Alex Good on Bitcoin's weakening thesis, proof-of-work AI coins, and how AI is changing market cycles

Like Rust for me is like a vacation destination that you have so many good memories.

Yeah. You don't want it remixed

and you're just like, "No, I'm good. I like going to this beach, this place."

Okay.

And I'm happy to keep going back.

He's a lite. He's a lite folks. Uh anyway, we have Alexander Good in the waiting room. Let's bring him in and see if he can hear us. How you doing?

Doing well. Can you hear me?

Yeah. Fantastic. How you doing?

And fantastic uh green screen background or Zoom background.

Yeah, it's it's the new AI. It's uh it's very powerful. Um, great to finally have you on the show. I've enjoyed your your uh posts and for a long time and some of your rants, your your you have some of the best rants on the internet.

Um, appreciate that. I've been been looking forward to this.

Yeah. Uh, h how would you introduce yourself these days? Like how much of uh just uh I I mean I guess set the table for people. Give us a little bit of backstory uh and uh and your thesis right now.

Cool. Um yeah, I used to be an army. Uh I worked at uh Cityroup, Palunteer, Balazny, started a company, sold it, and then I got into crypto. Um and AI changed a lot of obviously what I was working on and what many people were working on. And uh I was really just trying to figure out how to, you know, I I got very very pilled early on on the idea that AI would just destroy people's work product. Um, and a lot of people are getting on that train now, but I I guess I was on it in 2023. So, a lot of the rants were uh coming from processing that and going through my my own AI psychosis. And now I'm like now I'm like post psychosis. Now I'm just like in a co I'm in a codeex terminal uh working on uh building a protocol uh that hopefully can um move the needle for uh for people who who own it and uh also you know just our investors. a lot of money like flow a lot of like attention money everything flowed from crypto to AI there was even that meme like if you're in crypto pivot to AI has has AI strengthened your belief in crypto like what do you think the intersection is because a lot of people jump to like well the agents will need they would couldn't possibly figure out Stripe so they'll have to use a token and I've never really believed that fully I think that there's some opportunity there but it feels like the AI trends. You're a believer in that, but also your faith in crypto broadly has not been shaken and you haven't like pivoted or like gone to like the hot thing, which is not it's just less common these days, I think.

Yeah. I mean, let's be honest. Uh the intersection of crypto and AI has not been a particularly good intersection so far. There's been a lot of, you know, hacking events. Um, every human piece of code that's ever been written is being evaluated by agents right now, which is obviously created like an acceleration in hacking.

Um, X42 has kind of like gotten exploited. Uh, right. So, you know, there's been, you know, there's this kind of uh, you know, you can prompt engineer agents to send you money and that's like not a good thing if you're running a protocol. So, it's like, do you really want a protocol that can just like send people money? Um, I think the thing that's kind of kept me in the space or kept me excited about the space is I was always kind of fundamentally in this because of something I call the doom thesis, which is that, you know, AI might not cause productivity productivity to accelerate. It might actually accelerate, you know, the kind of onset of the surveillance state um or, you know, basically uh a type of world where capital is going to flee the system. Um, and I think that thesis is very much intact. I think that's why you're seeing a lot of crypto like Zcash uh run very very aggressively. Um and I think that's going to continue to be uh a story. Um you know I think that's why Teal moved to Argentina. I think it's sort of like um you know it's in motion.

Where are you on Bitcoin versus other coins? I mean obviously you mentioned Zcash. Like how strong do you think the Bitcoin thesis is these days? um and how how much maxiness needs to be considered in that thesis.

