Zuck's Open Source Playbook: Meta Muse Glimmer and 14-Page Manifesto

🔒 Check Out Our Sponsor: LEDGER AGENT STACK 🔒
https://developers.ledger.com/docs/ai-tools/overview/?utm_source=Audio&utm_medium=Podcasts&utm_campaign=Limitless

------
🌌 LIMITLESS HQ ⬇️

NEWSLETTER:    https://limitlessft.substack.com/
FOLLOW ON X:   https://x.com/LimitlessFT
SPOTIFY:             https://open.spotify.com/show/5oV29YUL8AzzwXkxEXlRMQ
APPLE:                 https://podcasts.apple.com/us/podcast/limitless-podcast/id1813210890
RSS FEED:           https://limitlessft.substack.com/

------
Let's unpack Meta’s return to open source AI and its latest laptop-compatible model, Muse Glimmer.

Alongside its focus on cheaper, private on-device inference, we cover Meta’s broader AI strategy, including personalization, compute spending, competition on price, and its AI glasses and hardware efforts.

------
TIMESTAMPS

0:00 Meta’s Open-Source Pivot
1:43 Why Local AI Matters
5:02 Meta’s Model Strategy
6:39 Zuck’s Open-Weight Thesis
10:43 Flooding the AI Market
13:00 Ads, Data, and Free AI
15:34 Cheaper Models Take Hold
18:41 The Earnings and CapEx Bet
20:33 Compute Is the Real Moat
22:24 What If Meta Fails?
23:48 Meta’s Bigger AI Vision

------
RESOURCES

Josh: https://x.com/JoshKale

Ejaaz: https://x.com/cryptopunk7213

------
Not financial or tax advice. See our investment disclosures here:
https://www.bankless.com/disclosures⁠

Josh works with Anthropic as a contractor. All views expressed are his own and do not represent Anthropic, its leadership, or its affiliates. Nothing in this episode is investment advice.

Episode Transcript

Josh: Meta has pivoted again for the third time and now back to an open source company. (00:00):
undefined

Josh: The story of Meta has been pretty insane. (00:05):
undefined

Josh: Two years ago they were totally open source, hellbent on wreaking havoc on the (00:07):
undefined

Josh: entire marketplace of closed source models. (00:12):
undefined

Josh: They were doing this through their model called Llama and for those who weren't (00:14):
undefined

Josh: familiar, Llama was just kind of like this close to frontier open source model (00:16):
undefined

Josh: that was expected to change the world in a meaningful way. (00:21):
undefined

Josh: It did not go as Zuck foresaw and therefore they went and they closed off all (00:23):
undefined

Josh: their model development and went closed source just like all of the other labs now (00:28):
undefined

Josh: they're going back they're transitioning back to open source and as a result (00:32):
undefined

Josh: they've spent a tremendous amount of money on this so what happened in the last (00:35):
undefined

Josh: earnings report after all the capex spend came in (00:39):
undefined

Josh: zuck just dropped a 14 page paper explaining exactly why they're making a re-entry (00:41):
undefined

Josh: into the open source world (00:46):
undefined

Josh: That entry is defined by a singular model that we're going to talk about right (00:48):
undefined

Josh: now called Muse Glimmer. And Muse Glimmer is a really impressive model that (00:51):
undefined

Josh: fits on your laptop, runs locally, and I think it's going to shift the world of open source AI. (00:55):
undefined

Ejaaz: I think this is a great move from Zuck and Meta. And this is coming from probably (01:01):
undefined

Ejaaz: the Meta's biggest bear in the past on this show. (01:06):
undefined

Josh: Yeah, definitely. We're professional haters. I should preface this with. (01:10):
undefined

Ejaaz: I'm a professional hair of Zuck and Meta for the last six months, (01:13):
undefined

Ejaaz: but I love to see this pivot. (01:17):
undefined

Ejaaz: And I'll explain why. Number one, Metz's entire philosophy around AI has always (01:18):
undefined

Ejaaz: been open source. And his simple reasoning behind that is he believes like everyone (01:24):
undefined

Ejaaz: should get access to this thing, right? (01:28):
undefined

Ejaaz: If everyone has access to this thing, there'll be no centralized core power (01:30):
undefined

Ejaaz: for AI. There's no AI overlord going forwards. He then kind of pivoted on that, (01:34):
undefined

Ejaaz: like you mentioned, and now he's back with not one, but two new open source (01:37):
undefined

Ejaaz: models. So what are the models? (01:41):
undefined

Ejaaz: Number one, the headline is called Muse Glimmer. It's an open-weight, (01:43):
undefined

Ejaaz: open-source, 30 billion parameter model. (01:48):
undefined

Ejaaz: Now, if you're not familiar with the sizes of these relative models, (01:51):
undefined

Ejaaz: this is a pretty tiny model. (01:54):
undefined

Ejaaz: It can actually fit and run on your MacBook. It's 20 gigabytes of data or RAM (01:56):
undefined

Ejaaz: that it requires, which is significantly smaller than any other frontier model, (02:01):
undefined

Ejaaz: which is like hundreds and hundreds and hundreds of gigabytes. (02:04):
undefined

Ejaaz: And the way that they were able to achieve this is they took a bigger model, (02:07):
undefined

