Bittensor Drama! TAO down 15%! | E2274

Summary

This Friday TWiST opens with a Bittensor crisis: Sam Dare and Covenant AI — the team behind the celebrated decentralized-trained Covenant-72B model and three of the network’s marquee subnets (3, 39, and 81) — announced an abrupt departure, accusing co-founder Jacob Steves (a.k.a. Const) of suspending emissions, deprecating subnet infrastructure, and applying economic pressure via token sales. TAO dropped from roughly $335 to $271 (about 19%) on the news. Jason Calacanis frames it as a governance and rug-pull adjacency problem: founders should be free to leave a subnet, but not with subscribers’ tokens. To get the community side, they bring on Gareth Howells, co-founder of Vidaio (subnet 85), who frames it as an “individual acting” rather than a network or community failure and walks through Vidaio’s video upscaling and compression product — 60% smaller files at the same perceptual quality, miners from Vietnam and around the world competing for compute jobs, and the math behind making subnet emissions subsidize a new kind of “256 ML engineers competing daily” labor market.

The middle of the episode is dedicated to Ole Lehmann (a.k.a. itsOleLehmann), the AI Solopreneur newsletter writer, who took Andrej Karpathy’s “LLM council” post and turned it into a Claude skill: a contrarian, expansionist, first-principles thinker, executor, and outsider all answer your question, then peer-review each other’s responses anonymously, and a chairman synthesizes the result. Jason demo-tests it on a real founder question (how much equity to give a VP of Engineering with five years and a Google background at a $3M-raised seed), and the council recommends 0.75-1.0% with a board-approved option pool first — a level of detail Jason says he couldn’t get from his own board. Off the back of the demo, Jason puts a $1,000 bounty on the OpenClaude skill that produces the best enhanced TWIST show notes, then hires Ole on the spot for venture capital work.

Off duty: Alex hates Microsoft Copilot product placement on ABC’s High Potential (“a television commercial inside of a television show”), Jason raves about Disney’s new Maul animated series and explains why Lucas’s original 7/8/9 Darth Maul-led sequels would have been better than the actual ones, both hosts unbox the new MacBook Pro M5 14-inch (48GB RAM as the sweet spot for running local AI) versus the $600 plastic-feeling MacBook Neo aimed at the Chromebook market, and Jason riffs on Pacific Heights’ camera density (“Pak Whites”), why hi-fi music dies over Bluetooth, and why bone-conducting Shokz OpenRun headphones are the right answer when you’re walking in New York or pushing a stroller.

Highlights

”Some type of a defection from the subnets of Bittensor”

“I got a bunch of pings on my phone last night. Oh my god, TAO’s going, there was some kind of rug pull. So let’s get into it.” — Jason Calacanis, 1:01

yt-dlp --download-sections "*1:01-2:18" "https://www.youtube.com/watch?v=BaGVZI-sMK0" --force-keyframes-at-cuts --merge-output-format mp4 -o "twist-bittensor-rugpull-tee-up.mp4"

”An individual acting rather than the company or the community”

“I think this is an individual acting rather than the company or the community. I think that the subnets that Sam was heading up are probably going to carry on doing the same work and that they’re going to build upon that.” — Gareth Howells (Vidaio Subnet 85), 8:30

yt-dlp --download-sections "*8:30-9:30" "https://www.youtube.com/watch?v=BaGVZI-sMK0" --force-keyframes-at-cuts --merge-output-format mp4 -o "twist-gareth-individual-not-network.mp4"

”60% smaller file at the same perceptual visual quality”

“We can get say conservatively 60% smaller file at the same perceptual visual quality. So the user has no — like there’s no degradation for the user. They get a perfect quality file, but it’s 60% smaller. So all of the costs along that chain, yeah, they reduce.” — Gareth Howells (Vidaio Subnet 85), 22:21

yt-dlp --download-sections "*22:21-23:30" "https://www.youtube.com/watch?v=BaGVZI-sMK0" --force-keyframes-at-cuts --merge-output-format mp4 -o "vidaio-60pct-compression.mp4"

Permissionless global ML talent

“There’s incredible tech talent around the world. They might have a hard time getting a job at Google or Coinbase or whatever it happens to be, but they can permissionlessly join a TAO subnet and immediately start making money and put their tech skills to work, and they can do it anonymously, and they don’t have to worry about visas, nor do they have to worry about payments because it’s on a distributed crypto project where it’s permissionless.” — Jason Calacanis, 27:05

yt-dlp --download-sections "*27:05-28:15" "https://www.youtube.com/watch?v=BaGVZI-sMK0" --force-keyframes-at-cuts --merge-output-format mp4 -o "twist-permissionless-global-ml.mp4"

TAO as an ETF on decentralized intelligence

“Honestly, I don’t think right now I would have confidence to say, Jason, here are the top three subnets. These are the ones that are going to win. The other 125 or whatever, but I don’t mind having kind of like a, effectively a call option on decentralized global intelligence.” — Alex Wilhelm, 35:53

yt-dlp --download-sections "*35:53-36:30" "https://www.youtube.com/watch?v=BaGVZI-sMK0" --force-keyframes-at-cuts --merge-output-format mp4 -o "twist-tao-etf-thesis.mp4"

Karpathy’s LLM Council, ported to Claude

“I saw Karpathy’s post… he posted about his LLM council, which basically has the idea of you have a question, you send it to different models. Then the answers get anonymized and they get peer-reviewed. So every single model reviews all of the outputs in an anonymous way. And then they get to the chairman, that’s like another model at the end.” — Ole Lehmann, 40:07

yt-dlp --download-sections "*40:07-41:30" "https://www.youtube.com/watch?v=BaGVZI-sMK0" --force-keyframes-at-cuts --merge-output-format mp4 -o "ole-karpathy-council.mp4"

$1,000 bounty for enhanced show notes

“If somebody wants to make a skill for OpenClaude to do this, I will put out a $1,000 bounty and on let’s say May 1st I’ll give the $1,000 bounty for the open source project that proves they can do incredible enhanced show notes for me for TWiST.” — Jason Calacanis, 48:46

yt-dlp --download-sections "*48:46-49:45" "https://www.youtube.com/watch?v=BaGVZI-sMK0" --force-keyframes-at-cuts --merge-output-format mp4 -o "twist-1000-bounty-show-notes.mp4"

Key Points

• Sam Dare / Covenant AI exits Bittensor (1:01) - The team behind subnets 3, 39, and 81 publicly leaves the network the day before, accusing co-founder Jacob Steves of suspending emissions, deprecating subnet infrastructure, and dumping tokens
• Covenant-72B was the network’s marquee proof point (2:18) - 72-billion parameter model trained decentrally on Templar; that hype is what drove TAO higher in March
• TAO drops ~$335 -> ~$271 (6:18) - Roughly 19% in a session, market cap ~$2.93B; Alex frames it as “not the end of TAO” but a governance shock
• Subnet governance fix: stake collateral, lock founders’ tokens (5:22) - Jason’s prescription: subnets should have to stake TAO, like buying a franchise, with limits on what one operator can withdraw
• “You can leave; you can’t leave with everybody’s tokens” (7:23) - Jason’s adjudication standard for rug-pull-adjacent behavior
• Gareth Howells: it’s an individual, not the network (8:30) - The Vidaio co-founder argues the subnets Covenant ran will continue under different leadership
• Vidaio (subnet 85) compresses video 60% with no perceptual loss (22:21) - Plus upscaling and lip-sync; reduces streaming, storage, and CDN costs end to end
• Upscaling/compression TAM: $175M -> $1.1B by 2032 (23:04) - Competitors: Topaz on the desktop, Amazon’s offering in cloud
• 256 ML engineers competing daily (24:11) - Subnet emissions subsidize miner pay; “winner-takes-all” model lets Vidaio also collect open-sourced model weights from miners
• Vietnamese miner team (26:48) - 3-4 university-graduate engineers from Vietnam, never met face to face, doing top-tier work for TAO emissions
• Permissionless global ML labor market (27:05) - Jason’s thesis: TAO subnets give Google-quality talent in emerging markets a way in without visas, KYC, or payment rails
• Capitalism backstop on miners (34:21) - If buyers underprice, miners simply walk and take Korean/Japanese contract work; the system corrects on its own
• TAO as an ETF on decentralized intelligence (35:53) - Alex’s frame for retail investors who can’t pick winning subnets
• Council of Advisors skill is one model, five personas (41:06) - Claude Opus 4.6 spins up parallel sub-agents (contrarian, expansionist, first-principles, executor, outsider), then anonymizes and peer-reviews
• What did all five miss (45:12) - Ole’s own twist on Karpathy: a chairman question that surfaces shared blind spots across all advisor outputs
• Council nails option-pool advice (52:00) - “Create the option pool before this conversation. A 10-15% option pool documented and board approved” — the punch-up Jason says his own board wouldn’t catch
• $1,000 bounty for enhanced show-notes skill (48:46) - Demo or die: Jason will pay May 1st for the best open-source OpenClaude skill that produces deep linked TWIST show notes
• Cyprus 12.5% tax vs Berlin 50% (56:51) - Ole moved his AI Solopreneur business to Cyprus; Jason’s broader rant: Americans abroad still owe federal, only state taxes drop
• Microsoft Copilot crime-show product placement (1:00:14) - Alex flags ABC’s High Potential running what amounts to a Copilot infomercial mid-episode; Jason calls it “literally being bashed in the skull”
• Maul’s animation style is watercolor-into-engine (1:04:13) - Disney’s new Darth Maul series; Jason argues 7/8/9 should have been Maul, not Rey, per Lucas’s original treatment
• MacBook Pro M5 14-inch with 48GB is the new floor (1:08:02) - $3,500; Alex says the M5 (not Max) plus 48GB unified memory is what unlocks running OpenClaude and local SLMs
• MacBook Neo at $600 (1:09:34) - The first Apple device that “feels cheap” to type on; Jason argues the play is education/iCloud lock-in, not hardware margin
• “Pak Whites” (1:19:50) - Jason’s running gag about Pacific Heights’ camera density and 30-second police response

Mentions

Companies

• Bittensor / TAO (1:01) - The decentralized AI network at the center of the episode; ~$2.93B market cap after the drop
• Covenant AI (3:00) - Sam Dare’s project running subnets 3, 39, and 81; the entity that exited
• Templar (2:18) - Decentralized AI training subnet that Sam Dare came on TWIST to discuss in March
• Vidaio (Subnet 85) (8:11) - Gareth Howells’s video compression and upscaling subnet; vidaio.io
• Tao Stats / MOG (Hippias) (14:18) - Vidaio’s incubator on Bittensor
• Topaz (18:27) - Desktop video upscaling software referenced as a competitor
• Amazon (23:04) - Has its own video upscaling cloud offering
• AI Solopreneur (Beehiiv) (55:30) - Ole Lehmann’s AI media and B2B GTM consulting business; aisolo.beehiiv.com
• Anthropic / Claude / Claude Code / Claude Co-work / OpenClaude (42:32) - The runtime for Ole’s Council of Advisors skill; Claude Opus 4.6 specifically
• Substack vs Beehiiv (56:29) - Alex declares Substack “so 2024”
• Microsoft Copilot / ABC High Potential (1:00:14) - The product-placement skewered in off duty
• Apple (MacBook Pro M5, MacBook Neo, Mac Studio, Apple Watch, AirPods, AirPods Pro 2) (1:08:02) - Hardware ladder for local AI
• Disney / Lucasfilm (Maul, Andor, Rogue One, Mandalorian, Ahsoka, Clone Wars, Bad Batch, Rebels) (1:03:43) - Star Wars canon discussion
• Pono (Neil Young) (57:51) - Hi-fi streaming service Neil Young founded out of MP3 frustration
• Wirecutter (NYT) (1:17:55) - Jason tried to acquire it; New York Times got it
• Plaud (0:13) - Show sponsor; NotePin S; plaud.ai/twist + code TWIST
• Northwest Registered Agent (9:06) - Sponsor for forming a Delaware C-corp
• Pilot (28:54) - Sponsor; $1,200 off year one of CFO-grade startup accounting
• Squarespace (19:21) - Sponsor; Blueprint AI builder
• Ro.co (58:58) - Sponsor; GLP-1 insurance check at ro.co/twist