Uh the the Bitcoin thesis is getting weaker uh by the day. Um you have privacy coins accelerating in terms of their adoption. Uh I think it's notable also you know the European Union and other countries tried to kill Monero and were not successful at that. um you know the tornado cash uh rulings in the United States have given clarity to to coins like rail gun which are smaller uh but are good tech and Vitalic uses it and you know their big privacy summits and bitcoin is fundamentally not uh private technology and it also has the unfortunate sort of leverage concentration from Michael Sailor uh who has you know issued a lot of uh preferred and complex convertible uh financial instruments on top of it which have kind of concentrated the ownership and the liquidity profile um and now is kind of driving a market downturn, right? So, so today's STRC drop as well as Micro Strategies equity situation are driving down moves in Bitcoin and unfortunately like it's hard to see that uh structurally reversing. Um,

do you know if there's been a material impact on crypto from not just the the the capital markets moving around and and founders and employees pivoting to AI, but the actual mining operations pivoting to AI because a lot of the a lot of the neoclouds that we know, they got their start in crypto mining and they're sort of out of that uh market. Is that actually affecting crypto in any way? I haven't really followed it. Yeah, I mean it's interesting. Cororeweave um you guys might know uh Vance at Framework, you know, he had a fantastic return for his fund partly because uh you know Cororeweave literally pivoted out of crypto into AI. Um and they're the OG pivot story. So they they were like prejason calacanis. They they did it they did it um

um

yeah you know the JCAL post the if you're in crypto pivot to AI in like 2023. It was post chat GBT so it was a little bit late. Um but it turned out to be you know cracked for a lot of people.

Yeah.

A lot of companies launched very well

difference.

Um so yeah I mean like there is an there is an interesting thing going on right now. Um you guys are probably going to hear about something called Pearl

uh relatively soon. Um it's aggressively backed. There's rumors that Thrive is in it. uh uh and you know the TLDDR is that it's a proofof work AI coin and basically there's a paper that came out in late 2025 um by SG Lang which talks about determinism uh in in basically GPUs which allows you to use AI to mine cryptocurrency. And so this is something my my brother started working on with Ambient which got backed by A16Z and then Pearl uh is an Israeli team which is doing something kind of like similar. They're using map wool mining and they have a partnership with together AAI that got tweeted out. Um right and so so you know essentially the the the argument that a lot of these crypto AI protocols increasingly are trying to make is is something called useful proof of work uh which is the idea that um you know you have a minor that can also deploy a service. So, you know, my brother is delivering verified inference. And then the pearl argument is that you can basically buy uh GLM uh not GLM, it's uh Gemma um on Together AI for a 25% discount because the miners are subsidizing it. Okay.

Now, the funky thing is that like you can also buy it for like 50% less than their subsidized version on Open Router, right? So, so you're like you're kind of like market cap dependent. you need this thing to really really move in order for it to provide a meaningful subsidy. Um, but you know there's an argument that uh I think the the the comp compelling zoomed out argument is that you know TSMC back in 2017 or 2018 you know 10% of their of their product was Bitcoin AS6 and now it's like way less than 1%. And so if you're like okay you want to secure money with proof of work very very high level it's like why wouldn't you secure it with GPUs? um um you know there there arguments why you wouldn't right because you're like it's because you're literally burning GPUs which are productive assets to secure this thing which is even worse than burning energy for Bitcoin mining but that's kind of a separate argument right

how how separate is this from uh the render token render network from Octane where uh individual computers or or folks with compute would would render CGI frames and then earn a token. it was not nearly as decentralized. And then I know Prime Intellect was working on this as well, trying to sort of wrap spare compute and and put it on a on a network that was uh sort of transacted with a coin uh intermediated by that coin. Uh it sounds like this is at a much lower level. Is that correct?

It it's at a Yeah, it's an architecture level. One render and a lot of these coins including Bitensor actually, okay, are are stake coins, right? So, so they're not actually secured by the operation of a GPU. So, so this is kind of like this Pearl thing

is a regime shift. Um, and it's trading at, you know, like a $2 billion FTV on on private, you know, there's there's there's token sales. So, it's starting to move. It's a it's a relatively new story in crypto AI.