Ejaaz: and they made it smaller. It's something called quantization. (02:11):
undefined

Ejaaz: Now, why did they create this model? (02:14):
undefined

Ejaaz: Well, he believes or Meta believes that locally run private AI models. (02:16):
undefined

Ejaaz: Should be the future of how AI models are dispersed amongst everyone. (02:22):
undefined

Ejaaz: Right now, we pay subscriptions to get access to a model that lives in the cloud, (02:25):
undefined

Ejaaz: that lives in a centralized company like Anthropic, like OpenAI. (02:30):
undefined

Ejaaz: With this model, you can kind of run it locally. (02:33):
undefined

Ejaaz: And that brings several different advantages that we've mentioned quite a few (02:36):
undefined

Ejaaz: times on this show. The first one being it's cheap as hell. (02:39):
undefined

Ejaaz: The second one being it can run on your private data, so it can become a lot (02:42):
undefined

Ejaaz: more smarter and attuned to you specifically. (02:45):
undefined

Ejaaz: And the third thing is it's very efficient when it serves you answers and results (02:49):
undefined

Ejaaz: to your different prompts. So I think it's a good model, but it's certainly not frontier in any way. (02:54):
undefined

Josh: Yeah, it's far from frontier. And I imagine they're probably not really concerned (02:59):
undefined

Josh: about that. We mentioned this on previous episodes where there is this fight (03:02):
undefined

Josh: for the frontier, but now there are different frontiers. (03:06):
undefined

Josh: There is the frontier of intelligence, which Anthropic and OpenAI are kind of (03:09):
undefined

Josh: working on. Then there's the frontier of cost and efficiency per token. (03:12):
undefined

Josh: That's kind of where I see meta of place in themselves. In fact, (03:16):
undefined

Josh: they're fighting on multiple fronts at once. (03:19):
undefined

Josh: And it seems like Muse Glimmer is sitting at the very base of that because it (03:21):
undefined

Josh: is such a small model, it's so quantized, it's able to actually run on a local machine. (03:25):
undefined

Josh: I think the fully capable largest version of this takes about 55 gigabytes of (03:29):
undefined

Josh: memory, which runs on my MacBook that I have sitting on my desk right now. (03:34):
undefined

Josh: And that's really impressive. (03:37):
undefined

Josh: This puts them in a race that not many other companies are competing on, (03:38):
undefined

Josh: which is just like that edge inference local compute thing that I think we can (03:41):
undefined

Josh: expect to see out of Apple Fairleague soon. (03:46):
undefined

Josh: Like with the new Siri going to be running on local models, this is very much (03:48):
undefined

Josh: feels like a response to that. (03:52):
undefined

Josh: Where now for the first time you're able to run this free inference that's fairly (03:53):
undefined

Josh: capable it has a somewhat large context window i believe a couple hundred thousand (03:57):
undefined

Josh: tokens i'm not sure the exact number but a couple hundred thousand (04:02):
undefined

Josh: Tokens of context it has you know pretty good outputs in terms of tokens per second (04:04):
undefined

Josh: and it runs at a level that you would expect a frontier model would have ran at (04:09):
undefined

Josh: maybe say 18 months ago so it's not going to do any crazy difficult complicated (04:13):
undefined

Josh: tasks but for something that can run on your machine you could use anywhere (04:17):
undefined

Josh: in the world at any given time and you can trust it to manage all of your data. (04:21):
undefined

Josh: So for example, if you have a machine runs a lot of sensitive information, (04:25):
undefined

Josh: you want to put health information through that, you might not want to give (04:28):
undefined

Josh: to a public facing model. (04:32):
undefined

Josh: This is a really easy way of doing that in a way that we haven't really had before. (04:33):
undefined

Josh: And I think that's kind of step one in this new strategy that Zuckerberg's going (04:37):
undefined

Josh: for in, I guess you could call it like scorched earth 2.0, where why do you (04:41):
undefined

Josh: open source models? Well, you want everyone to go off, build on them to remove (04:45):
undefined

Josh: margins away from these frontier labs. (04:49):
undefined

Josh: And this seems like the first version of that that people are are kind of (04:51):
undefined

Ejaaz: Excited about and you see him apply this logic to quite a few of his models (04:54):
undefined

Ejaaz: now like he also open weighted or he plans to open weight. (04:59):
undefined

Ejaaz: Muse spark 1.2 which if you have been following along with the show (05:02):
undefined

Ejaaz: was a brand new model metas frontier model that they released i think it was (05:07):
undefined

Ejaaz: like a week and a half ago and what made muse spark 1.2 really special is it's a really good (05:11):
undefined

Ejaaz: coding model in the way that they've kind of engineered AI agents to kind of (05:17):
undefined

Ejaaz: like replicate amongst themselves, (05:20):
undefined

Ejaaz: take your code and go work in parallel. (05:22):
undefined

Ejaaz: So it's a really efficient way of working. And that's kind of like Meta's core (05:25):
undefined

Ejaaz: thesis, I feel, with a lot of these models that they're releasing. (05:29):
undefined

Ejaaz: It's like they're not trying to get the most intelligent model, (05:31):
undefined

Ejaaz: but they're trying to produce the most cheapest and affordable model that still (05:34):
undefined