Products & Technologies

• Subnet emissions / TAO (6:00) - The mechanism Sam allegedly weaponized when leaving
• Covenant-72B (2:18) - 72B-parameter LLM trained decentrally
• Council of Advisors skill (42:32) - Ole’s open-source Claude skill, on his X profile and GitHub
• Andrej Karpathy’s LLM Council (40:07) - The original blog post; one model for chairman, multiple models as advisors
• Plaud NotePin S (13:53) - The wearable conversation recorder; auto-syncs and charges in dock
• MacBook Pro M5 14-inch / 48GB RAM (1:08:02) - Alex’s $3,500 setup
• MacBook Neo (pink, $600) (1:09:34) - 2.7lb cheap-feeling Apple laptop aimed at Chromebook market
• Shokz OpenRun bone-conducting headphones (1:17:42) - Open-ear safe alternative to AirPods for city walking and stroller pushing
• Hyperion (Dan Simmons) (1:16:00) - Sci-fi novel Alex pushes on Jason; movie adaptation in development
• These Burning Stars / Kingdom Trilogy (Bethany Jacobs) (1:16:00) - Sci-fi audiobook Alex listens to while pushing strollers
• Designer’s Guide to Creating Charts and Diagrams (Nigel Holmes, ~1983) (1:13:03) - The Businessweek infographics seminal work Jason found at a thrift bookstore
• Ogilvy on Advertising (David Ogilvy); My Life in Advertising and Scientific Advertising (Claude Hopkins) (1:14:53) - Other seminal advertising books in Jason’s coffee-table stack
• Democracy in America (Tocqueville) (1:15:39) - Alex reads it with his dad; argues you should write in your books
• Mythos (Anthropic preview) (54:30) - Alex’s “Hiroshima for software” question to Ole; sets up the next episode

People

• Sam Dare (2:18) - Covenant AI / Templar founder accused of selling subnet tokens on the way out
• Jacob Steves (Const) (3:00) - Bittensor co-founder; the actor on the other side of Sam’s accusations
• Akos (10:09) - Issued a community response on the Covenant exit
• Gareth Howells (8:11) - Vidaio (Subnet 85) co-founder; @GarethHowells on X
• Hippias (14:18) - Past TWIST guest; built inside Tao Stats MOG like Vidaio
• Andrej Karpathy (40:07) - OpenAI co-founder, ex-Tesla AI head; LLM council post
• Ole Lehmann (39:33) - AI Solopreneur; Cyprus-based via Berlin
• Steve Jobs / Steve Wozniak / Homebrew Computer Club (Menlo Park, 1975) (50:30) - Jason wants to do an “acquired-style” oral history this summer
• George Lucas / Kathleen Kennedy / J.J. Abrams (1:04:58) - Jason’s grievance with the Star Wars sequels
• Carrie Fisher (1:06:22) - Why Star Wars sequels never recovered
• Sam Parr (Hustle / My First Million) (1:12:55) - Recent guest who put Jason onto seminal-works hunting
• Nigel Holmes (1:13:03) - “Godfather of design”; Businessweek infographics pioneer
• David Ogilvy / Claude Hopkins (1:14:53) - Old advertising masters Jason scribbles in
• Dan Simmons (1:16:00) - Hyperion author
• Bethany Jacobs (1:17:10) - These Burning Stars / Kingdom Trilogy
• Neil Young (57:51) - Founded Pono streaming over MP3 quality complaints
• Dario Amodei (54:30) - Anthropic CEO; subject of Ole’s “Hiroshima for software” tweet about Mythos
• CEO of Zoom (Eric Yuan) (1:20:18) - Alex met him at a “fancier part of Pacific Heights” VC dinner

Surprising Quotes

“It would be like investing in a startup and then the startup just says, ‘You know, I don’t like the way capitalism works. I’m out,’ and then they take the money.” — Jason Calacanis on the Covenant exit, 7:23

“We’re essentially paying 256 machine learning engineers to outdo each other on a daily basis. And that’s an unbelievable powerful tool to have behind you.” — Gareth Howells (Vidaio Subnet 85), 24:11

“If everybody in the world could work at your local Starbucks or anybody in the world could magically sell coffee beans to your local coffee shop and they basically had a competition, you would have a barista for $1 a day, $2 a day living like a king or queen in their native location.” — Jason Calacanis, 31:01

“Stop ruminating, don’t go to therapy, don’t overthink things folks, just go for it and build. It’s all that matters is building and making your venture capitalists more money faster.” — Jason Calacanis on R-word maxing, 45:00

“I just think it’s a point in time… it’s one of these points in time where it just makes sense to take a step back and take your security seriously, because the downside if you don’t is just so immense.” — Ole Lehmann on calling Mythos “Hiroshima for software”, 54:47

Transcript

Jason Calacanis: 0:00 All right everybody, it’s Friday, it’s Twist, April 10th. Alex is here. We’ve got a great show, lots of important things going on in the Bittensor world. There was a rug pull, we’ll get into that. Obviously tons of AI stuff going on. But just as we kick off the show, Alex, I always like to applaud-plaud, I put my Plaud pin on, I press my button, I get a nice little haptic, and I’m recording. Now I have every detail so I can say, hey, after the show, make sure you don’t forget to go over the newsletter stuff for the this week in AI newsletter, boom, now it’s an action item, I won’t forget it. And I do that all day long and Plaud is an amazing service, thank you to them for supporting the show.

Alex Wilhelm: 0:41 Yeah, and also, if you want to get a Plaud for yourself, you can at a tasty discount if you go to plaud.ai/twist, P-L-A-U-D.ai/twist, use the code twist, T-W-I-S-T, save 10%. Be cool like Jason and I, don’t forget things, get more work done, less pain, Plaud.

Jason Calacanis: 0:58 Use that code twist for 10% off and they know we sent you. All right, let’s get to work here. There was some type of a defection from the subnets of Bittensor. People have heard me talking about this very promising crypto project called Bittensor, the token in it is called TAO. There are subnets. What are the subnets doing? You may have heard us talk about it here or on All-In, or you see it on your X feed. Some of, you can create subnets, there’s 128 of them available, and those subnets will do different projects. One of them is a co-working or a co-pilot for doing coding like Codex or cloak code. And why are these important? Why should they be created using cryptocurrency? Well, it’s a distributed network. And just like Bitcoin, you can have people compete to provide that service at the lowest possible price, which is super deflationary. And my thesis, and I’ve got an investment in this space, I’ve got a couple different investments in the subnets and I’ve bought TAO directly and I might even be a little underwater now based on where I bought it. Doesn’t matter to me, but I got a bunch of pings on my phone last night. Oh my god, TAO’s going, there was some kind of rug pull. So let’s get into it.

Alex Wilhelm: 2:18 We had Sam Dare on the show, Twist, it was March 27th, 2026, so just about two weeks ago, Jason. And he was talking about Templar, a project that was doing decentralized training of an AI model, of an AI model. This was Covenant AI, three different subnets, subnets 3, 39, and 81. They formed kind of a triumvirate, if you will, of things that work together, pre-training, compute, and post-training. And what they built was the Covenant 72 billion parameter model. In March, this got a lot of attention. Look, decentralized training of an AI model, not just the work of the major AI labs. Instead, something that, you know, maybe anyone can participate in. It drove the price of TAO up. And then, Jason, yesterday, we got a major statement from Sam and the group at… Covenant AI, which is right here, and it makes a number of very important claims. It says that Bittensor is not as decentralized as it claims. It also says that one of the co-founders of the project has been essentially blocking their ability to do their work. Really, really big claims. There’s been a response to it, but what really matters is scrolling, scrolling, Jacob Steves, also known as Const, one of the co-founders of Bittensor, has taken a series of actions against Covenant AI’s operations. These include suspension of emissions to our subnets, removal of moderation capabilities, unilateral deprecation of our subnet infra, and economic pressure via token sales. He responded to Jason, but I want to get your take first on basically what Covenant is claiming.

Jason Calacanis: 3:43 There’ll have to be a little bit of an investigation here as to what’s happening, but the control of the subnets and making sure there aren’t rug pulls like we see all the time with, or we saw all the time with ICOs or meme coins. This is a major vector of attack and a major concern people have in the crypto community. So what Tao is trying to do is remove that possibility. So what system do you create to remove that possibility? Well, you have to have some control over the subnets and there has to be governance. So that governance stuff is being worked out from what I understand from the insiders and sometimes you’ll have a bad actor. I’m not saying this person is a bad actor, but some people are claiming, hey, this person just ran off with a bunch of the subnets’ tokens and some Tao and it just is something that happens in crypto sometimes as people build a project, do you buy some tokens, it’s not regulated, and all of a sudden you get rug pulled. That still exists. Sometimes you’ll have lawsuits come because of these things and attorney general in Florida or somewhere will find who was impacted by a rug pull. You remember all the celebrities who were promoting different coins? They all got tagged. So this is the kind of stuff that can happen and this will be in my estimation, and this is my hope and what I believe will happen is that this will help solidify how to handle and avoid these things in the future.

Lon Harris: 5:14 This is a trip and a fall over something they’re going to remove from the pathway so in the future no stumbles of this sort will crop up.

Jason Calacanis: 5:22 Yeah, I think that they can, you know, eliminate this from happening in the future by making it such that the subnets have to stake a certain amount of Tao, and that the people who are participating have some control over the subnet. And subnet governance is going to be a very delicate thing because you want entrepreneurs to be able to start a subnet, put up some collateral for it, almost like buying a franchise, and then you want those folks to believe that they have some level of ownership, but not so much ownership that they can take a bunch of people’s tokens and then just say, you know what, I don’t agree with how this is being run, I’m just going to take all these tokens and leave. So, yeah. You know, I’ve… as I’ve said to everybody, the opportunity in these kind of projects is that these questions have not been answered. This… they are imperfect. That’s why it’s at a whatever it is, you know, 2 billion dollar market cap, not a 500 billion or trillion dollar market cap.

Alex Wilhelm: 6:18 2.93 after the declines. A good segue, Jason. TAO is trading at roughly 335 before this news broke. As we talk to you, it’s worth 271. So, a pretty sharp decline, but certainly not the end of TAO. This is not, you know, the project falling apart. But a pretty large drop in value. The thing that I’m trying to sort out in BitTensor-land, Jason, because I’ve talked to probably 10 different subnets in the last couple of weeks just prepping them for the show and getting to know what they’re working on and how the economics work, is there does seem to be at times moments when they say, ‘Oh, the community wouldn’t tolerate that type of bad action.’ And so often I feel like I’ve run into… there’s some humans in this system that is not yet fully decentralized. And I wonder if this argument is essentially coming to the point to which that’s no longer acceptable and it has to become even more decentralized to maintain trust. Because even though a lot of people in the BitTensor community are I think pretty negative about Sam right now—and we’ll bring on a subnet guy in just a second—I don’t… I’m struggling to get that mad at him for what he did, because it seems like he made something cool and then wants to go host it somewhere else. And you and I don’t throw a fit when someone leaves Substack.

