More or Less Podcast

SaaS Companies Beware: AI Is The New UI (Anthropic's Claude Code and Cowork)

2463 summary words 11 min summary Watch video

Start with the signal

11 min read

Summary

At-a-Glance

  • Verdict: Watch fully
  • Core thesis: Desktop agents such as Anthropic's Claude Code and Cowork are turning AI into a user-generated interface and control layer over existing systems, eroding standalone SaaS UI and pricing power while making security, trust, data access, and distribution the new strategic moats.
  • Why it matters: The discussion directly connects local agents, browser and file control, model flexibility, SaaS replacement, API economics, security boundaries, and venture strategy—the core issues governing agent control planes and OpenClaw-style systems.
  • Best use: Use it to pressure-test agent architecture and permissions, identify SaaS contracts or workflows vulnerable to internal replacement, and refine investment theses around durable moats beyond software features.

Executive Summary

The hosts argue that AI is becoming a dynamically generated UI over existing products and data. A user can give an agent a device, file, browser session, or poorly documented service and ask it to discover the relevant interfaces, authenticate, process the data, and generate a better front end. Their examples include rebuilding Netgear Mural management software, extracting Tonal data, reading an MRI stored in a legacy format, researching the seed-fund market, and analyzing hundreds of podcast transcripts.

That capability threatens SaaS businesses whose defensibility rests mainly on workflow logic or user experience. The sharpest anecdote is a prospective SaaS contract reportedly falling from roughly $150,000 to $70,000 before the buyer's CTO reproduced the needed functionality with Claude Code in about a week. The speakers contend that merely adding a generic agent to every SaaS product will not restore differentiation; value is more likely to accrue to proprietary data, network effects, community, physical or scientific assets, distribution, and control of the primary agent interface.

The counterweight is security and ecosystem control. A local agent that can operate the browser, read messages, access files, and use the keyboard and mouse is effectively being granted the user's own authority. The hosts expect serious breaches and subsequent consolidation around trusted vendors such as Apple and Google, while platforms may restrict APIs, packages, and app-store access to preserve monetization. They also acknowledge that current tools remain brittle and recreational: setup can require GitHub and environment knowledge, a Garmin watch-face attempt failed, and one host says most of his AI-built utilities are enjoyable toys he rarely uses.

Key Takeaways

  • Claim: A local agent can serve as a generated interface and execution layer over devices, cloud applications, files, and undocumented systems rather than merely answering questions. | Evidence: Dave describes Clodbot as a local system running on a Mac mini, reachable through WhatsApp, Telegram, Signal, or iMessage, with plugins for services such as Eight Sleep, Notion, and Google and access to local projects and multiple models. He says it reverse-engineered Netgear Mural's API and unusual authentication flow and produced a working photo-management web interface in under 20 minutes, with the system usable in under 30 minutes. | Implication: Ken should treat the agent as the emerging control plane and UI, with tools and SaaS applications increasingly becoming resources behind it rather than destinations users open directly. | Caveat: The transcript sometimes refers to the tool inconsistently as Clodbot, Claude Bot, or Claude, so the exact attribution of individual runs is not always clean.
  • Claim: The same permissions that make desktop agents valuable create a root-equivalent security and identity problem. | Evidence: The hosts describe granting screen recording, full file access, browser control, and access to Signal and iMessage. One experiment asked the agent to pull group-chat content and summarize it; another speaker characterizes the system as having the user's keyboard and mouse and being able to perform anything the user can perform. | Implication: For OpenClaw-style deployments, scoped identities, capability boundaries, isolated execution, approval gates, logs, and revocation must be first-class architecture—not controls added after the agent works. | Caveat: The discussion identifies the danger but does not provide a mature permissioning, audit, credential-isolation, or incident-response design.
  • Claim: AI-generated internal software is beginning to weaken SaaS pricing power where the purchased value is mostly UI and straightforward workflow logic. | Evidence: The central commercial anecdote concerns a recognizable but unnamed SaaS vendor offering a package for about $150,000, reducing it to roughly $70,000, and then losing the deal after the prospective customer's CTO reportedly rebuilt the needed functionality with Claude Code in approximately one week. The speakers also cite Netgear Mural's poor application being replaced by a custom interface in minutes. | Implication: Software renewals should increasingly be evaluated against an agent-assisted internal build cost, but the comparison must include ongoing operations and governance rather than prototype speed alone. | Caveat: The contract story is anonymous and unverified, while enterprise compliance, support, reliability, integrations, maintenance, and audit obligations can make replacing production SaaS much harder than building a demonstration.
  • Claim: Platform owners are likely to respond to agent-driven disintermediation by restricting programmatic access and monetizing scarce control points. | Evidence: A Garmin watch-face project became blocked by development-environment and package dependencies controlled through VS Code rather than Cursor, and the eventual Claude attempt also failed because of Windows binaries. The hosts compare the dynamic to RSS feeds being withdrawn after aggregators captured the user relationship and made publisher monetization harder. | Implication: Agent products need an explicit dependency map covering APIs, browser automation, package registries, authentication paths, licensing, and platform policy because access can become the primary business constraint. | Caveat: It remains unresolved whether services such as Asana or Notion will charge agents, benefit from increased usage, negotiate reciprocal access, or simply restrict integrations.
  • Claim: Agentic leverage already extends well beyond coding into research, knowledge synthesis, market mapping, and asynchronous back-office work. | Evidence: One host says Cowork crawled seed-fund websites and extracted positioning, check sizes, and investment categories while she continued working. Dave says he processed 131 podcast transcripts and generated essays from them in under an hour, while Lenny reportedly performed a similar exercise across 250 transcripts. | Implication: Ken should prioritize agent workflows where broad collection and first-pass synthesis are expensive but outputs can be checked against sources, rather than immediately delegating final judgment. | Caveat: The speakers offer little evaluation of factual accuracy, source coverage, data freshness, or the human review required before using these outputs for consequential decisions.
  • Claim: If software features become rapidly copyable, durable company and investment value must come from assets outside the code itself. | Evidence: The proposed founder framework has two poles: quietly exploit a non-obvious, technically leveraged insight that competitors cannot see, or build in an obvious emerging category while loudly owning its narrative. The hosts separately name community, network effects, biology, life sciences, human connection, robotics, hardware, defense, space, materials, and scientific breakthroughs as more defensible areas. | Implication: Investment diligence should separate implementation advantage from durable advantage and require a credible answer for what remains scarce after competitors gain equivalent models and coding agents. | Caveat: The 'secret business' category is difficult to verify and often unsuitable for venture financing because public distribution and fundraising can expose what makes it work.
  • Claim: Current agent creation volume materially overstates durable utility, and incumbents may capture more value than a long tail of new applications. | Evidence: One host says he spends many hours building AI utilities but very little time using them, likening the moment to the iPhone's 'fart app era' and AI building to entertaining Lego play. The Garmin project failed, GitHub remained difficult for an 11-year-old building a Game Boy title, and the panel predicts Google and other trusted incumbents can combine existing user data and distribution more effectively. | Implication: Agent initiatives should be judged on repeated usage, completion reliability, and economic substitution—not demos built, tokens consumed, or novelty engagement. | Caveat: Low retention among today's experiments does not negate the underlying reduction in software-production cost or the eventual emergence of durable workflows.