Ejaaz: gets about 90 to 95% of the work done. (05:38):
undefined

Ejaaz: And Zuck has said in many times over the last couple of months that this is (05:41):
undefined

Ejaaz: Meta's core philosophy. They believe that locally private-run, (05:44):
undefined

Ejaaz: effective AI models will eventually outcompete the frontier. Now. (05:47):
undefined

Ejaaz: When you kind of compare these models, right, you're probably listening to this (05:51):
undefined

Ejaaz: and you're thinking, okay, who cares? Meta has a model. I haven't used Meta (05:54):
undefined

Ejaaz: AI intelligence in God knows how long. (05:56):
undefined

Ejaaz: Why should I use this model? Well, if you are using, say, the Anthropic or OpenAI (05:58):
undefined

Ejaaz: API and you're spending a couple thousand bucks a month, even a couple hundred (06:03):
undefined

Ejaaz: bucks a month, if you're an enterprise that's spending 10 to hundreds of millions of dollars per year, (06:06):
undefined

Ejaaz: this might be a model that might be more bang for your buck. (06:11):
undefined

Ejaaz: And listen, you're not going to use it for everything. You're not going to get (06:13):
undefined

Ejaaz: it to plan your entire company strategy. (06:15):
undefined

Ejaaz: You're not going to get it to build a brand new app from scratch and expect (06:17):
undefined

Ejaaz: it to be the best thing, but you'll use it as a workhorse, maybe in conjunction (06:21):
undefined

Ejaaz: with some of these more intelligent models. (06:25):
undefined

Ejaaz: So it's great to see Zuck kind of like reprise Meta's role in open source. (06:27):
undefined

Ejaaz: If we remember, like in the past, it was Lama and then he kind of like pivoted (06:32):
undefined

Ejaaz: and everyone was like kind of super angry about this. I'm glad that he's doing this. (06:35):
undefined

Ejaaz: The question is like, why is he doing this, right? Like, you know, (06:39):
undefined

Ejaaz: why would this make sense for Meta, which is a for-profit publicly traded company, (06:43):
undefined

Ejaaz: he sucks, still wants to make a lot of money. (06:46):
undefined

Ejaaz: How does this make any sense at all? And he kind of dropped his entire thesis (06:48):
undefined

Ejaaz: yesterday when he announced that he's open weighting a bunch of these models. (06:52):
undefined

Ejaaz: He goes, I believe everyone should have access to super intelligence. (06:56):
undefined

Ejaaz: And I wrote a long piece about Meta's philosophy and values for building that world. (07:00):
undefined

Ejaaz: And essentially like this future that he describes balances on like three core (07:05):
undefined

Ejaaz: pillars. Number one is he believes everyone should be individually empowered, (07:09):
undefined

Ejaaz: aka get access to their own private AI model and that it shouldn't be centered (07:13):
undefined

Ejaaz: in specific frontier AI labs. (07:17):
undefined

Ejaaz: Number two, he believes the thing that pushes humanity going forwards is invention. (07:19):
undefined

Ejaaz: So humans' ability and intention to invent brand new things should drive who (07:23):
undefined

Ejaaz: gets access to these different models. And that could be anyone and everyone. (07:28):
undefined

Ejaaz: And the third thing is he just believes in a core balance of power, (07:31):
undefined

Ejaaz: which is something ironically that adversaries over in China have been advocating (07:34):
undefined

Ejaaz: for a while now. So it's nice to see like a frontier American AI lab advocate for the same. (07:39):
undefined

Josh: It is, but it feels a little bit like I can't, I can't feel like this is a high (07:44):
undefined

Josh: conviction post because the sentiment has changed so much. (07:49):
undefined

Josh: It's not like he's like they started this way, but then they changed this way (07:52):
undefined

Josh: when it wasn't working. And now that things are working, they're going to go (07:55):
undefined

Josh: back and start to believe this. It feels like they're just not really standing strong. (07:58):
undefined

Josh: On morals over time and this this could be the wrong guess but i mean he's he's (08:01):
undefined

Josh: here saying basically the exact arguments that everyone's been arguing for a (08:06):
undefined

Josh: very long time it's like we don't trust a single monoculture to handle this (08:09):
undefined

Josh: we don't believe that the future of humanity is going to be (08:12):
undefined

Josh: best served by two to three companies we we don't believe that any company who (08:16):
undefined

Josh: thinks that ai will eliminate most jobs would rush to build it faster than anyone (08:20):
undefined

Josh: else and these are the arguments that we've heard for a really long time and (08:25):
undefined

Josh: we just get these echoed over and over again (08:28):
undefined

Josh: as the company starts to fit more of this mold i mean the essay was interesting (08:31):
undefined

Josh: what i will be most interested in is just seeing what the actual downstream effects are of (08:35):
undefined

Josh: these new essays and like hopefully it seems like zuck is getting more confident (08:40):
undefined

Josh: that they're going to figure this out that they're going to actually be releasing (08:44):
undefined

Josh: models that can meaningfully (08:46):
undefined

Josh: shift the way people engage with ai and this is probably like the starting gun (08:48):
undefined

Josh: he's like hey new era new company new ethos we're going to start running with (08:53):
undefined

Josh: this and see how it goes and so far it looks like they're kind of on their way (08:57):
undefined