Jason Calacanis: 7:23 Yeah, it would be though if you took all the subscribers’ money and then didn’t provide the subscription product. So that’s what has to be adjudicated here, is how much TAO he collected from people, if they invested in his company. It would be like investing in a startup and then the startup just says, ‘You know, I don’t like the way capitalism works. I’m out,’ and then they take the money. So these would be the allegations that are all going to be worked out over time. You and I can speculate about it, but yeah, you should be able to leave your subnet and go do something independent if you want, of course. But you don’t get to leave with everybody’s tokens. And then there might also be like some dumping going on here. So again, all allegations, all to be worked out in the court of crypto adjudication.

Alex Wilhelm: 8:11 But let’s get some of the participants. Let’s bring on Gareth Howells. He’s from video subnet 85. And we’re going to talk to Gareth about what he’s building and why it’s cool and how he’s using BitTensor in a second. But I wanted to get his take on the situation. Gareth, hey, you’ve been tuning in. Quickly, who’s closer to the truth, Jason or I, and did the people behind Covenant AI make off with the tokens?

Lon Harris: 8:30 I think Jason has… is very… got a very good understanding of the situation. Yeah, absolutely. I think this is an individual acting rather than the company or the community. I think that the subnets that Sam was heading up are probably going to carry on doing the same work and that they’re going to build upon that. So, yeah, I think Jason’s synopsis of what happened was was pretty close.

Jason Calacanis: 8:53 Can you give me your synopsis? Try… because it’s always good when these new technologies come out to have multiple people try to… …try to describe and use analogies or, you know, their own handicapping of the situation. It’s easier than ever to build a great new product, and launching a startup even as a solopreneur is getting easier and easier. But, as my experienced founders already know, there’s a lot more to starting a company than just putting up a website, or even building a product. If you’re serious about going into business, you need a Delaware C-corp. That’s going to give you a serious leg up on your competition, and you’ll be taken seriously by investors. That’s where Northwest Registered Agent comes in. Just a few clicks, they’re going to give you the perfect start to your new enterprise: a real identity, a domain, a custom website, a business email, all the public filings done, and even a phone number. And you’ll complete this process in under 10 minutes. Plus, you’re going to get all sorts of free tools and resources, from step-by-step guides, compliance reminders, and an online account that’s going to keep everything organized even as your business grows. So get all the advantages of a Delaware C-corp regardless of where in the US you’re operating out of. northwestregisteredagent.com/twist.

Lon Harris: 10:09 Yeah, Sam was a subnet owner, you know, he had access to the wallets and he sold the tokens. The reasons that he gave for selling the tokens, I’m not 100% convinced by. I think Akos has made a response, saying actually some of the accusations he wasn’t able to do and they weren’t true, and it was impossible for him to like take some of the actions that Sam’s accused him of. So I think possibly he may be pointing the blame somewhere just to excuse his actions, but you know, we don’t know everything about it. This is just from obviously what I know anecdotally like you guys, I’m not right in the nitty-gritty of the exact situation.

Jason Calacanis: 10:47 And why did Tao go down 20% or whatever it went down, you know, after when this happened? Was it just bad vibes and people were like, ‘Oh my god, this is an attack vector, this is a weakness of the BitTensor system’, or was it because the market was flooded with Taos being sold and there weren’t as high of a bid to absorb them?

Lon Harris: 11:07 Yeah, I think there was a big element of fear there. I think obviously Covenant has been absolutely killing it recently with the retail investors, and it was right on the top of the emissions and it has been for a long time. It was getting amazing press outside from people who didn’t have a full understanding of BitTensor, but they obviously were understanding what he’s doing and very impressed by that. And I think that when it was found out that he was selling all his Tao and taking it, I think there was a big fear amongst investors, not necessarily the community, but investors were very worried that this was going to be catastrophic for the Tao price, and a lot of people got out. It wasn’t just him selling his Tao, but that was an element of it.

Jason Calacanis: 11:53 Gareth, is it fair to say that Covenant AI and its three subnets were kind of a marquee or like apex project for BitTensor?

Lon Harris: 12:00 I think they were, but I think there’s also they’re only three subnets out of 128 and we’ve got some incredible subnets doing some amazing things and they’re going to come to the forefront very soon. So I really think this is more of a blip than, you know, anything saying that BitTensor is in real trouble. I think we’re going to bounce back very quickly. The community is amazing and some of these the other subnets are going to come to the forefront very quickly. So I really think this is—this is a very short-term issue, but I also think it’s going actually make the community much stronger. We’re going to put things in place to stop this happening again and I think we’re going to learn a lot from that.

Jason Calacanis: 12:41 All right, I really appreciate that. Now, let’s talk about video subnet 85. Uh, first of all, give us the TLDR about what it does and then, uh, we’re going to show a quick, an upscale of, uh, Mr. Lon Harris.

Lon Harris: 12:51 We are, we are. So yes, video subnet 85 is a video processing subnet. So essentially we’re starting out doing video compression and upscaling, but now we’re moving on to optimizing your video in any way that’s possible. So we’re—we’ve got an agent that we can connect to your storage, we read everything that’s on your Amazon S3 bucket for instance. We take a look at that and we—we tell you where we can optimize. Optimization can be many things. We can optimize for streaming on a VOD platform, we can optimize for archiving, you know, best quality picture saving you space. We can optimize for security, medical, education, you know, we’re covering all those areas of video. So, um, yeah, essentially what we’re doing is, um, using an AI—AI agents to make your video the best it can be for whatever your needs are.

Alex Wilhelm: 13:43 Just to recap, I always like to explain back to the founder so I understand the vision. Uh, Vid-AIO, V-I-D-A-I-O is how the name of the company is spelled or should I say the name of the subnet is, the Tao subnet, the BitTensor subnet. It’s numbered 85, just so people understand there’s 128 subnets. A person like Gareth can start a subnet, stake with some Tao, and they may or may not have a company behind it. So are you also running a startup that is doing this or a consulting firm?

Lon Harris: 14:18 Yes, so we’re incubated by Tao Stats MOG, who you had on the show before with Hippias. So we’re in exactly the same situation as Hippias.

Alex Wilhelm: 14:27 Gotcha. And this could become a company. It’s inside of a different company right now.

Lon Harris: 14:30 Absolutely.

Jason Calacanis: 14:31 So as investors, uh, we could invest in a parent company, or you could buy Tao, or you could buy the subnet token. Three different ways to invest in these opportunities. I’m looking at all of them. Uh, I have—I own Tao directly myself, uh, through Coinbase. It’s not on Robinhood yet. And then Stillmark Capital, uh, which I’m a partner in with Mark Jeffrey, they are going and buying subnets. To buy subnets, you gotta go to the subnet owner and do a deal to buy the— Tokens that are not trading freely on marketplaces yet, um, but when you buy Tao, it’s in some ways like buying a mutual fund or a index, an ETF, I guess would be the best way to say it, an ETF of a bunch of different startups. Okay. This particular subnet is challenging people on the internet or globally to give you compute and then you take that compute and you will take video files and render them into, you know, different, um, a- different versions of it. Now, this seems like a very narrow, uh, use case, like I can on my desktop, I can use VLC to save a file as a different file type, but you’re doing something a little more sophisticated and the customers of this product are what? So let’s just talk really tightly about who are the customers for upscaling of a video or just essentially making different versions of a video, and how big is that market?

Lon Harris: 15:56 So I mean, the market is enormous. A- 85% of internet traffic is video. So, you know, it’s not going anywhere. More video is getting created every day and it’s going into storage buckets on the cloud that are just getting bigger and bigger. Um, people’s, you know, bills are getting bigger, no one’s really doing anything about that. So to answer the question, upscaling would be anyone with a, a library, a library of old content. You could be, you know, the National Archives in the UK, you could be a big TV broadcaster who made 20 years of programming in SD, um, you could be a- a consumer who’s got video tapes-

Jason Calacanis: 16:40 Got it. So like the BBC has a giant archive, BBC News has a giant archive, um, Getty Images has an archive, CBS owns an archive. Those archives might be in very low quality video.

Lon Harris: 16:56 Yeah, exactly that. So and and they won’t translate to like a 4K screen, like our expectations these days are of high picture quality. So we can go in there with our agent, we can scan your archive, and we can tell you what we can do to that. We can make these all these videos 4K, we can turn your black and white videos into color, we can change the frame rate if if the video’s jerky, we can clean up noise. Or if you want to keep it in black and white, we can do that, you want it in sepia, we can do that also. We can do subtitling, we can do transcription, we can do tagging of metadata so then you have a searchable content library. So we can do all of those things within, you know, within our ecosystem.

Alex Wilhelm: 17:27 Got it. So the miners here are not just providing compute, Gareth. Aren’t they providing the actual models themselves that you’re using for those specific tasks?

Lon Harris: 17:35 Yeah, exactly that. So we’re now moving on to the new iteration, um, which is a- a winner-takes-all or winner-takes-most model for the miners. So essentially what they’re doing now is optimizing models. So rather than we were using their compute before, but that became very difficult to scale. So now we’re we can scale with Shutes, which is another subnet, it has a trusted execution environment.

Alex Wilhelm: 18:00 which means that it’s a safe environment that our clients will be happy with because it’s 100% safe. It’s private for us and the client. So now we can optimize models so we can say we want a new upscaling model or a colorization model or subtitling. Lip sync- we’re looking into lip sync models as well. So there’s so much more versatility now in the subnet if we move to this kind of winner-takes-all model.

Jason Calacanis: 18:27 And there are desktop pieces of software like Topaz that I think people will use for this type of product or service. So if you had an old video of a Bob Dylan concert or I have all these old Dire Straits bootleg videos, they’re all over YouTube and they’re grainy. Now AI can take them and upscale them. So about three or four years ago, I started to see this on YouTube where people would say, upscaled Dire Straits thing and it’s pretty amazing to see literally a VHS, shaky VHS turn into something that looks like, okay, maybe HD or close to HD. And it’s only going to get better from here. This is the worst it will ever be. So very fascinated- oh okay here, so let’s see an example. AI tools are making it easier than ever to run your own business, even as a solo founder, but you still need a beautiful attention-grabbing website to help your new company stand out in a very crowded field. And you don’t want AI slop, nope, you want to use Squarespace. That’s the easiest and fastest way to turn your idea into a real business because the team at Squarespace cares deeply about design and functionality. And a plain looking or generic or AI slop website, man, that’s going to be a red flag for your customers, for your investors and people who want to come work for you and join your team. But Squarespace will take all the guesswork out of designing your first website with the Blueprint AI builder, which has been finely tuned to make beautiful websites. Squarespace isn’t just going to help you make a new website either, they’re going to be your all-in-one platform for launching your business. They’re also going to help you set up your email, they’re going to handle invoicing, paperwork, all your needs. Go to squarespace.com/twist for a free trial and when you’re ready to launch, go to squarespace.com/twist for 10% off your first website or domain purchase. So here’s an example of it from the subnet. Here’s Lon, if you’re on the audio version, we have a grainy Lon on the left and…

Alex Wilhelm: 20:24 Rest in peace Lon Harris, he was an amazing collaborator here and so we’re celebrating him today. He’s in surgery as we speak, who knows if he’ll get out. He’s wearing his plaid pin, just so you know, he’s wearing his plaid pin in surgery.

Jason Calacanis: 20:34 Two plaid pins in one show. There you go.

Lon Harris: 20:37 To me it doesn’t feel like Gaph is anywhere near ready to work by himself. Like even if he was in a Mac studio, I feel like I would still need to be there giving him feedback, pointers, asking him to do things. I don’t think he’s ready to replace me, he’s ready to do like 80% of what I do.