Um,

you know, I my my protocol is like a completely different bet. you know, my essential I think the there are some interesting intersections of crypto and AI that people are probably less focused on which is that you know all of crypto is open source. So actually when you use codecs or you use these AI agents they are like phenomenally good at uh you know crypto work because it's in the train because it's in the training data. It's all open source. So if you screw something up, you can just be like, "Hey, can you just double check this with rail guns implementation to make sure that we're doing the same thing and then like you know 5 minutes later you have like a working, you know, orchard shielded transaction on

some forran code that's locked in some old financial institution. It's never seen the light of day, not going to be in the training set whatsoever.

Exactly. And and so there's like an opex efficiency argument in crypto because historically crypex crypto has been like a very wasteful industry. uh you know there's like a lot of money spent on conferences and and just like you know the Ethereum Foundation is like a good example of like how not to run a company and it's it's sort of like

uh there's a lot of waste in crypto and AI is b like crypto is like the ultimate software industry where you're like okay I could credibly reduce engineering by like you know 60% here.

How are you thinking about AI psychosis? You said you went through it, you got to the other side. Like what's the state of AI psychosis? Where's it showing up?

Is it bearish if someone hasn't fallen in and then climb their way out?

True. A true lite versus the, you know, white pill got through the trough of disillusionment is in the plateau of the trough of uh

illumination. Yeah. Something I don't know. But how are you thinking about it? Where where is it cropping up? Where is it bad? What does it take to get out of it if you're in it? Well, you kind of converge on this like conclusion, right? You start with a we start with a psychosis and the threat of the psychosis is that my work is not meaningful and that everything that I've done is basically meaningless because an AI will be able to do it better.

Sure.

And then you actually start encountering the tools and you're like, okay, in the areas I don't understand, AI is really good, but in the areas I deeply understand, AI is like pretty bad, right? You know, there's not like a TVPN where uh there's like two AI avatars that are doing a better job than you guys. I mean, maybe the new A16Z1 is is that and we just don't realize it, right? It could just be that those guys are are are running a really really advanced uh AI system. Um uh there's some evidence. I think I I think it's there's some evidence of that. I'm not going to discount the possibility. Um but

but you know, essentially you're like, "All right,

this is a this is kind of directionally going to obsolete my work."

Yeah. And so I better be the first person to obsolete it, right? Like that's sort of where you end up.

So the way that you get out of AI psychosis is you say like the only way out is through, right? So you're like, all right, how do I how do I literally automate everything that I did before and then how do I also just like

start to think about network effects that were impossible previously. Um so so I'll give you an example like um you know I'm working with uh there's a pretty robust community with my crypto project and we have what we call the data lake um and the data lake is basically like you know hundreds of contributors who are doing tasks all day and then their aggregate intelligence gets kind of like you know summed up by an AI and it comes up with actionable trade ideas right and and it's it's sort of like you're like okay you can make all kinds of alt data now I I think you know palunteer kind of inspired me originally when they started talking about how well they're doing with corporates when they're like, "Hey, like we have the biggest, you know, uh we have the best demand story in history to take unstructured data and turn it into structured data."

And I'm like, "Okay, that's cool in a corporate and government context, but like nobody's ever done that with our Wall Street bets, right? Or like nobody's ever done that with trading information." And I'm like, "Okay, like

I'm just going to do that. I'm gonna take, you know, the palunteer ontology method and apply it to, you know, a group of individuals who create, you know, market alpha. And so then that's how you get out an AI psychosis. You you can sort of identify like what new capability does AI enable that allows you to generate the the type of edge that you would have historically just with a new kind of data primitive. So it's like okay, like it you need to have a story for how you use AI to get out. That's the the short answer.