Detailed Brief

Open-source agent layer versus packaged consumer agent

  • Claims: The speakers portray Anthropic's Cowork as a simpler consumer version of capabilities already available in the open-source Clodbot community.; The local open-source layer is model-flexible: it can route programming work to small or large models rather than being inherently tied to a single provider.; The likely point of concentration is not necessarily model inference but the trusted middleware holding local permissions, credentials, and execution authority.
  • Evidence: Clodbot is described as difficult and 'nerdy' to configure, whereas Cowork exposes packaged connections to services including Asana and Notion.; A suggested mitigation was to install the local agent on an otherwise empty Mac mini, although the panel immediately noted that usefulness still requires access to real accounts and data.
  • Caveats: Hardware isolation alone does not solve identity risk when the machine receives privileged cloud tokens, browser sessions, messaging access, or access to sensitive network resources.; The claim that trust will consolidate around Apple or Google is a strategic prediction, not evidence that their future implementations will be safer by default.
  • Implications: The strategically valuable layer may be a model-neutral, policy-enforcing local broker that owns permissions and auditability while routing inference to replaceable providers.; Consumer convenience and enterprise governability are likely to diverge, creating room for hardened orchestration even if broad desktop-agent functionality becomes commoditized.

Automation's effect on venture sourcing and investment alpha

  • Claims: Routine venture work such as market mapping, outbound sourcing, founder matching, and spreadsheet analysis is expected to become highly efficient and therefore less differentiating.; The panel argues that seed-stage alpha should migrate toward 'illegible' opportunities requiring unusual networks, original judgment, or a willingness to underwrite risks consensus investors avoid.
  • Evidence: Boardy is described as a WhatsApp and email-based marketplace that can source 6,000 founders in a week, recommend matches to VCs, and charge companies $10,000 annually for business-development support.; Boardy reportedly also raised a $150 million fund based partly on LP interest in its sourcing mechanism.; The discussion estimates that roughly 2,000 seed funds plus multistage firms compete over mainstream opportunities, producing near-perfect pricing and 'synthetic beta' in the visible portion of the seed market.
  • Caveats: The Boardy figures and business description are conversational claims and are not independently substantiated in the transcript.; Prior attempts to automate venture investing, including an old Google Ventures effort mentioned by the hosts, did not eliminate the role of human judgment.
  • Implications: Better sourcing automation may increase deal flow without increasing returns if it routes the same legible companies to every investor.; Proprietary interpretation and access become more important as generic research and matching become available to all funds.