Jason Calacanis: 20:56 The one on the left and the one on the right, explain to us the difference.

Alex Wilhelm: 21:00 So the one on the left was

Lon Harris: 21:00 The original video that we upscaled, and which was taken obviously from one of your podcasts, and we compressed that down a bit to make- to make the picture quality worse, but then I put it through the subnet and we upscaled this up to 4K. So essentially what we’re doing is, it’s a Gen AI tool which is predicting what should be in the picture and it’s making all of the the, you know, the blurry lines clearer.

Alex Wilhelm: 21:28 Got it. And so this is a way if people are recording their podcast, just as a silly example, over Zoom, like many people do today, and that kind of downgrades your video a bit, you could take All-In or This Week in Startups here and Twist, and just boom, make it a beautiful look like it was using 4K Studio Red cameras or something, yeah?

Lon Harris: 21:48 Yeah. Absolutely. We can do that and then we can also compress that video. So you get a super high picture quality, at much smaller file size. So therefore, I mean, that’s going to help you with streaming, that’s going to help your storage and your CDN costs. So it’s going to create efficiencies all the way down, you know.

Alex Wilhelm: 22:05 But how does that play into low-connectivity markets? Like, you know, in parts of the world, like parts of Latin America or parts of Africa, connectivity’s a lot worse. Does Starlink make it so that everyone has unlimited bandwidth now, we don’t need to worry about it, or are there still places where people are pretty bit-conscious?

Lon Harris: 22:21 Unfortunately not, yeah. So Africa being the perfect example. I’ve got a lot of experience working in Africa in and in emerging markets. So connectivity there’s a real problem. Also everyone uses mobile phones, there’s no laptops, there’s no 4K TVs. So everything’s needs to be compressed down to the small screen. Connectivity’s an issue, data’s an issue, phone battery’s an issue. So these- all of this, everything we’re doing to these files is solving those problems essentially. We can get say conservatively 60% smaller file at the same perceptual visual quality. So the user has no- like there’s no degradation for the user. They get a perfect quality file, but it’s 60% smaller. So all of the costs along that chain, um, yeah, they reduce.

Jason Calacanis: 23:04 Yeah. And if you look at this space, uh, it’s predicted to grow from 175 million in 2025 to 1.1 billion in 2032. So this is going to be com- this is going to be like a compounding vertical where people are really going to want to be upscaling videos on a regular basis, and there are many competitors here already from desktop software to Amazon has an offering. You, in terms of your pricing, walk us through the economics now because you’re using another subnet for compute and so you’re using your miners - miners are individuals who want to earn Tao for doing a task. So just so people think about it, that’s the supply side. So that’s the Uber drivers, that’s the Airbnb hosts. They can just sign up here and compete for these jobs. And then there’s a validator that confirms that they did the job and then picks the winner? Am I broadly correct in how subnets work and how this one works?

Lon Harris: 24:11 The thing with video is compute heavy for sure, but there’s—it’s very hard to put a price per minute on what we’re doing, which is essentially the old way of pricing, say, that’s what AWS would do. Because we now have this optimization model where we could be using three, four, or five different models on one piece of video. It goes kind of on a case-by-case basis. But what we have on our side, not only do we have the innovation of BitTensor, which is, of course, the incentive mechanism and the miners creating better and better models—so we’ve already done benchmarking with all our competitors and we’re already better, so we know that we’re good there. But also, you have the emissions from the subnet, which can also subsidize our earnings so we can pay our miners. So we’re essentially paying 256 machine learning engineers to outdo each other on a daily basis. And that’s, you know, that’s an unbelievable powerful tool to have behind you.

Alex Wilhelm: 25:12 What are the models that they’re using? These are open-source models? They’re tweaking their models? They’re the model originators? They’re making SMLs out of large language models? Who are these—what are these models they’re using?

Lon Harris: 25:23 It’s some and some. So when we launched the subnet, we always start with a base model. For upscaling, we use an open-source model. For compression, we created our own because there was none existing. The miners have taken those and they’ve built upon them, and they’ve tweaked them, and they’ve made them better, or they’ve created their own models. Sometimes we don’t even know what models they’re using, we just know that the output is better than, you know, that we can do ourselves. So with the new winner-takes-all, we will ask them to open-source their models to us, so then we will be able to use that in a private environment for our clients.

Jason Calacanis: 25:59 Have you ever met these miners and the people who are participating?

Lon Harris: 26:02 Face-to-face, no, but we’ve had some meetings. There’s a couple of pro-mining teams that are working on the subnet, which come from the video world, actually, not machine learning engineers, which is very interesting. So we have, yeah, we do talk to them and we, you know, have meetings.

Jason Calacanis: 26:22 And just describe for us who these type of people are. Are these people who work at IBM and they make a great salary and they’re doing a side hustle? Or, you know, they work at a Hollywood studio and on the weekends they do this? Or they’re just bored at work? Or they’re living in Thailand and they don’t have to make a ton of money and they’re just nomadic, entrepreneurial CTOs and in their spare time they, you know, contribute to these subnets because it’s a fun way to make a, you know, an extra couple of thousand dollars a week?

Lon Harris: 26:48 I mean, we’ve had a team from Vietnam doing exactly that. There was, I think, three or four guys working solely on this. I think they were university students or graduates, essentially.

Alex Wilhelm: 26:58 In from Vietnam? Or people who were— Like nomads.

Lon Harris: 27:01 No, no, they were Vietnamese, yeah. And they—

Alex Wilhelm: 27:03 They were Vietnamese?

Lon Harris: 27:04 Yeah, yeah, they had a really good team.

Jason Calacanis: 27:05 So that’s a very interesting thing to pause on there, Alex. Is there’s incredible tech talent around the world. They might have a hard time getting a job at Google or Coinbase or whatever it happens to be, but they can permissionlessly join a Tao subnet and immediately start making money and put their tech skills to work, and they can do it anonymously, and they don’t have to worry about visas, nor do they have to worry about payments because it’s on, you know, a distributed, you know, crypto project where it’s permissionless. Permissionless, fancy way of saying, um, I can just participate without somebody, you know, approving me. And that’s a real, you know, unique thing in all the world.

Alex Wilhelm: 27:49 This has been my biggest learning talking to the different subnets in the last few weeks is that there are so many people out there who can contribute to machine learning technology that don’t only work at Google. Cause I’ve asked this question: where are you finding these people? Where are these ML engineers that have all this spare time on their hands? Cause in typical startups, Jason, they’re desperate for local talent, everyone’s fighting over the same eight people. But you go to Vietnam, you have a bunch of smart people who can do great work, and Gareth doesn’t care.

Lon Harris: 28:16 Yeah, they come to us. That’s the beauty of it is that, like, we built it and they come to us. So, you know, they obviously have an understanding of the ecosystem and the network, but you create a new subnet, people register, they start mining, you’ve no idea who they are and you don’t really need to. You just, obviously you need to see their proof of work and you need to validate it. And that’s the only thing that matters. And that’s why, you know, you can get the best talent and maybe they are working for Google, maybe they’re doing this on the side, but sometimes we don’t know.

Jason Calacanis: 28:45 Alright, Gareth, anything else we missed? If not, we’ll drop you off and thank you for your time.

Lon Harris: 28:50 Absolute pleasure. Thanks guys.

Jason Calacanis: 28:54 There are lots of accounting firms out there that will help you maintain your books and do your taxes, but when you’re a founder, you’re facing a lot of unique challenges, so you want a partner who understands the landscape that startups operate under. You want to be able to trust them with the reins so you can focus on building your product. Well, Pilot is the largest accounting firm that was built just with startups in mind. You’re not just getting your taxes done, no, you’re getting CFO-level guidance for your startup. Pilot will help you track where all your cash is going, increasing your profitability and helping you spot looming issues before they become real problems, hey, like your runway. You gotta be on top of this stuff, folks. And when it’s time for you to raise your next round, Pilot is right there to help you with due diligence and scale up and make that process easy-peasy lemon squeezy. So, start focusing on your product and let Pilot handle the bookkeeping. Plus, Twist listeners get twelve hundred dollars—twelve hundred dollars—off in their first year. What a deal! Go to pilot.com/twist to get started. That’s pilot.com/twist. And we’ve been going through all the subnets. I believe in this project. I have— I’m very, um, you know, since the administration’s changed and the folks who were running crypto policy have changed,

Lon Harris: 30:08 True.

Jason Calacanis: 30:09 I am now officially open for business in crypto. I’m doing a stablecoin project, getting involved in Tau, I’m specifically looking though only for projects where there’s a customer who is either making money, saving money, entertained or otherwise delighted by a product or service. So I’m not looking for meme coins, I am not looking for pump and dumps, I’m looking for, you know, actual products that create value and use the technology in a way that it was intended that gives it an advantage. And really when you look at the Bittensor universe, incentives matter, tapping into a global workforce, tapping into global compute, having it be distributed and, you know, again, there’s like fancy words used in crypto like permissionless…

Alex Wilhelm: 31:00 Mhm.

Jason Calacanis: 31:01 …and, you know, or, you know, peer-to-peer or distributed. All this stuff means is anybody can participate, the compute or storage or transit can come from anybody, and that means it moves faster and it more violently removes cost. So if you just open your mind up for a second, if everybody in the world could work at your local Starbucks or anybody in the world could magically sell coffee beans to your local coffee shop and they basically had a competition, you would have a barista for $1 a day, $2 a day living like a king or queen in their native location. And if the beans could magically be transported without all the distribution between where the coffee beans are and the roaster and eventually the cafe, just allow your mind to think about how powerful that is. Your cup of coffee and that experience would be delivered faster and better and cheaper. That’s all. And you wouldn’t have the friction of I have to hire this person, I’ve got a union, I’ve got to pay taxes, whatever. It just boom, just happens magically in this global cloud, this machine known as the interwebs.

Lon Harris: 32:08 Yeah.

Alex Wilhelm: 32:09 Do you think that we’re going to see a little bit less winner-takes-most or winner-takes-all? Because I’ve seen several different subnets moving towards this model for their miners, and I fully understand why it’s good for the subnets. But when I think about expanding economic opportunity, Jason, kind of what you’re saying, to me if it is going to be winner-takes-most or winner-takes-all, there’s less low-hanging fruit for people that are trying to get to that top rung of a particular miner task for a subnet. I’m just curious if that’s a concern for you as well or just good economic consequences.

Jason Calacanis: 32:37 Not, yeah, no. I mean you’re a big fan of capitalism. What happens if the buy side doesn’t make a good enough offer to labor? The supply side.

Alex Wilhelm: 32:44 Oh, they don’t hire enough people.

Jason Calacanis: 32:45 They don’t leave their house. So if you’re going to pay $10 an hour to be a barista and people can work from home for 15 bucks an hour, they’re not coming to your Starbucks to work. Apple stores faced were faced with this and so were Starbucks and—

Lon Harris: 33:00 Coffee bean and everybody else when Uber and DoorDash came out. It turned out working for one or two hours picking your own schedule and making, you know, $21 an hour was better than getting up at the crack of dawn, showing up at 5:30 AM at a Starbucks and having to work an eight-hour shift and not knowing what shifts you were working. It just was a better offering. So this actually empowers, I believe, labor if you consider labor, you know, the supply side on these things.