Yeah, it's interesting. I mean we're sort of at this moment where yeah the AI is is replacing a lot of things but I mean at least as an entrepreneur like I don't know that the moes have actually changed because people are saying like oh like SAS apocalypse like your big pile of code is no longer a moat but if you go back and look at the literature whether it's 0ero to one or Hamilton Helmer seven powers like big pile of code was never a moat it was always cornered resource brand network effect complex coordination like these other things that were not just oh yeah you were the first person to build a lot of momentum around a piece of SAS and you've you've accumulated so much technology and and software that that is your mode it was always something else but I don't know how are how are you thinking about entrepreneurship outside of trading outside of uh the startup

before I wanted to ask how you think market cycles are evolving Uh we were joking in Q4 of last year that uh we were like great like the Bears win like the bubble the bubble did pop right. It was it did feel like it deflated. You had a big sell-off

and then you got a new paradigm that

that came you know it came at the right moment uh and kind of re reignited a lot of you know um excitement and you know revenue and and all the all the good things that you want to see during a technology cycle. Um but it feels like you know everything's been massively sped up. Um, but I'm wondering uh I'm wondering how how your what your kind of mental model is for market cycles today.

Well, the interplay of of AI and the market itself is kind of accelerated. So, the way that people interact with markets has changed. So, they're now very commonly pasting a transcript or pasting a company's earnings release into uh an AI system. And those AI systems have biases, right, that are quite separate from the biases that humans have. So, I'll give you an example. Like the the company ASML um is extremely complex. Like as a human listener to the earnings call, you would be asleep halfway through the earnings call. If you paste that into Claude Opus 4.8, it loves it. It's like, wow, this is like the most interesting cutting edge. Like, I don't even believe this is true. Let me web search that. you know, Claude is highly engaged with ASML in a way that a human kind of wouldn't be. And the really esoteric bottleneck

Well, there are there are humans that would be. There's just like a very small number of them in the world. And now there's effectively infinite, right?

There are. And and so some of these stocks that are that are pumping like HPE, you know, like I used to trade this thing back when I was at a hedge fund is the most boring company in the world. and and this guy David Gills who lives down the street from me, you know, it's like we were trading meme stocks and now he's trading HPE, right? And so it it shows you, you know, this is like a buyback like 70-year-old guys like, you know, not nothing against wearing suits or anything, but you know, it it wasn't an exciting story. Now it's like, okay, now it has tripledigit, you know, data center growth and um it's it's the things that are going up are kind of um historically stocks that retail would have never thought of. And a lot of them like SanDisk and like Micron, you know, I I know one woman who's like, you know, I inherited uh an account from my my grandparents that had a ton of Intel, Micron, and SanDisk in. I'm like, yeah, it's like literally like, you know, 75year-old people were owning these stocks because they thought they were like reliable bellweathers and these are the things that are up, you know, 300%. So, so I think I think just like the market composition is is really different. Um, you know, if you look at Coinbase, it's like got substantial sequential revenue deceleration. So, you know, people are not investing in Salana, they're they're investing in HPE.

Um, and so, you know, the the sort of B2B uh focus of the markets has completely changed. And I think that's going to continue to be the case. Like, I think I think people are going to continue to engage with extremely complex equity stories in like a way that they just never would have before. Um, and I think that benefits like a lot of, you know, for example, in my industry, I think it is going to benefit coins with more complex technology stories. I I think Zcash is a good example. Um, where it's like, okay, like their encryption technologies that are moving forward that are esoteric, that are hard to understand, that wouldn't have generated engagement before, but now like the UX of the of the user interacting with is like, I just copy pasted this into Claude and it says it's the best.

Sure. Sure. Uh, how does lore play into those stories? I feel like I heard I don't know if it's Zcash, but there was one about like they brought eight people together and they all had different pieces of the key and they were all like they brought a journalist with and there were different computers and they broke the computers and it was like building all this lore. Obviously, the Satoshi story is like extremely lore fililled. Even the Vitalic story is like you know build something you want like and it feels like the like starting the perpetual motion machine of some of these projects is important. How does that change? you know, I I want to say lore becomes uh less important um just because the the stories that I just shared with you, you know, like HPE or some Intel, like nobody knows who Intel CEO is, right? It's like and the it's like nobody is like invested in this story, right?