Why SaaS erosion may be real but gradual

  • Claims: The strongest version of the panel's argument is not that organizations stop using software, but that software features cease to support premium standalone pricing.; Enterprise replacement will lag prototype capability because incumbent vendors bundle governance, support, reliability, and institutional acceptance.; Adding similar AI agents to every existing SaaS product is unlikely to create meaningful differentiation when the agents perform overlapping tasks.
  • Evidence: Operators previously told the hosts they could not cancel SaaS contracts because audits, compliance, and security requirements still favored established vendors.; The panel points to Constellation Software's share-price decline from its 2025 high and return toward earlier levels as a market signal that investors are questioning the durability of SaaS roll-up economics.; A host summarizes the customer's experience as having five SaaS providers whose agents all do approximately the same thing—usually little, occasionally something useful.
  • Caveats: The transcript's market-price figures are loosely stated, include a corrected year, and should not be treated as precise financial analysis.; The speakers use 'software is dead' rhetorically; their own examples preserve value in software-enabled businesses with network effects, proprietary assets, distribution, or operational accountability.
  • Implications: The relevant strategic question is which part of a software vendor's revenue compensates for scarce capability versus packaged convenience that an agent can recreate.; SaaS compression can coexist with continued use of incumbent systems as systems of record beneath a new agent-controlled interface.

Notable Concepts & Terms

  • AI is the new UI: Users can ask an agent to interpret data or control a system and generate the needed interface on demand, reducing dependence on a vendor's fixed application.
  • Clodbot: The transcript's name for an open-source local agent that connects messaging channels, computer access, plugins, projects, and multiple models; it is the closest analogue to an OpenClaw-style personal control plane.
  • Claude Cowork: Anthropic's more accessible agent experience for operating across connected work applications, characterized by the speakers as a packaged but less powerful version of the open-source approach.
  • Hacking yourself: The panel's shorthand for granting an agent the same broad access a malicious intruder would seek, but doing so voluntarily to automate personal work.
  • Software as a pricing mechanism: The idea that packaged features and UI historically justified recurring SaaS fees; agent-generated substitutes may remove that basis for premium pricing without eliminating software itself.
  • Walled-garden response: Platforms may close APIs, control packages, impose app-store rules, or charge for access once agents begin disintermediating their interfaces and monetization.
  • Secret company versus obvious company: The proposed strategy under rapid copying: either quietly exploit an undisclosed economic insight or openly dominate the narrative in a category everyone knows will exist.
  • Illegible alpha: Investment returns found in opportunities that automated sourcing and consensus filters cannot easily categorize, price, or distribute to every fund.

Operator Notes / Why Ken Should Care

  • Run desktop-agent experiments only inside an isolated environment with dedicated service accounts, short-lived credentials, restricted network access, and no personal messaging or wallet data.
  • Create a capability matrix for file access, browser control, shell execution, messaging, credential use, external communication, and destructive actions; assign approval and logging requirements to each capability.
  • Select three upcoming SaaS renewals and benchmark vendor cost against an agent-assisted internal replacement, including maintenance, security review, uptime, support, and audit burden.
  • Require every agent prototype to report 30-day repeated usage, task success rate, human correction time, and avoided labor or spend before it graduates into production.
  • Map all external dependencies in OpenClaw workflows and maintain fallback paths for APIs, browser automation, package registries, authentication changes, and provider policy shifts.
  • In investment memos, add a mandatory 'post-code-scarcity moat' section covering proprietary data rights, network effects, distribution, physical assets, regulatory position, trust, and switching costs.
  • Monitor whether Apple, Google, Anthropic, and Microsoft introduce local permission brokers, agent identity standards, audit logs, paid connector programs, or restrictions that change control-plane economics.

Source/Metadata

  • Title: SaaS Companies Beware: AI Is The New UI (Anthropic's Claude Code and Cowork)
  • Transcript words: 19517
  • Duration seconds: 3226
  • Timestamp note: No timestamps or chapter markers were provided. The transcript contains speaker labels plus substantial duplicated passages and some inconsistent product-name transcription.
Search this video Resume transcript at 0%
Full transcript 10929 words · 86 min read
0:00

SPEAKER_01

The information we surveyed our subscribers so 67 would not move. You should change it.

0:05

SPEAKER_04

To be like 33 of our subscribers are billionaires. I do believe we have the highest daily active...

0:09

SPEAKER_01

Billionaire readership on the internet. D-A-U-B.

0:13

SPEAKER_00

No or is it yes? We'll debate the text that's best.

0:33

SPEAKER_01

Why hello friends, welcome to More or Less. It's been another week. How's everyone holding up?

0:39

SPEAKER_02

Great. How are you?