Josh: like things are pretty good i was looking at (09:01):
undefined

Josh: Muse glimmer and i actually i want to try this one so my plan is to install (09:03):
undefined

Josh: this on my local machine today and give it a go (09:06):
undefined

Josh: because one of the things that is unique about this one is for people who are (09:08):
undefined

Josh: still using open claw remember that thing um this actually is a very agentic (09:11):
undefined

Josh: model that runs end-to-end for task completion (09:15):
undefined

Josh: it uses agents it has tool use it has multi-step reasoning it has multimodal (09:18):
undefined

Josh: input and output it seems like a really strong and powerful model yeah we could (09:23):
undefined

Josh: see here kind of how good it is at creating visuals, how good it is in benchmarks. (09:27):
undefined

Josh: And it's an actually usable model in a way that I don't think a lot of these (09:30):
undefined

Josh: others have been. So I'm taking this 14 page paper with a grain of salt. I... (09:34):
undefined

Josh: Read some of it. I didn't get my way through all of it. EJS, (09:38):
undefined

Josh: I know you said you read the entire thing. (09:41):
undefined

Josh: I'm excited for Meta. I want them to win. So I'm hoping this is the new starting (09:43):
undefined

Josh: point for them to release better models, to really get back in the game, (09:47):
undefined

Josh: to compete. Because, I mean, yeah, duopolies are less than ideal. (09:51):
undefined

Josh: We want a lot of people competing for a lot of different parts of the pie. (09:54):
undefined

Josh: And if Meta is going to come out here and do it and do so open source, (09:57):
undefined

Josh: that's great. And you got to imagine Jensen sitting over here pretty happy too, Mr. (10:00):
undefined

Josh: Open Source King, that now there's another company in the running for making (10:04):
undefined

Josh: really compelling open source models in the United States. (10:08):
undefined

Ejaaz: I think it's actions over words, though. Zuck has talked this talk for a while, you're right. (10:10):
undefined

Ejaaz: But he's also the guy that's releasing like the frontier open source models in America. (10:14):
undefined

Ejaaz: You see all the other AI labs. Look, I'll give you an example here. (10:19):
undefined

Ejaaz: Satya Nadella, bunch of people signed that they're in favor of open source and (10:22):
undefined

Ejaaz: open weight movements in general, but they haven't released any frontier open models themselves. So (10:26):
undefined

Ejaaz: I look at actions over words and Zuck, yeah, he's flailed around quite a bit. (10:31):
undefined

Ejaaz: And I've been a big critique of that, but he's come back. (10:35):
undefined

Ejaaz: And I'm going to judge him by his actions going forwards and we'll see what (10:39):
undefined

Ejaaz: he actually uses this for. (10:42):
undefined

Ejaaz: Now, the bad case for this is this is like an old school business move that (10:43):
undefined

Ejaaz: he's making, which is let's flood the market with intelligence that is basically free. (10:49):
undefined

Ejaaz: So we commoditize the paid subscriptions that other frontier labs are paying for. (10:54):
undefined

Ejaaz: And that brings down their value, that brings down their profit margins, (11:00):
undefined

Ejaaz: their revenue, and it allows Meta more time to catch up. Now, (11:04):
undefined

Ejaaz: is this a costly and risky endeavor? Yes. (11:08):
undefined

Ejaaz: Meta has burned, what is it, like $30 billion in acquiring 150 staff to produce (11:11):
undefined

Ejaaz: a model that is sub-frontier. If you were a shareholder, you're probably looking (11:17):
undefined

Ejaaz: at this and thinking, Zuck, what are you doing? (11:21):
undefined

Ejaaz: But if you look at what he's been investing in outside of that. (11:23):
undefined

Ejaaz: I think he's on track to spend, I think, $210 billion this year through the (11:27):
undefined

Ejaaz: first half of next year on Datacenter CapEx. (11:33):
undefined

Ejaaz: He is one of the largest holders of NVIDIA GPUs. He's secured like one of the (11:37):
undefined

Ejaaz: largest arsenals of NVIDIA Rubins, which is their next-gen GPU architecture. (11:42):
undefined

Ejaaz: Second only to Elon Musk, you have to question, okay, what's he going to use (11:48):
undefined

Ejaaz: all of this compute for? Well, he could rent it out, which is what SpaceX is (11:52):
undefined

Ejaaz: doing and make a bag from that. (11:55):
undefined

Ejaaz: But he's also probably going to train a bunch of frontier models. (11:57):
undefined

Ejaaz: And the biggest signal for whether he's actually going to be able to pull this (11:59):
undefined

Ejaaz: off is model release cadence. (12:03):
undefined

Ejaaz: And in the last, I think 90 days, he's released two models, but he's going to (12:05):
undefined

Ejaaz: release another two more over the next, I think it's like one to two months. (12:11):
undefined

Ejaaz: I think you mentioned in some tweet thread, Elon Musk is planning to do the same. (12:15):
undefined

Ejaaz: So I'm seeing this trend where the companies that, even though they're a little (12:18):
undefined

Ejaaz: bit behind in frontier intelligence, who have the most compute, (12:22):
undefined

Ejaaz: who have the most capital, are still able to kind of ship models iteratively (12:25):
undefined

Ejaaz: and maybe respond to the market quicker than some of these other frontier labs. Now, (12:29):
undefined