Are you saying they still think Pat Gellzinger's running the place or what?

I I think people just don't know, right? I think that's the the sort of real story is that Micron like I don't think people know what Micron does. It's like a trillion dollar, you know, memory bottleneck company. And

um

uh so I think lore becomes less important in one way. The the more contrarian view that I have is that IP universes become a lot more valuable.

Um and so for example, uh you know, one of the there's a lot of AI stuff about Harry Potter distills, right? So you you can like extract 98% of Harry Potter out of Deep Seek or Quad if you're if you're motivated. And so it's like it wasn't supposed to train on it. There's the big libgen settlement with Anthropic like they did train on it and there's a big multi-billion dollar payout.

But it's like that kind of uh multi-billion dollar payout creates an a strong incentive to kind of have IP treatment.

Sure.

Um in a proper way going forward. And so, you know, Games Workshop, for example, has, you know, the cinematic universe, which is World of Warcraft that becomes infinitely generative when you combine it with an AI system.

And actually, like what you get from like AMD or where where this is, I think, interesting is that you've seen like a big structural crowding out of entertainment because of the enterprise focus of AI. So, Anthropic, for example, like if you listen to Dario, he's like, "Yeah, I'm not going to ever generate videos or images because like that's just a waste of time, right?" like he's not or you know OpenAI shut down Sora

we had we had Mikey we had Mikey from Suno on today announcing a big round but he has had the blessing of a lot of his wouldbe you know the the the most intense competitors he could have have had to just dep prioritize entertainment because it might be a 20 billion market instead of a you know$ two trillion dollar market.

Yeah. And and that's sort of like the interesting Givvon's paradox argument in my opinion is that like

uh right now for example Nintendo is down you know 40% over the last year right or or you know take 2 is like getting even though there you know we've got a new GTA game it's it's getting smoked and there was a narrative that like for a while that you know video games would be a big beneficiary of AI because intuitively we see all these like like wow like what if all NPCs were you know LLMs or you know what if you had in-game generative content and the reality reality is given the cost structures right now and the bottlenecks, you're like, "Okay, um that's just not affordable, right? You can't generate real time generative characters and expect to break even on that because that would be like, you know, $6 per game session and that's not a good investment."

And as AI gets cheaper, that that story is actually going to change, right? So, so I think that's like an interesting kind of like like where we're going is I actually think we're going to get like a big acceleration in entertainment stuff because the cost of AI stuff is going to go down.

Yeah, you'd have to imagine that uh like the the the the most like mildly bullish take on Take 2 would just be like, yeah, the next GTA game is going to take seven years instead of 10 or like our DLC is going to be pulled forward by three months and that's going to accelerate earnings. like that seems very that seems very believable but uh yeah people have been caught up in all sorts of oh maybe it'll just generate it all maybe it'll all go to uh some foundation model but that does seem further away uh even when we see you know amazing demos from Google but again it's like six seconds here and there very expensive not really persistent lot

how was that how was Roblox done over the last year because that was always my thesis there was just it's going to be a lot easier for somebody to figure out how to make compelling games, but uh

down 50% in the last 6 months. Not good.

Yeah, the these things are are very very bleak.

Very bleak.

Shouldn't be. Anyway, hopefully they build back. I'm I'm optimistic. I like Roblox. Uh well, thank you so much for coming on the show.

Great. Great to finally have you on. It was great. Let's do it. Let's do it again. Let's do it again soon. See you. Have a good one. Cheers. Goodbye.

Uh there's one more announcement we got to share. And NASCAR are teaming up to sell a $14,000 VR racing rig simulator. Uh Palmer Lucky's back in the VR business. Uh I mean, he already was with the Eagle Eye headset, but he's back in the consumer entertainment VR industry with this. Of course, he's not making the actual headset. I believe it is a

really three days after I just put down a deposit on my own sim.