0:44

SPEAKER_01

Great. We got a great. It's mid-January, but I feel like this year's already been long so far. I'm ready for the next one. You're ready to roll into 2027? How are you doing, Sam?

0:50

SPEAKER_04

You're wearing a lovely shirt. You're welcome. I actually have two shirts on. I'll show you the other one later, Sam. Lumberjack chic. Oh, is this a surprise reveal thing? Yeah, I have a reveal shirt only for Jessler. For all of us it's got to be a Jackson Hole shirt. I got one for Dave too. It's just for me and Dave. [SPEAKER_01] All right. It's another AI shirt. We need some shirts. All right, we'll make shirts. Sam's wearing a nice shirt and another shirt. Brit's feeling like the year's long, and Dave, how are you doing? [SPEAKER_03] I said great.

1:10

SPEAKER_01

I know, but I need more than great. [SPEAKER_03] They've got a haircut today. Who cares? He's got a hat on.

1:20

SPEAKER_03

Yeah, I've got a hat on. Got the podcast hat. I'm awesome. I'm mind blown by some of the stuff I've built with AI in the last few days. Oh, we're going down that? It's crazy how fast this stuff's moving. We don't have to rabbit hole into this too far, but I will be a witness to this as someone who also has been playing with all the vibe coding stuff but also watching Dave play with it. The productivity it's created for our household with the things he's built has been high. I'm very proud of my husband for thinking about new types of vibe coded apps that make everyone's life better.

1:33

SPEAKER_03

[SPEAKER_01] Okay, now I'm just curious. So Dave, if we look back on this week in tech, there has been big vibe coding news. We've got the continued rise of cloud code. We have Anthropic coming out with co-work, which I have not played with, but I actually feel like I should have taken a day off to play with because everybody's talking about it. So this is a great topic. Let's dive in, Dave. And there's some secret underground stuff going on.

1:38

SPEAKER_02

[SPEAKER_01] Well, now you've perked my ears up. We have to talk about the secret underground stuff, or I would call it just nerdy stuff going on. [SPEAKER_03] Nerdy. Okay, I want to know about this more. Household app. What's it doing? [SPEAKER_03] The interesting nerdy thing going on is that every once in a while in Silicon Valley you think you're living in the future, and then you try something and suddenly it leaps you forward even further in a way that melts your mind. Waymo would be that for me. But yes, there's this open source project called Clodbot. Clodbot's great. I was posting about it the other day. It's pretty cool. Are you using it?

1:55

SPEAKER_02

[SPEAKER_04] Yeah, it's cool. The Clodbot's impressive. It also, I like that when you open it up, it gives you the book needs funding.

2:01

SPEAKER_01

[SPEAKER_04] Well, npm needs funding. npm is always raising. But Clodbot also gives you a prop being like 37 projects looking for funding. [SPEAKER_03] Whatever. It's cute. No, Sam, that's npm. Are you sure? [SPEAKER_04] Yeah, the node package manager is always raising money for all of its various projects. [SPEAKER_03] I know that, but I actually maybe I misread it. I thought when Clawbot opens it also, which I think is clever. Okay, you guys, I need the TLDR on Clodbot, Dave.

2:28

SPEAKER_01

[SPEAKER_03] Okay, so Clodbot's an open source project built by this really talented engineer who's been around for a really long time, big contributor to the open source community. It's a local AI system that you can run on your own Mac mini. Then it has tons of skills and various plugins that you can plug into everything from your Eight Sleep to your Notion account to Google. It can use your own computer to access basically anything on your computer. Then you can use WhatsApp or Telegram or Signal or iMessage or whatever your favorite messaging app is to talk to it all the time. It has really smart and very powerful programming tools built into it. So you can use every single model you want to program. It can obviously use your computer so it can see all of your projects. If you've got a lot of vibe code projects in one folder, it can check out all of those. It can see what you've already been doing there. It can use small models, big models. It's really powerful. It's the first time I've used something where it just leaped forward my AI experience in a way that's been very productive.

2:34

SPEAKER_03

[SPEAKER_01] So this is an open source project? This is not a company? It's not a company. It's an open source project. Can you give me an example of something you've done with it?

2:45

SPEAKER_03

Yeah, so over the weekend I decided to throw a challenge at it that I'd been doing some things like this with cloud code, but I just decided to do it with this instead and see. I actually A/B tested it. I started a cloud code terminal window, a Gemini—or sorry, not Gemini—Codex, which is OpenAI's version of this. Then I started a Clodbot window. I wanted to see if I could reverse engineer the Mural frames, which are the photo frames that a lot of people have on their walls that have the world's worst interface and the world's worst app. [SPEAKER_01] You guys are the first persons. These are the big ones?