Ejaaz: I appreciate that that's optimistic thinking, but I do think Zuck is back in (12:32):
undefined

Ejaaz: the game and I'm excited to see what these models are like. I want to play around (12:36):
undefined

Ejaaz: with this Muse Spark 1.2 and this new one that I can run on my laptop and see (12:40):
undefined

Ejaaz: what fun things we can come up with. (12:45):
undefined

Josh: Yeah, I was trying to unpack the reasoning. I'm like, okay, why did he make (12:47):
undefined

Josh: this pivot again? It seems like there's so much uncertainty in the strategy of meta so far. (12:50):
undefined

Josh: And there really does appear to be some pretty compelling reasons that are kind (12:55):
undefined

Josh: of underlying this decision. (12:59):
undefined

Josh: Because when you think about meta the company i was like okay what does meta (13:01):
undefined

Josh: actually do meta is just an attention salesman (13:03):
undefined

Josh: their biggest part of the business is advertising and sales through that (13:06):
undefined

Josh: all their services are mostly free to use you can pay for a blue check mark if you want (13:09):
undefined

Josh: but most their users are free and they have this tremendous user base so they (13:13):
undefined

Josh: can sell a subscription to those people but those people are not used to and (13:17):
undefined

Josh: not very probably likely excited to pay (13:22):
undefined

Josh: twenty dollars a month to use an ai system that can be offered for free because (13:24):
undefined

Josh: you have to assume the average user doesn't need frontier class intelligence (13:28):
undefined

Josh: so if the model is attention (13:31):
undefined

Josh: And if you can improve your core business through AI, then that's probably what (13:34):
undefined

Josh: they want to do. And that's probably why open source makes sense. (13:38):
undefined

Josh: I mean, yeah, meta already announced it wants to use your AI chats to personalize the ads (13:41):
undefined

Josh: and it wants you to use the AI to build a large profile on you, to build (13:46):
undefined

Josh: a josh.md, an ejaz.md that has all of our preferences, all the things we care (13:50):
undefined

Josh: about, that gets updated in real time, and that has an increasing amount of (13:55):
undefined

Josh: ways in which you can collect context on you. (13:59):
undefined

Josh: And if you are trying to do this then you really just want these models to be (14:02):
undefined

Josh: free and accessible and open to collect as much data as possible to feel like (14:05):
undefined

Josh: there is no friction to getting in the game and then also just to kind of like just chip away at the (14:08):
undefined

Josh: the frontier labs margins where if there's someone who is deciding between a (14:14):
undefined

Josh: 20 model or just going into their instagram app and using this cool new chatbot (14:17):
undefined

Josh: that they have like chances are a lot of people are going to choose to use the (14:22):
undefined

Josh: chatbot and that puts them (14:25):
undefined

Josh: in a way i guess it's kind of unique to them where they're they're competing (14:26):
undefined

Josh: on edge inference in a way somewhat with apple and what they will be doing with (14:30):
undefined

Josh: the new siri but they're also competing with google in the ads business if (14:33):
undefined

Josh: they're going to have just as much if not more data than google has (14:37):
undefined

Josh: When it comes to search queries and like think about how custom the advertising (14:42):
undefined

Josh: can be and not even advertising but content creation in general if meta understands (14:45):
undefined

Josh: what type of content you like how you engage with certain types of content it (14:50):
undefined

Josh: can serve this very custom profile that allows custom generated content just (14:53):
undefined

Josh: for you all in the effort of selling you more ads. (14:57):
undefined

Josh: And I think this is the reality of Meta is as they've tried to pivot to hardware, (15:00):
undefined

Josh: they've done all these other exploratory things. (15:03):
undefined

Josh: It still just comes down to the core business is selling people ads. (15:05):
undefined

Josh: And it's like understanding who their user is, pairing them with a seller and (15:09):
undefined

Josh: collecting commission off of those sales. (15:12):
undefined

Josh: And I think this is probably just a natural extension of that. (15:15):
undefined

Josh: And they're going to continue to make these products more compelling and more (15:17):
undefined

Josh: capable through these very lightweight AI models. (15:22):
undefined

Ejaaz: At the end of the day, Meta is obviously a for-profit company. (15:24):
undefined

Ejaaz: They're trying to make money and ads, as you said, is like their main business. (15:29):
undefined

Ejaaz: So, you know, they're definitely going to focus on that. (15:32):
undefined

Ejaaz: But I also think in the wider market of AI, a bunch of things have shifted recently (15:35):
undefined

Ejaaz: that we just can't ignore. (15:40):
undefined

Ejaaz: The first one being, you don't need the most intelligent model to do anything (15:41):
undefined

Ejaaz: and everything for you. I was catching up with a friend who is pretty high up (15:46):
undefined

Ejaaz: at a top, I would say, Fortune 100 company. (15:49):
undefined

Ejaaz: And I asked him, like, you know, how are you using AI these days? (15:53):
undefined

Ejaaz: Like, do you use private models? Do you use the standard models from Anthropic and OpenAI? (15:56):
undefined

Ejaaz: And he said, I've shifted completely to some of these open source models, (16:02):
undefined

Ejaaz: as well as just cheaper models in general. (16:06):
undefined

Ejaaz: And I get it to do the bulk of my work whilst I use Accord or GPT 5.6 to brainstorm (16:08):
undefined