Really?

Yeah.

Did you actually?

Yeah.

No way. Uh but but I I don't think that's going to be

that's going to be very specific I imagine.

Yeah.

For I imagine NASCAR simulation, right? Because that's that's where they're focused. But

very

um very very cool. Matthew Prince over at Cloudflare says, "Well, that happened faster than I predicted. thought it would be end of 2027, then early 2027, but a Gentic traffic growing so fast that bots have now passed human traffic online for the first time in the internet's history.

That's crazy. Lots of agents running around on the internet.

Um,

yeah, bots are now, according to Cloudflare, 57 and a half%.

Every time you fire something off, it's it's hundreds of pages. You see it in the reasoning traces and the tool calls. Uh, let's watch this demo from Eric Lyman, friend of the show, sponsor of the show. Introducing Stack, the AI operating system that lets accounting firms take on more clients without hiring, learns your firm's process, runs the close, posts the journals, fully audit all auditable. Uh, we're living through the biggest shift in accounting since the spreadsheet. That's a good way to frame it. And they did a very cool uh video for this. All practically produced. I believe this is all shot with an actual camera. They got these actual pieces. They filmed this. Not CGI and accuracy are not negotiable. These one-off experiments.

Such a common voice.

I love Eric doing the VO stack.

It's great.

One secure place to orchestrate AI co-workers for accounting from day one.

This has been the vision of RAMP since they launched. And it and it feels like, oh, this is the moment. Of course, they have to do an AI thing, but they've been like like pitching this exact thing uh since what 2020 2019

working on it longer honestly.

We go back to parab and it learns over time always giving you the final say before anything gets posted and every action it takes is fully recorded and auditable. The more you build the lighter the work gets and the more clients you can take on. Stack

ram stack. They're saying it's god mode. It's god mode for accountants.

God mode for the

It's literally god mode. Uh well, go check it out. Uh we should also watch this one last video from Martin Scorsesei, Black Forest Labs. I alluded to it earlier. I saw this on Instagram and I really enjoyed it. 30 seconds uh of Martin Scorsesei, storied filmmaker. Jordy, name your favorite Martin Scorsesei movie.

Doesn't feel modern. A town, not a village, not a city. Medieval

Mafia. The movie Mafia. So,

isn't there Isn't it Godfather?

There is a movie.

Isn't it Godfather? The Godfather? No.

Uh, you're thinking The Irishman, maybe that is like a mafia movie.

Uh, Gangs of New York

or maybe Casino or Good Fellas.

Good Fellas.

Good Fellas.

Actually, storyboard.

Play this. Sorry, I was not paying attention.

Let's try it and we'll go from from what you think. I need a place that doesn't feel modern. A town, not a village, not a city, almost medieval. Even the the streets are narrower, cobblestone. The main road through the town is twisting and turning. Put the camera higher looking down. Deill would have his production designers do oil paintings. This is that in a sense conveys a cinematic a cinematic intelligence.

Cinematic intelligence is a good tagline for Black Forest Labs. I thought that was I thought it was really really good. Uh and yeah, what a what what a way to like, you know, hammer obviously in film making's deeply controversial, but you get Martin Scorsese talking about it and uh he's at least going to

perk up people's ears and they're going to listen to what he has to say. Yeah.

And think about it. Is it a tool? Could it be useful in a workflow? Could it speed something up? Is it going to make the next Martin Scorsese movie? Probably not this year. But will he potentially be using it when he's thinking about what to work on next? Sure. Uh so fun fun project and very cool video from Black Forest Labs. Anyway, tomorrow we have a special show, but we will still see you at 11:00 a.m. Pacific.

Sharp

sharp. And leave us five stars on Apple Podcast and Spotify. Sign up for our newsletter, tbp.com, and we will see you tomorrow.

We love you. Have a wonder evening. Flashbang.