2:58

SPEAKER_04

[SPEAKER_03] Yeah, actually speaking of a living in the future moment, walking into your house like eight or nine years ago and seeing those with digital photos was one of these. They've been around a long time. They're the only ones of their size. Netgear owns the company, which is a great company in itself, but has not really prioritized the Mural product and definitely doesn't build great software. It's more known for hardware. So the software is terrible. The app is terrible. You try to upload new photos and everything fails. So we're constantly trying to get new photos on our walls of the stuff we've been doing lately.

3:04

SPEAKER_04

[SPEAKER_01] Okay, back to you, Dave. The app is terrible. You upload photos, it fails for no reason. It doesn't tell you why it failed. You can't really put things into it. There's no feedback. It's always been this way. So you just never change your photos frequently because it's difficult to deal with. So my new favorite thing to try with cloud code or any of these tools is to just point it at the thing and say reverse engineer this and try to discover the API, try to discover the authentication method, build me a new web interface for it that's simple and easy to use. Maybe I describe some features that I want with it. In this case, I want to be able to sort the photos. I want to be able to upload photos and resize them to the correct size that works perfect for the frame. Just stuff like that, right? Easy management interface. So I started this project on all three different tools. Clodbot got it done in less than 20 minutes. I had a brand new web interface. It had reverse engineered not only the API...

3:09

SPEAKER_04

[SPEAKER_03] My new favorite thing to try with Claude or any of these tools is to just point it at the thing and say reverse engineer this and try to discover the API, try to discover the authentication method, build me a new web interface for it that's simple, easy to use. And maybe I describe some features that I want with it. In this case, I want to be able to sort the photos, I want to be able to upload photos and resize them to the correct size that works perfect for the frame, just stuff like that. Right, easy management interface. So I started this project on all three different tools and Claude got it done in less than 20 minutes. I had a brand new web interface. It had reverse engineered not only the API and the ways that people have been doing this, but it turns out that Netgear has not kept up their authentication scheme properly and so you can't authenticate into their API using a normal method. And so it actually figured out how to do it through a really weird thing I'd never even heard of and boom, I was up and running and managing our photos on our photo frame in less than 30 minutes. And then I spent the rest of the day just in my break times just adding little features to it like AI processing of the photos to make a better title, doing all kinds of things that are pretty cool. And now I've got this amazing mural manager project that I open sourced on my GitHub for anybody to use to manage their own mural frames if they want to. And it's mind-blowing to me that you can do a project this simple this quickly and have a better interface for something that you've always wanted.

3:15

SPEAKER_03

[SPEAKER_02] And that's just one example. Didn't you also reverse engineer an API into Tonal at home to do what? That's funny. So I wanted to get the data out of it and Tonal doesn't have an API and there is no documented API. So this brings up...

3:27

SPEAKER_04

Okay, finish this. This is effectively hacking Tonal, which is fine. So this actually brings up the point I would point, which is black hat Dave or white hat? What hat are you? Here's the big problem. I looked in the last day, actually a week ago, Mike Galpert was the one who was saying you got to use Claude Bot and play with it. So I started playing with it and then today I downloaded the Claude desktop agent. Here's my issue: it is sweet. It is a security nightmare and terrifying.

3:32

SPEAKER_01

[SPEAKER_03] Yeah, totally. And I did think it's really funny. The first thing I did was use the interface with iMessage. Fine, I'm like get these files for me and I'm like oh my god, you now have access. If you get access to this it's a disaster. It's a backdoor to everything.

3:37

SPEAKER_03

[SPEAKER_04] It's really funny. I actually did a thing on one project I started working on with Claude and I'm working locally. It's pretty dangerous. It's just open Signal, pull all the group chats. I'm not reading fast enough. Get all the content from them, store it, and then summarize it for me because I can't read. Here's the problem: it's great. It's terrifying that it can do that right. And it's the same thing. You're going through it. It's oh, you want access to screen recording? Okay, Chrome? Okay, this. And then you're like, I need a full system. I need full file access. And I'm like, well, I want to give this to you because I want what I want. But I am absolutely terrified of the security implications of this.

3:42

SPEAKER_03

So it becomes the thing where I'm like, this is actually the really important conversation. And then I'm like, do I actually need this? [SPEAKER_04] Part of me is like, you know what I need? I actually literally need to buy a new laptop and only run Claude Bot. But the problem is then the IT is just really interesting personal security problem because there's so much. Just put it on a Mac Mini that's an empty box.

3:55

SPEAKER_03

[SPEAKER_01] Yeah, but the problem is it needs access to. Okay, I want to slow this down for a sec for our non-engineer listeners for a second because I think this is an important conversation. So basically your point is that these AI bots are able to penetrate any sort of security or encryption on any individual service. It's sort of just on your computer, Jess. So it's just looking at the files on your computer or it's looking at... Right, but my Discord chats aren't on my files.