Ejaaz: kind of like the architecture of what I want to build. (16:14):
undefined

Ejaaz: And I'm seeing this pattern reflected across a number of different enterprises. (16:16):
undefined

Ejaaz: It's certainly been reported across social media like every single day. (16:20):
undefined

Ejaaz: We're seeing the shift of companies thinking, I'm not going to just needlessly (16:23):
undefined

Ejaaz: spend a bunch of money on frontier intelligence. I certainly will for some things, (16:27):
undefined

Ejaaz: but I want to explore some of the cheaper options, right? And you can look at (16:31):
undefined

Ejaaz: the business models of the frontier labs themselves, right? (16:34):
undefined

Ejaaz: I think over the last 90 days, eight models have been released, (16:37):
undefined

Ejaaz: three of them alone from OpenAI, right? Why didn't they just release GPT-5.6 (16:41):
undefined

Ejaaz: Sol? Why did they release Luna? Why did they release Terra? (16:46):
undefined

Ejaaz: Why did they slash the price of Luna by 80% a month ago? Why did Anthropic yesterday, (16:49):
undefined

Ejaaz: I think, announce that they're going to keep Claude Sonnet, I believe, (16:56):
undefined

Ejaaz: at the cheapest price ever indefinitely. (17:01):
undefined

Ejaaz: I think these companies are realizing that cheaper models are definitely applicable (17:03):
undefined

Ejaaz: and have an advantageous incentive for the customers that they're trying to (17:07):
undefined

Ejaaz: sell to, whether it's enterprises or consumer customers. (17:11):
undefined

Ejaaz: And you need more bang for your buck. Zuck realizes this, and he's seizing the (17:14):
undefined

Ejaaz: opportunity, but he's being more aggressive with the open source side of things. (17:19):
undefined

Ejaaz: And I think Zuck has enough capital at Meta to be able to keep this train going (17:22):
undefined

Ejaaz: for a while now. So I'm interested to see how what meta and specifically SpaceX (17:26):
undefined

Ejaaz: looks like over the next couple of months. (17:32):
undefined

Ejaaz: I think SpaceX is also teasing their next model and they've released like two (17:34):
undefined

Ejaaz: over the last like three weeks today, which should hopefully be releasing. (17:38):
undefined

Ejaaz: But yeah, I think there's something that (17:41):
undefined

Ejaaz: they're going to pull off something cool. (17:44):
undefined

Josh: Yeah i'm looking forward to it i'm looking forward to all the agents that they're (17:45):
undefined

Josh: going to spawn from these new models because if you are working with agents (17:49):
undefined

Josh: you know you know who we got to talk about and that's ledger our sponsor of (17:51):
undefined

Josh: this episode um because they have this infrastructure for agents in which it's kind of three parts (17:55):
undefined

Josh: agents propose then the human approves and then (18:00):
undefined

Josh: the ledger signer enforces the ledger signer is this mostly comes in the form (18:03):
undefined

Josh: of a hardware wallet in which you can approve or deny we talk a lot about security (18:07):
undefined

Josh: and how a lot of these agents can kind of go rogue and do things you may not want (18:11):
undefined

Josh: ledger offers an agent stack which is a series of open source tools that allows you to (18:14):
undefined

Josh: evaluate what these agents are doing across the way you could tell it to do (18:19):
undefined

Josh: things like rebalance your wallet (18:23):
undefined

Josh: or adjust whatever parameters you want to adjust and then it will propose you (18:24):
undefined

Josh: approve and the ledger signer enforces it works with cloud code (18:29):
undefined

Josh: codex all of the places that you work with ai on a daily basis so thank you (18:31):
undefined

Josh: so much to ledger for sponsoring this part of the episode you can get their (18:36):
undefined

Josh: products at the link in the description below. (18:39):
undefined

Josh: EJS, you mentioned capex spend, or you actually didn't mention capex spend. (18:41):
undefined

Josh: You mentioned strong earnings, but I'm going to mention capex spend, (18:45):
undefined

Josh: which is their free cashflow. (18:48):
undefined

Josh: Like we have to talk about the earnings quickly because like we do want to get (18:51):
undefined

Josh: a little bit economic here. Like how viable is this strategy? (18:54):
undefined

Josh: Their free cashflow this year, which means basically the money that comes in (18:56):
undefined

Josh: and out is down 91% year over year. (18:59):
undefined

Josh: They're spending, They're spending big on a few key pillars. (19:04):
undefined

Josh: One of them is this open source bet, which does not have an immediate ROI because, (19:08):
undefined

Josh: I mean, again, they're offering these weights for free. (19:12):
undefined

Josh: The other is on their hardware platform. One of the things that is like still (19:14):
undefined

Josh: shocking to me, and I'm not sure that many people know about this, (19:19):
undefined

Josh: Meta owns 70 to 80% of the entire AI glasses market. (19:22):
undefined

Josh: And this is not a big market, but this is a market that is meaningful and that (19:27):
undefined

Josh: probably shouldn't be overlooked. (19:31):
undefined

Josh: This is done exclusively with a company named Luxottica. Together, (19:34):
undefined

Josh: they have more than tripled Meta's AI glasses sales just in the year of 2025. (19:37):
undefined