Jason Calacanis: 33:29 And the customers are the people who want their videos to be upscaled. So the people who lose are the people who are charging a higher price or and/or giving a lower level of service. The winner is best price, best service. And sometimes it’s some combination of that. If the validator, which is the person who basically arbitrates, did you actually perform the task? Again, in crypto, uses a lot of fancy terms or a lot of wonky terms. A validator is just the person who checks the work. Okay, you made a flat white. Is this a good flat white or is it garbage? Is it an acceptable flat white, you know, or fuck off. You know, so the validators will determine that. And then the person who’s buying the service will say, well, like, this wasn’t good enough, I’m going to go use another service. So they could go use the desktop software or Amazon’s offering.

Alex Wilhelm: 34:18 Sure.

Jason Calacanis: 34:21 Amazon’s offering might be faster, it might be better, but it might also be twice the price. And then the BBC might say, you know what, these these things haven’t been upscaled for 50 years. We’re in no rush here. If it takes, you know, a week or a month or a day, it doesn’t matter versus an hour, a minute, or two hours. So here we are, folks. This is capitalism at its finest. And if you ever dreamed of, you know, one human race working on projects beyond borders, beyond governments, beyond regulation, that’s what makes this so compelling to me. It kind of cuts across humanity and just makes a global competition for capitalism. And if the people who lose are people in the West who are coddled, overpaid, lazy, whatever, and the people who win are the ones who are the biggest hustlers who, you know, will do it for the right price. But if you don’t offer a decent price, people don’t participate. So you have that backstop, right? So before anybody gets, you know, their hands wringing too hard here and they rip the skin off their knuckles. The the guys in Vietnam doing this four-person team, if they get a job, you know, for a Korean company or Japanese company and they pay them more, they stop participating in the subnet.

Alex Wilhelm: 35:37 Yeah.

Jason Calacanis: 35:39 Again, back to capitalism. Back to free markets. This is free markets, like, unconstrained.

Alex Wilhelm: 35:46 What do you and I like, Jason? Free markets.

Jason Calacanis: 35:51 Free markets.

Alex Wilhelm: 35:53 And unconstrained. Hey, those are our favorite things. What I like though, just as a last thing, I know we’ve got to move on, but your point that buying Tao is a bit like an ETF into the Bittensor economy, I like that a lot because frankly… Honestly, I don’t think right now I would have confidence to say, Jason, here are the top three subnets. These are the ones that are going to win. The other 125 or whatever, but I don’t mind having kind of like a, effectively a call option on decentralized global intelligence. So that translates well for me. I appreciate that.

Jason Calacanis: 36:15 Yeah. And you know, listen, everybody wants to debate me and like, I’m not writing the code of Bittensor. I haven’t been like a market participant in crypto aside from, you know, my, you know, some small bets we’ve made as a family office, but it’s, my job is to explain it in plain English and to explain my thesis. So for people who are like, you’re pumping this, for people who are like, you don’t know what you’re talking about, of course I don’t know what I’m talking about. I am an investor. I am a commentator. So the way I choose to place bets is not by writing code, not by being a subnet participant or launching a subnet. My way is studying, placing bets, losing bets, learning, placing more bets. And that’s the job of an investor. You learn by betting. My job is to learn by betting and to have conversations, which is what I’m doing here. So to the mids and the weirdos, now I’ve got all these crypto weirdos back in my replies again.

Lon Harris: 37:21 You poor thing. It’s the worst.

Jason Calacanis: 37:24 I’m trying to think of like what’s worse, like the MAGA bots who are still loyal to Trump after he starts a forever war or hopefully not a forever war, but you know, that group of people in my replies or the crypto people in my replies saying have fun staying poor. And I’m like, guys, I’m worth collectively more than all of you. I’m doing okay guys.

Lon Harris: 37:36 Yep. Yeah.

Alex Wilhelm: 37:50 It’s so tribal though at times. And I think I really hope that maybe Bittensor can help break some of this, but people get so parasocial about their favorite token that it becomes nigh religious.

Jason Calacanis: 38:02 It’s the nature of humans, Alex. I mean, I literally when I tell Mac people like back in the day, like how some of the Mac decisions were, they would lose their minds. When I would tell them how like pathetic it was that Microsoft couldn’t build a UX and have less than 300 things in submenus, like they got upset. I’m like okay, this is just valid criticism. Like people want a headphone port. People want an on-off button. You know, like just give them the volume buttons. You know, it’s not like a big thing to keep a headphone port on your laptop. Like bad decision to remove the headphone port. I’ll die with that. Yeah.

Alex Wilhelm: 38:40 I’m still mad about it and I own several pairs of wireless headphones, but I also like to listen to audio sometimes, Jason, for more than 20 freaking minutes.

Jason Calacanis: 38:46 Exactly. Alright, let’s keep going. Alright. And I welcome the feedback from crypto mids. Congratulations on hitting, I was some, I was getting into a fight with a guy about this stuff and then I looked, he had eight followers and he started… So this account in 2013, I said, ‘Congratulations, I’m almost hitting double digits in your follower account.’ I’m like taking this account seriously for 15 minutes of my life.

Alex Wilhelm: 39:11 That’s Jason, that’s less than one follower per year net.

Jason Calacanis: 39:13 Literally, it’s 0.6 followers per year. And by the way, I get 20, I probably get 100 bots following me per day, right? It’s not hard to get past 10, bro. And then I look, he’s tweeting 20 times a day with eight followers. I’m like, what’s the point?

Alex Wilhelm: 39:33 You have to be pretty toxic to tweet that much and end up with no followers unless you’re literally blocking people that are following you. Anyways, all right, next up, we’re going to bring my new best friend, Ola Leiman, on the show. Ola is a social media poster. He builds with OpenClaude, builds skills for Claude Code, runs classes for people and built, Jason, what I think is the single coolest skill for OpenClaude I’ve ever yet had on the show, though we are going to show it off inside of Claude Co-work. Ola, welcome to the show. Unmute your microphone and tell us all about Andrej Karpathy’s idea behind the Council of Advisors.

Lon Harris: 40:07 Yeah, great to be here. So I saw Karpathy’s post. For some context, Karpathy was co-founder of OpenAI, later head of AI with Tesla. So AI Giga Chad. And he posted about his LLM council, which basically has the idea of you have a question, you send it to different models. So Grok, Claude, ChatGPT, all of that. They all come up with a different answer. Then the answers get anonymized and then they get peer-reviewed. So every single model reviews all of the outputs in an anonymous way. And then they get to the chairman, that’s like another model at the end. And the chairman just comes up with a final verdict. So that was the idea of Karpathy’s council. And I really liked that idea, but I thought, okay, it would be cool to build something in Claude that’s more based around basically giving you basic advice for everything in your business and also for general life advice. So…

Alex Wilhelm: 40:58 Is your council of advisors using different models for each of the advisors that debates the question or is it one model with different personas attached to it to kind of talk amongst themselves?

Lon Harris: 41:06 Yeah, so this is just one model. So I’m just using Claude Opus 4.6 right here. It’s because most part of my audience are really non-technical, so I wanted to build something that’s still kind of simple but still uses the same concept that Karpathy uses. So I’m just using Opus right here.

Alex Wilhelm: 41:26 Okay. So let’s ask it a question. Let’s let Jason pick a question and then as it thinks, I want you to talk us through what it’s doing. So Jason, what’s a good startup question for founders that they be asking themselves today?

Jason Calacanis: 41:41 Sure. How much equity should I give my VP of Engineering who has five years of work experience and who previously worked at Google? I’m a seed stage startup that’s raised $3 million. I am a solo founder. And I have 85% of the equity and my investors have 15%.

Alex Wilhelm: 42:04 I wonder how many tokens this is going to consume?

Jason Calacanis: 42:06 Not a lot. It’s not so bad. And and this is how, you know, the more specific you get, the better the models tend to do. And then what you’re doing, Oly, is you’ve given one model, Anthropic’s exceptional model, but you you’ve have it have four or five different personas. So you’ve loaded into those personas a contrarian, like why would this fail? You’ve got a first principles thinker, an Elon Musk on the team. You’ve got an expansionist, which would be like a JCal: how does this grow faster? You got an outsider…

Alex Wilhelm: 42:43 …who’s like, I don’t I don’t even know what you’re talking about, I’m dumb.

Jason Calacanis: 42:46 Yeah, it has no context. And then you have an executer…

Alex Wilhelm: 42:49 …who’s like a sycophant, who says ‘hey, I’ll get it done for you’ or something.

Jason Calacanis: 42:53 I would put the sycophant in there too, but okay.

Lon Harris: 42:56 Yeah, this is the basic idea. And again, what’s important that you have the step here to anonymize all of the responses before the peer reviews happen. That’s like one of the the key steps here. So once that happen, we can see if they already… in the beginning you could spin up five different sub-agents. So this is pretty cool because Claude Code or Claude CodeWorks. So this also works on Claude CodeWorks for everyone watching. Um, I just like to use Claude Code more but it’s just the preference here, it totally works too. And again this is because Anthropic now has these agents working inside all of their products so it just spins up different sub-agents.

Jason Calacanis: 43:31 So it does it in parallel, saves a lot of time. You anonymize the five different responses because when you go to the peer review stage you don’t want it thinking ‘Oh this is the expert, this is the contrarian,’ you want it to just evaluate it, you know, without any baggage, correct?

Lon Harris: 43:52 Exactly. And I read Karpathy’s thoughts on that and he was quite surprised that in his version he used different LLM models and they often times rated other LLM models higher than themselves once this peer review thing kind of got started. So it really helps because most of these models have some kind of bias, especially when it comes to rating other stuff. So with this this step it really helps to to get to a better like more balanced opinion. This is also why I chose these, JCal already like talked about this a little bit. I tried to include opposites a little bit so the contrarian is kind of what is the downside, the expansionist is what is the upside, the first principle thinker is kind of takes a step back it goes deeper into the thinking and the executer is just going forward and just full throttling, not really thinking just doing. And the outsider…

Alex Wilhelm: 44:41 Maxing. Just call it what it is, Oly.

Lon Harris: 44:44 I’m actually gonna drop an article today about JMax so that’s funny you talked about that.

Jason Calacanis: 44:51 Yeah, I’m sorry to everybody who’s offended by the word but it’s in popular culture right now. It’s not going backwards. Um, it is what it is. But for people who don’t know… Stop ruminating, don’t go to therapy, don’t overthink things folks, just go for it and build. It’s all that matters is building and making your venture capitalists more money faster. Let’s go.

Alex Wilhelm: 45:12 Yeah, so you can see all the advisors now responded and is running the peer review is also complete so now it all comes to the final stage here, stage three, which is the chairman. And so the chairman gathers all the data and all of these different responses, so the council members rated okay who has the strongest response, who is the weakest also what I can blind spots that some of the reviews missed and also I added this thing so this is not from Kapathi but I thought it’s kind of a good data to have is the what did all five miss so they all sit down and try to find things that all of them missed. So that’s kind of an extra angle here.

Jason Calacanis: 45:51 Ooh, love it. Now wait how many people are on the peer review and did the peer review people have personas?

Alex Wilhelm: 45:59 So the peer review still happens in these personas.

Jason Calacanis: 46:01 Ah got it, so the five are evaluating each other’s answers without knowing them.

Alex Wilhelm: 46:06 Exactly yeah and they go through these three questions here, there’s a little bit more context in the skill but this is kind of the basic principle of it and then they go to the chairman and once the chairman has all the data he creates a report for you and this will be shown on a dashboard so hopefully we’ll get there yeah it’s still doing the hard work.

Lon Harris: 46:26 Well while it works Alex and while we wait for it to finish doing its number crunching, can you just quickly break down for folks when you like to use Claude Code/Co-work and when you prefer to use Open Claude because I’m seeing a lot more folks in my circles start to tinker more with Co-work and I’m still a Claude guy so I’m curious pros and cons and what I’m missing.