4:08

SPEAKER_03

[SPEAKER_02] No, it is actually. And same with Co-Work by the way. Just like yesterday I was trying Co-Work, the new thing from Claude, and you can connect Asana and Notion and all kinds of things into this. And Co-Work is actually the consumer version of this. When Co-Work came out, all the people that have been on Claude Bot community for many weeks were just like, oh, this is just the simple consumer version of what we're all doing, but it's not as powerful, but it does the same kind of stuff. It's basically you are literally just hacking yourself.

4:23

SPEAKER_03

[SPEAKER_01] Is what you're doing, Napster? And then we'll all sort this out. What are some analogies? Claude Bot—it's like a lobster claw Claude Bot? No, he's just saying what's the moment here? Here's the deal. Just are they the Napster moment? And then I want to know what's going to happen to the billion-dollar-valued Anthropic. If the open source version of what it does is more powerful and you never have to pay Anthropic— [SPEAKER_02] It still needs Anthropic's models.

4:35

SPEAKER_03

[SPEAKER_04] No, you're pointing to any of them. It's a top-of-my... Well, I mean, here's the upshot. I think when you play this all out, here's what's going to happen. It's actually all just a massive security breach and because of that, actually what's going to matter more than anything is who do you trust, right? And we all trust Apple effectively because they already have this already, right? Yeah, but Sam, they really need. I think they actually have to start thinking about this pretty seriously.

4:48

SPEAKER_03

[SPEAKER_04] No, I agree. But just let me finish. So the thing is, when you're like who's going to win this interface layer, I'm giving Claude access. I'm basically giving myself root with Claude and with full knowledge that there's a whole stack behind there that can do terrible things to me. And then I'm like, okay, well, I do actually trust Google more. So I actually think when you think about the AI wars, which happens a lot, what's going to happen is in order to do good stuff beyond the chatbot, you ask it a question in a chatbot, you give it access to such a level of scary security stuff that the people that win are going to be the people with trusted brands and security and all that type of stuff. And I don't think it's going to make it even harder for startups to compete because look, I basically just rooted myself with Claude bought a bunch of other stuff. And even I'm like oh my god, I'm not giving you root file access. That's too much. I'll give you other stuff. But it's terrifying what you do and don't give access to and then what it can do with speed and automation. So I just don't think it's a place startups are gonna be able to play. And I think it's gonna get locked down super fast because it is super powerful but it is super dangerous.

4:54

SPEAKER_01

So I understand. So that helps me.

4:59

SPEAKER_03

[SPEAKER_04] people that win are going to be the people with trusted brands and security and all that type of stuff and I don't think it's going to make it even harder for startups to compete because look I just rooted myself with Claude bought a bunch of other stuff and even I'm like oh my god I'm not giving you root file access that's too much I'll give you but it's terrifying what you do and don't give access to and then what it can do with speed and automation so it's like I just don't think it's a place startups are going to be able to play and I think it's going to get locked down super fast because it is super powerful but it is super dangerous so I understand so that helps me

5:05

SPEAKER_03

[SPEAKER_01] understand like you think that we're going to have to really trust the people who as AIs get into these operating system level plus agents plus all this you really have to trust them but I have a different question which is the open source version of this you guys say is better so why would anyone use the cloud version because they don't trust the back end of this open source or it's too hard to like it's really nerdy it's too hard it's really difficult like it would it's really difficult to set up you're not going to set it up Jess

5:16

SPEAKER_03

[SPEAKER_04] it's not that difficult but I think the problem the real story is just going to be like what's going to happen at some point is someone's going to come out with one that's great right and it's open and whatever and it's free and then you're going to be and it's going to be totally controlled by some hacker network that's going to do terrible terrible things like everyone's like it's just a disaster and so what's going to happen is there's going to be some blow ups on this and then everyone's going to retreat and be like I can't use this stuff unless it's coming from Google or Apple right and I think that's how it's going to play out now the point about the tokens this goes into the whole how the ecosystem evolves because I actually do trust anthropic but not as much I definitely don't trust OpenAI but they either have the spectrum of trust here's the thing I'll still use the free compute token resources of almost anyone and if someone comes to the better model great that is its own security threat but the middleware that sits on your computer and has all the file access that's critical and so it'll end up being like maybe you have some neo clouds or whatever that do AI requests or specific things that these things route to but who do you trust and who do you have to trust it's like the thing that's sitting on your computer and has full root access is the most scary part that's the

5:23

SPEAKER_03

I guess dystopian take but the optimistic take here is this is a really powerful tool I guess I'm trying to find the middle ground here of like how do we explore what this future looks like I mean Sam I think you're right that we have to get the security question correct the Apples and the Googles have to give us they have to engage in this but I think the reason I'm excited about this and you're excited about this and MSG and everyone I know that's a tinker or hacker is excited about this is that they're really powerful it's a really cool way to think about this and I really want this I want it but