Josh: We haven't gotten 2026 numbers yet, but like that is kind of noteworthy it's interesting to see the (19:42):
undefined

Josh: floor of the capex guidance being raised too in 2026 meta is guiding for 145 (19:48):
undefined

Josh: billion dollars of spend which is just this like unfathomably large number (19:56):
undefined

Josh: and when you think about the places that it's going to land well it's certainly (20:01):
undefined

Josh: open source it's certainly hardware but like what else are they really just (20:04):
undefined

Josh: going to go all in on building out data centers and gpus to make larger models is that (20:08):
undefined

Josh: is that the end game? Because I don't see them competing on the frontier. (20:13):
undefined

Josh: So do they really need that much compute in order to compete for open source, open weight models? (20:17):
undefined

Ejaaz: I think so. I think the number one factor that determines how good your AI offering is, is compute. (20:22):
undefined

Ejaaz: At the end of the day, Elon's going to prove that with SpaceX. (20:29):
undefined

Ejaaz: And I think Zuck is planning to do the same thing. (20:31):
undefined

Ejaaz: We saw NVIDIA or Jensen Huang announced yesterday that he secured an additional (20:34):
undefined

Ejaaz: $500 billion of revenue and capital investment for future compute. (20:39):
undefined

Ejaaz: We see OpenAI partnering with NVIDIA for a similar amount, half a trillion dollars. (20:44):
undefined

Ejaaz: You see Anthropic securing deals with Riot yesterday with an old cryptographic (20:49):
undefined

Ejaaz: mining company to secure even more compute. (20:53):
undefined

Ejaaz: Compute is the ultimate indicator on whether you can create a better model. (20:55):
undefined

Ejaaz: It is the one thing that is consistent across all competitors in the AI race. (20:58):
undefined

Ejaaz: And I think Zuck believes if he amasses enough of it, if he secures enough allocation (21:02):
undefined

Ejaaz: of frontier NVIDIA GPUs, he'll be able to pull off and build better models. (21:06):
undefined

Ejaaz: Now, whether the open source strategy plays out, again, it's that strategy that (21:12):
undefined

Ejaaz: I think I mentioned earlier, which is flood the market with commoditized intelligence, (21:15):
undefined

Ejaaz: cut down or cut into revenue and margins of your competitors. (21:20):
undefined

Ejaaz: Yeah, Meta's CapEx takes a beating in the short term, but hopefully that plays (21:24):
undefined

Ejaaz: better in the long term when you have personalized agents that all run on Meta's (21:29):
undefined

Ejaaz: infrastructure and consumer apps. (21:32):
undefined

Ejaaz: Also, remember, Meta has like 4 billion consumers right now. (21:34):
undefined

Ejaaz: I was scrolling on Reels yesterday, which I don't like to admit, (21:37):
undefined

Ejaaz: but a ton of them are now Meta glasses. (21:41):
undefined

Ejaaz: I don't know if you've seen this, Josh, but it's this POV from a bunch of different people. (21:44):
undefined

Ejaaz: And I think that they don't have the best hardware, but they have one of the (21:48):
undefined

Ejaaz: most amazing distributions ever. I think 4 billion, and then you compare that (21:51):
undefined

Ejaaz: to Apple, which has, I think, 3.5 billion live Apple devices right now. (21:55):
undefined

Ejaaz: It's probably a lot more. I think that data is outdated. (21:59):
undefined

Ejaaz: They have a real competing moat here. And I look at Google, who has recently (22:01):
undefined

Ejaaz: gone into negative cash flow. Meta isn't quite there yet. (22:06):
undefined

Ejaaz: So you could argue that they're being somewhat conservative. (22:09):
undefined

Ejaaz: They have other businesses that are earning a ton of money that can support this for a while, but, (22:12):
undefined

Ejaaz: They don't have infinite runway. And, you know, maybe we'll see Meta raising (22:16):
undefined

Ejaaz: external capital similar to how Intel and Google has over the last couple of (22:19):
undefined

Ejaaz: months. I don't know. We'll see. (22:23):
undefined

Josh: Yeah. And I was going to ask the question whether or not this CapEx spend is (22:25):
undefined

Josh: existential. Like what happens if they fail again? (22:28):
undefined

Josh: Because this isn't their first time trying to get a frontier model that works (22:31):
undefined

Josh: really well and the first time did not go as planned. And then I was thinking, well, you know what? (22:34):
undefined

Josh: At least they have the GPUs and they could just go resell those and make a great business out of it. (22:39):
undefined

Josh: So there's this really weird part of the market now where it's very advantageous (22:42):
undefined

Josh: to just spend money on building data centers. And even if you can't figure out (22:46):
undefined

Josh: how to use them yourself, you just sell them off to the next best guy and you (22:49):
undefined

Josh: can make a tremendous amount of money from that infrastructure build out. (22:52):
undefined

Josh: So I think at the very least, you could find a lot of solace in that idea in (22:55):
undefined

Josh: terms of users. Like that's a great point that, I mean, 4 billion users is half of the planet. (22:58):
undefined

Josh: There's not many companies in the world that I'm not sure if there's any companies (23:02):
undefined

Josh: in the world with that quite as many users as meta, or there's a very small handful of them. (23:06):
undefined

