06
Ejaaz: We have a thesis coming out on Amazon next week.
33:09
Ejaaz: So if you want to read that and be notified about that, subscribe to the newsletter.
33:14
Ejaaz: That is all from my end, Josh.
33:16
Josh: That's it. Yeah, and that's Alpha. A few weeks ago, or even a month ago now,
33:20
Josh: he just posted the Google Bowl thesis. And since then, the stock is up quite a bit.
33:24
Josh: So there are some good alpha in these newsletters. I would highly advise you
33:27
Josh: to subscribe. The link is in the description below.
33:29
Josh: But as always, thank you. Seriously, thank you for a great year.
33:32
Josh: We're going to keep on going. The year's not over. It's still early December.
33:34
Josh: We got a lot more to talk about. So we will be back next week.
33:37
Josh: Have an amazing weekend.
33:38
Josh: And we will see you guys then. Thank you.