5:29

SPEAKER_03

[SPEAKER_04] here's the problem again I just go back to like yeah here's the thing what's it worth to you it's super cool that you can backdoor your and get data off of your Tonal is it worth all your Bitcoin of course not right and so it's worth like what risks it's asking are you willing you ready you ready I'm going to bring it to Sam topic to get some clicks on the day oh the poor and the poor you ready it's a super hot chick are you willing to have sex without a condom

5:36

SPEAKER_01

[SPEAKER_02] I thought you were going to say something about it using being native in your browser and seeing what you're actually looking at all day Sam [SPEAKER_04] yeah well that I mean that's the mild version of it

5:47

SPEAKER_03

[SPEAKER_02] that's the mild version oh god can someone take over moderating I got nothing it was weird when Claude took over my browser I was feeling that was I haven't had that totally happen before where it's just running a whole window but that's the frontier okay so Dave I want to go deeper into what you are excited to envision being built and I think that that is important it's also of course important to think of the other side of the coin and it's possible to do both but what about the company whose proprietary ecosystem of software is their business like the other thing that comes to mind well like Tonal I mean I guess to the point that maybe if Tonal's expertise is the hardware in this situation but my reaction is that there's swaths and swaths of businesses where some of the moat around their business is this UX and software that are just going to evaporate right this is the

5:53

SPEAKER_02

Napster are part of the analogy just because like Dave just made a new version of Mural's app and at some point so many people are going to be doing this and it's going to be open source are you really going to police all these people like everyone that burned CDs back in the day I don't think so

5:59

SPEAKER_02

[SPEAKER_01] well we did ultimately I mean again that there was iTunes but this is the real thing that I don't know maybe we can add some value to the Silicon Valley conversation about this today which is I heard a story yesterday that a big SaaS app we all know won't say which one had I don't know something like a 150,000 contract out to try to sell their package to another company and the company was like yeah no too high so they lowered the price to 70K or something like that and then the CTO shows up in the meeting and is like we can vibe code this in three weeks using Cloud Code I just let me at it he got it done in a week and then they just let go of the contract and everybody's been talking about that this is a risk but I think the risk is actually here and it's real now that's why SaaS is trading at zero right now in the public markets because everyone gets this what do we do like everyone's saying look it's happening but I don't know that anybody's talking about like what's actually software is not a business it's just not a business

6:03

SPEAKER_02

[SPEAKER_04] I think that's right that's what it is it's very simple and that doesn't mean you can't build businesses and that doesn't mean you don't use software to build businesses you use the written word to build companies but it's just not a thing right it's just a

6:09

SPEAKER_02

[SPEAKER_01] well so I you know in my heart I always have I guess I'm a legacy media disruptor but I like to make sure we air the point of view of these legacy companies who by the way and maybe I'll be at liberty to say more depending on what my lawyers say but these software companies are just scared right now they are running around like chickens with their head cut off and from my point of view in my day job I see it constantly and it's honestly baffling to me and I think it only makes sense in the context in this broader context so they are paranoid and there is a lot of

6:14

SPEAKER_03

[SPEAKER_04] word to build companies but it's just not a thing right. It's just a well, so I you

6:20

SPEAKER_03

[SPEAKER_01] know in my heart I always have a, I guess I'm a legacy media disruptor, but I like to make sure we air the point of view of these legacy companies who, by the way, and maybe I'll be at liberty to say more depending on what my lawyers say, but these software companies are just scared right now. They are running around like chickens with their heads cut off and from my point of view in my day job I see it constantly and it's honestly baffling to me. And I think it only makes sense in this broader context. So they are paranoid and there is a lot of behavior that they're undertaking. In what way? Which software companies are you talking about? Like which realm? You know, I'm not at liberty to state at a podcast in the moment but not companies but what examples of running around like what do you mean? Just think of what my day job is and think of how I would encounter such people. I mean, I think they're telling narratives and they have their narrative and you know they're fighting really hard to preserve their narratives. What narrative do they have? They're like the software is still a business, that their AI is taking them to the moon, instead of what you said, Dave. But right, and now that it's defensible. But to the point I just want to make a point, Dave. So last summer I, any person who worked at a forward-thinking company, I'm like, have you killed your SaaS contracts yet? Like what are you doing? And most of the people at that time who I talked to said no, we're playing with things, but at the end of the day there's compliance, there's security. You can't actually do much on the margin because you might get audited and so and so has all this stuff. So hey, I heard that answer and I was like, wait three months, because I do think a lot of those pieces can be filled in. But there are some modes. So software is dead, but the death will be long. Maybe. Although, Dave, if you're hearing examples of companies that are really shutting down those contracts today, that's obviously fascinating. But that's why they're nervous. And so what are they doing? They're all pivoting to try and be AI first or integrating with other AI providers. But then you talk to any business out there and they're like, I have five SaaS providers who all have agents who all do exactly the same thing, which is mostly nothing but sometimes something. But they all do the same thing and it's not going to be your savior. Adding AI to your SaaS business is not going to save you and suddenly make software a business again.