Alex Wilhelm: 46:44 Yeah that’s that’s a really good question, I would say for me the most like more serious in-depth work I like to do in Claude Code or Claude Co-work just because most of my open Claude is in my telegram, I think the big upside of Open Claude is that you have the memory system built in, you can also get there with Claude Code using things like Obsidian for example or using projects now in Co-work so it’s kind of it’s becoming a little bit more alike over time I would say. Um but in general to me for for more in-depth work I like to use Claude Code just to be very honest with you because I like the interface more. You can probably do most of the work I’m doing inside Claude Code or Claude Co-work inside Open Claude. It’s also a little bit I would say habit because the truth is I could potentially do most of the work I’m doing inside Claude Code or Claude Co-work inside Open Claude.

Lon Harris: 47:45 Yeah.

Alex Wilhelm: 47:47 One big difference I would say is especially for non-technical people who I mainly talk to is you have better or like easier control inside Co-work. So you can just go into your connected apps so let’s say Gmail or Notion and you can just see these are the permissions I gave my Co-work but if you look at Open Claude sometimes it will just do things and you

Lon Harris: 48:00 Don’t really understand like which permissions you gave to it or you maybe gave it an API key, it’s still in there. So there’s some files, some stuff is saved but you don’t have a good overview. So I just think that Anthropic is doing a great job of just making it more accessible and yeah, just easier to use for non-technical people.

Jason Calacanis: 48:17 Now is this all open source Ole and is it a skill in OpenClaude? Where does this exist if people want to build or are you building this as a proprietary startup that I can invest in?

Lon Harris: 48:29 Yeah, so you can just get this from either my X account. So you can find me at @itsOleLehmann and I pinned it to my profile here. So it’s an article here that I wrote about it and you can find this as a Google Doc in there. I also put it onto GitHub but again I’m mainly talking to non-technical people.

Alex Wilhelm: 48:44 And we will put this in our enhanced show notes. I have given Producer Jacob and Producer Oliver explicit instructions that I want enhanced show notes.

Jason Calacanis: 48:54 I want enhanced show notes, please.

Alex Wilhelm: 48:57 Yeah.

Jason Calacanis: 48:58 So I want them running the show notes and asking our own LLMs, our own council, what did we miss in the transcript that could have been linked to? And I want to do that like twice. And I want to have absurd show notes going forward with tons of links in it. Why? I take notes myself Alex when I’m listening to a podcast or I’m listening to an audiobook. So let’s really do that with our show notes. I want enhanced show notes, please. And if somebody wants to make a skill for OpenClaude to do this, I will put out a $1,000 bounty and on let’s say May 1st I’ll give the $1,000 bounty for the open source project that proves they can do incredible enhanced show notes for me for TWiST. Okay, there. I’m doing this new… this is a new program we’re going to do here Alex which is the bounty program. I’m going to offer money for OpenClaude skills so that we can keep the capitalists who I’m invested in from killing OpenClaude. This is our format here is demo or die folks and so in the tradition of the… what was the original computer hacker club called Alex in… this is like a good Jeopardy… was the computer something club and it was in like Palo Alto or whatever in the 70s and like Steve Jobs used to go to it and they would just have demo or die. What was that called?

Alex Wilhelm: 50:24 Chaos Computer Club?

Jason Calacanis: 50:26 No, it wasn’t the Chaos Computer Club. It wasn’t like 2600 out of New York.

Alex Wilhelm: 50:30 Homebrew Computer Club! Founded in 1975, Menlo Park, Palo Alto, premiere 1970s hobbyist group. Steve Wozniak, Steve Jobs, the Homebrew Computer Club.

Jason Calacanis: 50:37 I want to do an episode on the Homebrew. We’ll do like an acquired episode, you know like when they go like do a deep dive. Jacob, put this on for the summer schedule. Deep dive Homebrew. And I want to do an oral history of the Homebrew. This is new from this week in startups from the mind of executive producer J-Cal. I want to do an oral history documentary on it and we’ll debut it as a wonder-

Lon Harris: 51:00 Two parts during our summer session.

Alex Wilhelm: 51:02 And now we have the council report and it… it gives a recommendation first. We can just have a look at it.

Jason Calacanis: 51:09 Mhm. Take a look here, the recommendation 0.75 to 1%. Offer 0.75, negotiate up to 1% ceiling if pushed. You’re a seed stage with 3 million raised and a series A coming that will dilute everyone by 20-25%, okay, that seems likely. I would have said 15 to 20%. Google 10 year and five years experience are real signals, correct, that’s why I gave them. But they don’t justify top of the market. You need room for two to three more senior hires before or after your series A, that’s true. Structure it as ISOs with a four-year vest, one-year cliff, that’s standard double-trigger acceleration, that’s a fancy way of saying if the company gets bought they get their vesting pushed forward in case like the acquirer doesn’t need the CTO. If the candidate pushes beyond 1%, shift the conversation to base salary first. Moving base up is cheaper long-term than moving to equity, that’s true if you succeed because equity compounds through dilution rounds and salary does not. The one thing to do first: create the option pool before this conversation. A 10 to 15% option pool documented and board approved so the VP offer is priced against real post-pool cap table, not a number that will change the moment you formalize the hire. This is critically important. This is something I wasn’t even thinking of, so that’s a great, great punch-up there that I put in yellow. Where the council agrees: every advisor landed in the same mechanical range, 0.5 to 1.5. And I didn’t do CTO, right, I did VP of engineering I think, which is like a step below CTO. What the council clashes on: the expansionist push for 1.5 to 2%. That’s what I was going to say, I was going to say 2 to 3% but I am the optimist pushing for like getting great candidates, I would have said 2%. Blind spots council caught: cash-equity tradeoff, option pool math, and role trajectory. Does this person eventually become the CTO? Wow. Oh my lord, this is exceptional. Like if you had a discussion with your board as a founder you would not get to this answer as crisply and as cleanly as possible. This is why I think you should create a digital board of directors from day one of your startup and having it get complete access to your Slack and and working on these decisions, really genius. Amazing.

Lon Harris: 53:18 Can you click on just one of these for me, Ola, so I can see what it looks like?

Alex Wilhelm: 53:22 Yeah, so this is one response. The output, the individual outputs of the… of the different council members. And it also gives you some highlights here and some some blind spots. And I… I think the general idea is, everyone who like makes important decisions has a council, right? Like the president has a council, right? Like if you… no… in every decision is made by having influence by these people that have in-depth knowledge and if you have a startup you would have never had these opportunities or these resources before.

Ole: 54:00 So like J-Cal said, now you have an LLM and LLMs are so good to like stepping into another role and also it saves a lot of your own biological compute because if you have to think and empathize with a different, different opinion every day, it takes so much of your, of your own energy. So for me it’s just helpful to let these different council members run through, through ideas because yeah, it’s a, it’s such a big, big advantage and most people don’t use LLMs for that and they’re extremely good at that.

Alex Wilhelm: 54:30 You said that Claude Mythos is quote Hiroshima for software. Um, that’s a very vivid image you might say. Uh, why are you that concerned? Right now it’s under lock and key. Jason and I talked about Project Glasswing, and seems like we’re in a bit of a holding pattern. Why are you so scared?

Ole: 54:47 I just think it’s a point in time, it’s, I don’t want to be too alarmist, but it’s one of these points in time where it just makes sense to take a step back and take your security seriously, because the downside if you don’t is just so immense, right? Like there’s just so immense. Uh, if you have a big moment of uncertainty, it’s usually smart to take some measures. And I think, um, just taking these basic security measures is something everyone would do, but most people never do them. I think if you, if you just ask the random, random person like, what are your security measures? They are incredibly bad. So this is why I made this post. The headline was kind of extreme just to get attention, but the, the meat of the post was just hey, these are the basics of security, and this is a good point in time to take them seriously. On the other hand, we don’t really know how advanced Mythos is, right? Like maybe you do all the security measures, but it’s still going to be so advanced that it’s just a, a drop in the ocean, right? So that’s the big question here.

Alex Wilhelm: 55:43 And just Ole, before you go, what’s your story? Where are you based, what do you do for a living?

Ole: 55:47 Yeah, so, um, I’m based in Cyprus. I’m German, but I’m in Berlin right now. Um, I’m the founder of the AI Solopreneur, which is an AI media and training company. So I mainly, mainly do media around AI, um, I do some B2B consulting for companies. I’m focusing more on GTM engineering, so everything AI automation-based, um, around marketing, around building distribution.

Jason Calacanis: 56:11 Alright, you’re hired. Everything you want to work in venture capital, you’re hired. You’ll be the first and only European employee I’ve ever had.

Lon Harris: 56:22 You seem brilliant. Uh, really appreciate you coming on the program. And everybody sign up for Ole’s Substack.

Alex Wilhelm: 56:29 Yes, it is AI solo dot beehiiv dot com. B-E-E-H-I-I-V dot com. Screw Substack Jason, that’s so 2024, everyone’s now on the Hive. Yeah, Beehiiv’s a great product too, and…

Jason Calacanis: 56:41 I love a little capitalism and…

Ole: 56:43 I’m trying to bring capitalism back to Europe I’m saying.

Jason Calacanis: 56:46 Yeah, good luck. Be careful, you might get shivved. Uh, you know… What are you paying in taxes? What’s the, what’s the high tax rate where you are?

Ole: 56:51 I mean, that’s why I moved my business to Cyprus and I have residency there. This is like 12.5%, so I’m, I’m pretty fine with that.

Jason Calacanis: 56:59 This is amazing. Now if you- If you’re still in Berlin, you’d be paying 52% or something?

Ole: 57:02 Yeah, it’s around 50, yeah, that’s what I’ve paid before, yeah.

Jason Calacanis: 57:04 Yeah. See this is one of the important things that people have to understand. The most savvy entrepreneurs, especially ones outside the United States because unlike Americans, Europeans can move to Dubai and pay like one or zero percent, Singapore, whatever, and they get tax treatment Alex that’s extraordinary. If you’re an American and you try this, you still have to pay your federal. So you’re still on the hook for 30% or so. And the Americans move abroad, the only thing they get a benefit from is state taxes. What do you pay state taxes Alex, what’s the vig up in Rhode Island?

Alex Wilhelm: 57:43 I don’t know, but let’s find out.

Jason Calacanis: 57:44 Okay, yeah that’s how rich Alex is, he doesn’t even know, he’s just pouring in this money from all his incredible gains.

Alex Wilhelm: 57:52 Ah, so 3.75% up to 6%. Top rates for income over 180… oh, that’s not very high. Okay. All right.

Jason Calacanis: 57:58 I mean, it’s half of, you’re basically half of California and New York which I think are at 13% state tax.

Alex Wilhelm: 58:01 Yeah, but our state is 1% as large, so whenever we try to do high-speed rail, it’s only eight cars long, you know? So it’s pretty cheap to run a state.

Lon Harris: 58:10 Basically the tip of the train is already at the next station and people are still getting off the last caboose.

Alex Wilhelm: 58:18 I can’t believe we have two senators. It’s incredible to me. Like you can literally see across Rhode Island and we have two.

Jason Calacanis: 58:21 Shoutout to our founding fathers. Like what’s the least populous state? Is it still Wyoming or Nebraska or something? It’s unbelievable how our system works that a state with, you know, a fraction of the population of these other states get the equal… yeah, it’s Wyoming.

Alex Wilhelm: 58:37 Yeah, it’s Wyoming. 580,000 people. That’s so few, I’m pretty sure they have a lottery and you draw it out of a hat and if it’s your turn you have to be representative in Congress.