Josh: So if they're able to figure out how to, I mean, weaponize is like a terrible (23:10):
undefined

Josh: word for this, but like able to leverage their AI tools to further create compelling (23:14):
undefined

Josh: offerings on their social media platforms, like that's a pretty big value unlock. (23:20):
undefined

Josh: So I'm excited. I'm cautiously optimistic on meta. (23:23):
undefined

Josh: Like I'm still looking for signals that they are going to figure it out. (23:27):
undefined

Josh: But like you mentioned, when you're watching reels, a lot of it are coming from (23:30):
undefined

Josh: the glasses. It's this novel form of creating content. It's this cool new paradigm (23:33):
undefined

Josh: that every other company is working for. And right now, I mean, (23:36):
undefined

Josh: say what you want, like they are in the lead. (23:39):
undefined

Josh: It's maybe not the best piece of hardware in the world, but they have the best one available. (23:41):
undefined

Josh: And we'll see what happens with the company. But that's kind of the update. (23:44):
undefined

Josh: Meta is going open source again. The boys are back and they are spending a lot (23:49):
undefined

Josh: of money and they are very clearly hellbent on making the best open source models (23:53):
undefined

Josh: that they can and implementing them. (23:57):
undefined

Josh: Across the entire product stack while continuing to iterate on their hardware (24:00):
undefined

Josh: products and weird science projects like remember the the pupil dilation tracker (24:03):
undefined

Josh: and like the this brainwave scanner that they're using to like understand how content affects you and (24:07):
undefined

Josh: it's going to be a weird future with meta but they're working on it nonetheless (24:13):
undefined

Josh: full speed ahead and um yeah we'll just continue to cover it as we go along this journey here (24:17):
undefined

Ejaaz: Also i will just say this if you're listening to this episode and you are a (24:22):
undefined

Ejaaz: builder, there has never been a better time to be alive. (24:27):
undefined

Ejaaz: Like all these tools, all these models, all these products that are being released (24:31):
undefined

Ejaaz: over the last couple of months that are being increased in frequency are all (24:35):
undefined

Ejaaz: for you to trial test out. (24:39):
undefined

Ejaaz: If it's too expensive for you to buy a Claude Mac subscription, (24:41):
undefined

Ejaaz: play around with some of these open source models. If you don't think that the (24:44):
undefined

Ejaaz: American frontier open source models are good enough, play around with some (24:47):
undefined

Ejaaz: of these open weight or open source Chinese models. (24:49):
undefined

Ejaaz: There is no excuse to not be using these things. (24:52):
undefined

Ejaaz: If you have access to it, if you have a project or a hobby or an idea in your (24:56):
undefined

Ejaaz: mind, go out there, try these things that we're talking about and honestly give (25:00):
undefined

Ejaaz: us feedback because we'd love to hear from builders to see what you guys are (25:04):
undefined

Ejaaz: actually doing with the thing. (25:07):
undefined

Ejaaz: Maybe we're missing something and maybe you can kind of like shine a light on (25:08):
undefined

Ejaaz: something that we're completely missing. (25:11):
undefined

Ejaaz: But we will be covering anything and everything on Limitless going forwards. (25:13):
undefined

Ejaaz: That is the end of this episode. You're all caught up on all of things meta (25:17):
undefined

Ejaaz: related. They've done a pivot. (25:22):
undefined

Ejaaz: They've turned it around and I'm excited to see where the company is in a couple months' time. (25:23):
undefined

Ejaaz: That being said, wherever you're listening to us, if it's on YouTube and you (25:28):
undefined

Ejaaz: aren't subscribed, if the notifications aren't turned on, please do so. (25:31):
undefined

Ejaaz: It helps us out massively. (25:34):
undefined

Ejaaz: If you haven't left us a comment on Spotify, Apple Music, or even YouTube as (25:36):
undefined

Ejaaz: well, please do so. We have been loving the comments, by the way. I think we've (25:39):
undefined

Ejaaz: gotten in the in the realm of like i don't know it's like a thousand two thousand (25:43):
undefined

Ejaaz: plus comments over the last like five episodes keep them coming we love hearing (25:48):
undefined

Ejaaz: from you guys we respond to anything and everything, (25:51):
undefined

Ejaaz: as soon as we have like free time available um and then yeah we have a newsletter (25:54):
undefined

Ejaaz: that goes out to over a hundred thousand of you we post twice a week i think (25:57):
undefined

Ejaaz: this week we getting three times because josh and i can't keep our gobs shut (26:00):
undefined

Ejaaz: but um josh anything anything well and most. (26:03):
undefined

Josh: Most of all don't forget to gift this to a friend if there's someone else who (26:06):
undefined

Josh: you know that would benefit from watching this every single day who wakes up (26:11):
undefined

Josh: in the morning goes up with a little walk for their dog or like on their morning (26:14):
undefined

Josh: commute that wants something to listen to 25 minutes long that will get sometimes (26:17):
undefined

Josh: a little bit longer but that will get them up to speed on everything like (26:21):
undefined

Josh: it's always much appreciated it's how we grow the best so as always thank you (26:24):
undefined

Josh: so much for making it to the end of the episode real ones only here at the conclusion (26:27):
undefined

Josh: thank you for watching and we will see you in the next episode (26:31):
undefined