Jason Calacanis: 58:46 I mean, it’s basically, you have a… yeah, if you have two, you have a one in 250,000 chance of being a senator. I’m moving to Wyoming. Senator JCal from the great state of Wyoming. All right, Ole, we’re dropping you off, but you’re coming back on this program because you crushed it.

Ole: 58:58 Thanks, oh, I appreciate it, man.

Jason Calacanis: 59:00 All right, listen, everybody wants to know. JCal, you look great. You’re no longer a fat bastard. You know why? Because of ro.co. I have been doing a GLP. Yes, it’s true. And I lost 40 pounds. And I’ve added muscle. If you are GLP curious, ro.co is my partner. And they do a great job. Now, you don’t need to get Ro Sparks if you’re great. You might want to pick up some Ro Sparks while you’re there, but I’m going to say GLPs are life-changing. Do your research and then use the checker. They have an incredible insurance checker. If your insurance doesn’t cover it, man, it’s getting more and more affordable thanks to folks like ro.co. What’s the offer?

Alex Wilhelm: 59:42 It’s ro.co/twist if you want to check your insurance, ro.co/twist. It’s very affordable, very good, and you should all go check it out.

Jason Calacanis: 59:50 Uh, well let’s go off duty. This is your first chance to go off duty. We’re doing off duty at the end of the program. Why are we doing it? Uh, people love to know what… You know, what we’re into when we’re not doing podcasts or investing in companies or running newsletters. So, you’re first, Alex. What are you consuming, obsessed with, or otherwise find yourself enamored by?

Alex Wilhelm: 60:14 I think for me, it’s the egregiousness in a recent television show of a Microsoft bit of astroturfing. Now, we all know that Apple has rules. If you have an Apple computer in a movie, you’re not allowed to have the bad guy use it, that sort of thing. They’re very image-conscious. Microsoft, on the other hand, in a clip that our dear friend Lon pulled for us, that I watched and it’s terrible, goes to show how not to do product placements. So if you are on the Microsoft Copilot team, let me tell you this is egregious, Jason. Get your vomit bucket ready and tune in.

Jason Calacanis: 60:44 Now wait, this is yours or this is Lon’s?

Alex Wilhelm: 60:47 Well, we both agreed on it. We talked back and forth on the show today. Yeah, cause we co-produce because he has surgery and such, but—

Jason Calacanis: 60:52 But you’re endorsing this or no?

Alex Wilhelm: 60:54 I’m endorsing the badness of this. I’m saying that… Have you watched this? Okay, here we go.

Jason Calacanis: 60:57 Oh, several times. Okay, here we go.

Alex Wilhelm: 60:58 Several times. Oh yeah, watch. Just… Please enjoy.

Jason Calacanis: 61:01 Okay, here we go.

Ole: 61:02 Albuquerque, New Mexico, San Angelo, Texas, Yakima, Washington. You know, I can run it in Copilot. I’ll have it categorized police reports by location, date, and name, then color-code them by location and create a summary of the findings. Okay, here we go. Wait a second. I’m finding reports of alleged scams in multiple cities he’s lived in. He posed as a startup investor in Phoenix, a high-end real estate partner in Ojai. So he’s not a businessman, he’s a con man. So Frank was his next target.

Alex Wilhelm: 61:31 That is essentially a television commercial inside of a television show. And Jason, I’m not a big procedural guy, but if I was watching ABC’s hit show High Potential, I’d be pissed that they just sold me out to Microsoft. Who’s that for?

Jason Calacanis: 61:46 Wait, is that a Microsoft Copilot? Is that Claude Co-work?

Alex Wilhelm: 61:51 Oh no, I think it’s Microsoft Copilot. They were in Excel there. And they said Copilot directly. That’s not Co-work.

Jason Calacanis: 61:58 Got it. They did say Copilot. Okay.

Lon Harris: 61:59 Also, don’t you think Anthropic would have a slightly better taste than that?

Jason Calacanis: 62:04 Yeah, listen. Product placement is a great way for these folks to make money and reduce the number of ads, but it does feel smarmy to me at times. I’m a fan of like when they just have a Heineken in the background or something to that effect. But this type of product placement is like literally being bashed in the skull in the middle of a show. Takes you right out of the experience. I’m going to go ahead and say disgrazia at ABC.

Alex Wilhelm: 62:32 Yeah, terrible. But Microsoft is the one who should have just more quality control here. Like, you have trillions of dollars in market cap. Hire someone who has taste. Like, this actually makes me doubt their ability to make good products because they can’t advertise them well. Maybe that’s too much Apple in my head, but—

Jason Calacanis: 62:47 But I want to hear one thing. If you are spending more time promoting your product than making a great product—and, you know, I don’t know, I haven’t used this product so I’m not making a judgment on it— That is what I call pouring kerosene on a giant, you know, tree that’s been felled in the forest. You could start a fire, you might even start a forest fire, but generally, you’re going to wind up with a charred giant log. You want to build your company and your product through what I call kindling. Each piece of kindling, those little tiny sticks are the like first 10 users. You get that fire going, then you put sticks. After you have sticks, then you might put like branches. After branches, you might put logs, but you want those hot embers, and you want to go nice and slow, you know, and get those users in there who are the early advocates who give you great feedback. You don’t want to just go right to marketing with your startup. That’s a good poll. I like it. Here’s my poll. Uh, I have always loved the character Darth Maul. Darth Maul is the guy with the two-sided lightsaber. And they have created a series called Maul. And I have now watched the first episode three times.

Alex Wilhelm: 63:57 Three times!

Jason Calacanis: 63:58 With my 10-year-olds, uh, because I watched it myself, then they wanted to watch, then they wanted to watch it again. The first two episodes are out. I think they’re dropping two episodes for five weeks in a row on Disney. Here’s a trailer, I’ll talk over it. The first episode, super banger. Second episode, incredibly good. And, uh, you can see that the animation is very unique. The animation was created because they did some watercolor paintings and then they put that into the animation software and they created a very unique new look that I’ve never seen. The Clone Wars had its own style, but if you look at this, it’s very cyberpunk. Reminds you of Blade Runner. And this is like Andor in that it’s telling another story that occurred during the timeline. And it’s the story of Darth Maul who survives when he gets cut in half, uh, by a young Obi-Wan Kenobi. It’s like a dark underworld underbelly show. Now why is this also important? It turns out that the sequels, episodes seven, eight, and nine of Star Wars, were supposed to be having Darth Maul as the lead. Not bringing the Emperor back, not all this nonsense. They threw away George Lucas’s treatment, they destroyed the franchise with these disgusting episodes seven, eight, nine. Those should be decanonized, but instead they took the elements of George Lucas’s sequel and they just placed it between episodes three and four, when Andor occurs, when Rogue One occurs, when the Han Solo movie occurs, essentially when Luke and Leia are, you know, babies or toddlers, uh, before Obi-Wan goes and, uh, starts teaching Luke Skywalker to be a Jedi. It’s extraordinary, uh, already. I have great hopes for this one and, uh, you know, they’re just trying to get this, uh, Kathleen Kennedy disaster of a and J.J. Abrams who basically these guys just, they took Star Wars off the rails. They’re trying to fix that with Andor, Maul, etc., and I…

Alex Wilhelm: 66:00 I am here for fixing the uh disgrace-yad of Kathleen Kennedy’s horrible reign post-George Lucas and throwing Lucas’s genius in the garbage. Um, yeah. And so it’s—

Jason Calacanis: 66:14 Which one had Rey in it?

Alex Wilhelm: 66:16 That was the disgrace-yad sequels. So 4, 5, and 6 are the original Star Wars.

Jason Calacanis: 66:21 Uh-huh.

Alex Wilhelm: 66:22 7, 8, 9 were supposed to be the Darth Maul ones that would have Luke and Leia reestablishing the Jedi Academy training a new series of Jedi. We didn’t get any of that. We didn’t get any of that while Carrie Fisher was alive. There’s a lot of resentment in the Star Wars community because of all of this. They didn’t even have the story arc for 7, 8, 9, you know, done. They just did a money grab because Star Wars was bought for whatever 4 or 5 billion by Disney and they just killed the entire IP. Now they’re going back trying to fix it. Um, and I think they just have to—

Jason Calacanis: 66:57 Jar Jar Binks though was in the prequels, so these are post-prequels.

Alex Wilhelm: 67:00 1, 2, and 3.

Jason Calacanis: 67:01 Okay. Thank you for correcting my Star Wars knowledge.

Alex Wilhelm: 67:04 So there’s the original trilogy. Yes. There’s the prequels and then there’s the sequels. The sequels some people will refer to the prequels as, but if you look at it as a timeline, it’s episodes 1 through 9. He released 4, 5, and 6, the most interesting piece in the 70s and 80s and into the 90s, then he released the prequels in the 2000 era, and then they released the sequels and anyway, it’s well worth checking out. And if you want to, if you’ve got kids or if you’re just a Star Wars fan and you’re on the treadmill, highly recommend Clone Wars, Bad Batch, Rebels and now Maul. These are like the animations that fill in the series, but these are serious animations telling serious stories in the way Japanese folks do with their anime. So it’s very influenced by all of that. And then Mandalorian, Ahsoka, the TV show Ahsoka, those are live action. Those kind of thread into here, Andor is live action, those thread into this awesomeness. The second thing I’ll give you is I just bought the new MacBook Pro. Uh, and here it is. I bought the 14-inch. I spent $3500 on a computer, on a laptop, for the first time in a while. Actually, I have a Dell Alienware that I spent a similar amount on, but this MacBook Pro 2 with the great screen is amazing. And I think the future is going to be SMLs, small model languages.

Ole: 68:08 Yes!

Alex Wilhelm: 68:10 I think doubling your spending, getting one of these M5s, you don’t need the M5 Max, that’s for video, but the real key is getting to that 48 gigs of RAM. And you know, very simple things in your workflow, you’ll notice are much faster. On the other end, there’s this neo MacBook that uses like a micro— a mobile chip, that’s for $600. And then here’s your six times more, like a $3600 laptop when you put it at max memory for that M5 chip, you get to 48. There’s like a larger one you can get to with 128, but then you’re in the 6—

Jason Calacanis: 69:00 inch, which is like 5 pounds. It turns out the MacBook Pro is essentially the same weight as the MacBook Air. I was a huge MacBook Air fan, that’s what we were standardized on, but I think in the future we’re going to be putting everybody on higher-powered ones so we can run Open Claude and local language models in the future.

Ole: 69:16 Yeah.

Jason Calacanis: 69:18 So I just want to let folks know who are wondering if they should do it, in my opinion, you should just go ahead and upgrade to a MacBook Pro and spend the 3,500 bucks.

Alex Wilhelm: 69:27 3,500 bucks is like a month of child care, get yourself a real computer. Although I will say, my MacBook Neo that I bought very recently just because I wanted to have what I call—

Jason Calacanis: 69:34 You bought the Neo?

Alex Wilhelm: 69:36 I bought the Neo! I have one.

Jason Calacanis: 69:38 What are your impressions? Oh my god. This is like a perfect Off Duty, go. What color did you pick? And you paid 600 bucks for this thing?

Alex Wilhelm: 69:44 I got the pink one. I got the cheapest one. Uh, I got pink for two reasons: one, it’s my second favorite color after blue. And also—

Jason Calacanis: 69:52 Yeah, you wear pink shirts.

Alex Wilhelm: 69:53 Yeah, yeah, I can give it to my oldest daughter when she’s computer age because she loves pink, so problem solved. Uh, I bought it because I wanted something that I can just throw and take with me. Uh, I have a lot of big heavy computers, desktops, large monitors, my twist MacBook Pro—

Jason Calacanis: 70:05 What does that thing weigh? A pound and a half, two pounds, something like that?

Alex Wilhelm: 70:09 Well, it’s not in the case of the twist computer, it’s not that I care about the weight, it’s more that I don’t want to do—

Jason Calacanis: 70:13 No no, the Neo. What does the Neo weigh?

Alex Wilhelm: 70:15 The Neo weigh, I will find that out for you. Uh, but the problem is, Jason, it is the first Apple computer or Apple device I have ever used, 2.7 pounds, that feels cheap. Like when you type on it, it does not have that Mac polish.

Jason Calacanis: 70:28 Ah, so not—

Lon Harris: 70:29 Okay, yeah, so making a $600 computer means there’s going to be compromises. But clearly they’re going after the Chromebook market and, you know, kudos to them. I said years ago, what should they do with their, you know, massive cash war chest, you know, the hundreds of billions of dollars, they probably still do, it was the largest hedge fund in the world at a point in time, like their treasury. I just said why don’t they just lose money on iPad Pros with keyboards and put that into the education sector and come out with a $200 iPad and just make it the standard? Because remember the iPad was kind of like this forgotten kind of compute device and I was like, what if they could just get many more of those into the operating system and make the money downstream on services. I believe that’s what they’re doing with the Neo. I don’t—maybe they make 100 bucks on this thing or 150 bucks, whatever they make on it, it’s not going to change the fate of Apple, but it’s going to train young people, it’s going to get them onto iCloud, it’s going to get them to buy additional storage, it’s going to get them to buy Apple Music, it’s going to help the App Store. Why not get another billion people in the emerging world to use Apple computers as opposed to Windows or Chromebooks? Brilliant idea by them, but hey, listen, there’s going to be some compromises. But 600 bucks is the right price. That’s the price of the Mac Mini.

Ole: 71:46 Yeah.

Jason Calacanis: 71:47 So here we are folks, you know, Apple is really trying to get market share away from Windows and Android.

Alex Wilhelm: 71:54 And I’ll just say, it’s the first cheaper Apple device, it’s still great, but you can just tell if you go from living in a very— Apple heavy world, that this is a step up from HP and a step down from your MacBook Air or MacBook Pro. But also here’s the other thing Jason, I’m an enormous nerd. Like I use computers all day long. Different keyboards, different setups, whatever. Most people think AI is ChatGPT and so for them I’m pretty sure this computer is amazing and I don’t think, I cannot imagine recommending any Windows laptop when the MacBook Neo exists. It’s too cheap, it’s too good, and it comes in fun colors. What more do you want in life?

Jason Calacanis: 72:29 I mean, it’s pretty great. I’m going to give one more shout out here. So, I have been looking for… you know I like to understand the seminal works in an industry, but a lot of them are out of print, and a lot of them you don’t even know about them, so I ask Claude or Grok or Gemini what are the seminal works in advertising, in typography, in design. And so we’re building a newsletter for This Week in AI, we’re going to launch a paid newsletter, and I’ve been just trying to get some inspiration. And we had the guy from the Hustle and My First Million on, gosh I’m sorry I’m blanking on his name.

Alex Wilhelm: 72:55 Oh we always forget his name… I’m on it don’t worry.

Jason Calacanis: 72:58 Not Ben, gosh. Sam Parr. Sorry. And so there’s a guy Nigel Holmes as an example. And he was the person at Businessweek who came up with infographics, right? Where they would use wine bottles as the bar charts, all that kind of stuff. Would use actual slices of pizza and it just was a way to make graphics more interesting. I love infographics. So I buy, I go online and like, hey it’s not in print, and it’s like well you could buy the version from 1983 for five bucks. And then I buy the book and I don’t feel guilty about writing my notes right in it. So Nigel Holmes is this, you know, he’s the godfather of design. You can pull up this book, ‘Designer’s Guide to Creating Charts and Diagrams’ and I was going to give this, but I’ll just tell Oliver and Ismail who are working on the newsletter to just go ahead and feel free to buy old books, put them in the library. And this one, you know, came from a thrift bookstore, I don’t know what I paid for it but I paid basically nothing for it, and I think the book came out in like 83 or 84.

Alex Wilhelm: 73:45 Yeah, you can copy it for 54 bucks on Amazon if you want. Totally affordable.

Jason Calacanis: 73:54 Well that’s 54. I got it for like 20.

Alex Wilhelm: 73:57 Oh no I’m just saying like usually when I look at a book from 91, I’m adding a zero to that price. So this is super affordable for a book that old.

Jason Calacanis: 74:07 Yeah I think, does it say 54 bucks on there? What does it say there?

Alex Wilhelm: 74:11 Yeah so if you look at the used, you’ll be able to find one that’s not 54. This is just the first one that came up. There’ll be like, there’s a button for other sources or for the original library version. I’m trying to not accidentally show off my address on live.

Jason Calacanis: 74:51 Oh no problem. So then I got Ogilvy on Advertising, this was like an old book on advertising and then… And there was this, My Life in Advertising and Scientific Advertising by David— these are two works by Claude Hopkins. And you know, my thesis is there are a lot of lessons that were learned in the 70s, 80s, 90s when people were using other mediums that I find inspire me. They just give me great inspiration. And you know, if I have a stack of them sitting there and I’m having a cup of coffee, I just take one out, I get the highlighter, I write some notes in the margin and it’s my little workbook, it’s my little treat to myself. Maybe I’ll bring one of them with me when I’m on a trip and open it up, and you’re a fan of books I see.

Alex Wilhelm: 75:39 Well no, I’m a fan of scribbling in them. Like this is a book I read with my dad, Democracy in America. If you haven’t read it, you should read it. But you should scribble on every page of your books. Like take notes people, enjoy yourself. It’s a $10 book, it’s a $20 book, look at it as a moleskine and just get in there.

Jason Calacanis: 75:52 It’s a wear part, as we say in my house.

Alex Wilhelm: 75:54 Now, do we have time for one more off— okay.

Ole: 76:00 So have you read a book called Hyperion?

Jason Calacanis: 76:04 No, but I’m aware of it.

Ole: 76:06 Okay, well first of all Jason, my gift to you, read it. It’ll blow your mind, it’s crazy, you’ve never read anything like it. It’s fantastic. And one thing that I’m really excited about that’s coming down the road, wish we had Lon here, he would know what to say about this, but they’re making the movie adaptation. And I’m torn because on one hand I’m excited that they’re trying to take this monster story to film, but I really think it would have been better as like a one-season Game of Thrones-length series. So while we’re waiting on that though, there’s quite a lot going on in the world of science fiction. And what I wanted to give a shout out to on off topic today is this series of books from a woman named Bethany Jacobs. I’m reading the audiobook version while I push my kids around in strollers. It’s called The Kingdom Trilogy. First book is These Burning Stars. I am not a— I’m not into books that are like thrillers per se, like people chasing people around the solar system or whatever. This book has elements of that but it has some of the best world-building and characterization I’ve ever read. I’m in the middle of book two and I’m just absolutely hooked. It’s the type of book where like I don’t want to go to bed because I want more of it, Jason. And it’s fantastic they’re making such amazing art. And I just hope also that the mall show you just showed off, I hope that in time it becomes cheaper to make stuff that awesome so that way not just Hyperion, which has been a famous book for decades, but also things like these books from Bethany Jacobs can also get their treatment because her works here are not nearly popular enough for HBO to pay attention. But they’re good enough, you know? And so I just— I just want everyone to have the capability to make these awesome stories visual because we’d be so much richer, I think, as storytelling humans if we had that capacity.

Jason Calacanis: 77:42 Yeah. And one thing I can recommend to you, which I had on a previous off-duty, was the Shokz open-ear headphones. And when you’re walking with the baby, right? You need to hear if the baby wakes up, right?

Ole: 77:54 Mhm.

Jason Calacanis: 77:55 And so Wirecutter, which I tried to acquire but— I try to acquire everything— The New York Times bought it. Um, they have reviews of these bone-conducting ones at the Wirecutter. Uh, that’s the open-air fit.

Lon Harris: 78:09 I’d rather the wrong one.

Jason Calacanis: 78:10 I like… oh yeah, I look for the… the OpenRun. Yeah, you see the OpenRun right there?

Lon Harris: 78:15 Well, if their website would stop trying to teach me how to use the internet…

Jason Calacanis: 78:18 Yeah, do everything, yeah. So anyway, this thing looks a little weird but…

Alex Wilhelm: 78:21 Oh!

Jason Calacanis: 78:22 Goes over and it’s bone-conducting or it will just play it but your ear is completely open. So now you can be listening to this and you’re alert of your surroundings. So when I’m walking around a city like in Japan, it doesn’t matter if I’m wearing my fancy incredible high-end headphones and not listening to the outside world, but when you’re walking in New York or San Francisco and you gotta be alert—there could be like, you know—I’m not being facetious here, you do need to not have your AirPods on if you’re on the New York City subway or walking around certain areas of San Francisco. You want to be alert. And if you’re with a kid you want to hear if they’re crying, boom, you wear one of these open-air ones and you’re good to go.

Alex Wilhelm: 79:03 So I just went to New York City last week for a couple days and I was remembering what you said about it—a little gritty, a little, a little dangerous? I’m not gonna lie, Jason, I can happily report that it was great. And I even took the subway by myself, I didn’t get lost, it was super clean and now the display up that tells you what stop’s coming up, it’s—I don’t know, I was kind of proud of America.

Jason Calacanis: 79:25 Where were you? Manhattan or Brooklyn? Where were you? What parts did you stay in?

Alex Wilhelm: 79:30 I was in from Midtown to the Lower East Side and back around that area.

Jason Calacanis: 79:33 Yeah, I mean listen, it’s a bit of a crapshoot. Different like neighborhoods in New York could be phenomenally dangerous, different subway lines during some times and other ones would be totally safe. So, and the same thing for San Francisco. If you’re up in the Marina…

Lon Harris: 79:50 In Chicago too.

Jason Calacanis: 79:51 Yeah, if you’re in the Marina or you’re in Pack Whites, you’re gonna be fine. If you’re in the Tenderloin, you’re gonna be murdered.

Alex Wilhelm: 79:56 Did you just say Pack Whites?

Jason Calacanis: 79:57 Yeah, Pack—the joke of Pack Heights is literally if you walk around there it’s all whites.

Alex Wilhelm: 80:03 I used to live there!

Jason Calacanis: 80:04 Yeah, that’s what they called it when I moved there. They’re like ‘oh you should move to Pack Whites.’ I’m like ‘I’m sorry, Pack Heights?’ They’re like ‘no, Pack Whites’ because there’s no crime there. There’s zero crime and if you call the police they show up in 30 seconds whereas if you’re in any other part, you know, they might take a little bit longer.

Alex Wilhelm: 80:18 What? I met the CEO of Zoom, uh, it was at a dinner at a venture capitalist’s house in the fancier part of Pack Heights. So that tracks.

Jason Calacanis: 80:25 The Broadway billionaires row, yeah. I mean, billionaires row you have like a thousand cameras per block. The other areas of town, they’ve banned closed circuit cameras because, you know, privacy. And it’s like trust me, anybody who lives in those areas is willing to give up privacy to be a little bit safer. In Pack Whites you’ve got 17 cameras on the front of every house and like if you like spit out your gum, you’re gonna have 16 different videos of it submitted.

Alex Wilhelm: 80:49 Whereas when I was in a different neighborhood, someone spit in my mouth while I was walking down the street. Not on camera.

Jason Calacanis: 80:56 Yeah, I mean literally you could be shivved. All right. This’s been another great week for TWiST. We’ll see you on Monday. Bye-bye.