6:25

SPEAKER_03

Well, that's what they're all doing, right? And so there's no, I mean, but that is kind of fascinating.

6:30

SPEAKER_03

[SPEAKER_04] Look, I think the purest financial explanation of what's going on, if you want to understand, is Constellation Software. You watch Constellation Software recently? So Constellation, for many years, has been my absolute favorite stock in the world, right? It's a perfect growth chart from 2006 to May 2026. It's a parabolic, beautiful curve. But what they basically do is roll up SaaS, right, and make it more profitable. They have come down right from a 16,000x return. I'm sorry, 16,000 growth over this lifetime to 5,000 bucks. It's trading at 3,000 right now, right? And since May of last year, wait, May? Did you say May 26? Did you say May 2020?

6:37

SPEAKER_03

[SPEAKER_04] Uh, 2025.

6:44

SPEAKER_03

[SPEAKER_04] 25, got it. Again, it's trading back to where it was trading in 2023 or something. The basic point is it's a perfect illustration of what's going on in people's actual story on what's happening here. There are a few other things that fit in fact, but that's the story. And then, look, I there are plenty of good businesses to build in the world. That's not going to change at a high level. But this idea that software is defensible or remote or even a thing you can sell, I think we're just at the end of software sales right? That's the way to think about it. And that means you focus on community, you focus on network effects. There are things that are valuable. But there's no such thing as a software company anymore. It's a real change though, Sam.

6:50

SPEAKER_03

[SPEAKER_02] And I mean, I think that's the era we've been part of the internet in, like the last 25 years. It's been, and there's a lot of seed funds out there that only know how to invest in this. And so how should they think about this? [SPEAKER_02] Well, they're just going to go away. [SPEAKER_02] Well, a lot of people's portfolios are going to get really affected by this. That company that thought they could make 150k per partner in SaaS sales, not so great.

7:09

SPEAKER_03

[SPEAKER_04] What I've been telling founders for 2026 is there's only two business types I'm interested in right now that touch into technology. Right? One type is you have a secret about the universe and you should tell no one ever what that secret is, because it is an incredible idea and insight, a way to make money. It's like a secret incantation. And if you know what you're doing, it's a great way to make money with all the technical leverage. But don't ever tell anyone, and your hope is that no one ever finds out. [SPEAKER_04] There are lots of cool businesses being built right now that look like what's the example of that? Yeah, give me an example.

7:27

SPEAKER_03

[SPEAKER_04] No, I'm not saying a stealth one. What's one that already exists that was like this? Coca-Cola? They're all pretty stealth, yeah. I guess they have a formula, but even Coca-Cola, somebody on YouTube last week spent a year reverse engineering it and actually achieved it. Nothing lasts forever, but there are businesses. I keep hearing of people, I have a business like that, but it's not based on one of those. It's based on consistently generating those things that aren't known that are important. [SPEAKER_01] Sure, but people know your business.

7:40

SPEAKER_03

[SPEAKER_04] These are companies, or it's like someone has a brilliant insight about a way to make money using a bunch of AI tools, and they're minting cash. But they will never tell anyone what they're doing, because it's infinitely copyable. Everything's infinitely copyable, right? That's option one. Option two is there are these things that everyone agrees. So there are none of these in option one. There are, this is a new phenomenon that there's no such thing as an example that's big that you can point to. Probably are, but in some ways you don't know what they're like. There are a lot of businesses in America that make a ton of money that you've never heard of and will never find, right? Like these exist, right?

7:46

SPEAKER_03

[SPEAKER_04] Prevents the copying at launch. Like the second it launches, no one, because you don't even know, there's no website for these things, right? These are literally business people are minting cash off of very quietly using more and more technical leverage. [SPEAKER_02] Are you saying these are not venture types of businesses or just some of them might need a little capital? Some of them need none.

7:59

SPEAKER_03

[SPEAKER_04] Mostly not, yeah. It's mostly, and they're with AI leverage and tech leverage, you can now build things and connect things together with an insight. We're like, oh my God, I can get this type of data here and sell it there. I can do this here and do this there. I can make this. There are all sorts of. [SPEAKER_01] prevents the copying at launch like the second it launches no one because you don't even know there's

8:11

SPEAKER_02

[SPEAKER_04] no there's no website for these things right these are literally business people are minting cash off of very quietly using more and more technical leverage are you saying these are not these are

8:17

SPEAKER_03

[SPEAKER_02] venture types of businesses or just some of them might need a little capital some of them need none [SPEAKER_04] mostly not yeah it's mostly and they're with ai leverage and tech leverage you can now build things and connect things together with an insight we're like oh my god i can get this type of data here and sell it there i can do this here and do this here i can make this there are all sorts of things going on with tons of leverage that small teams are doing they're making fortunes that's type one there are those things that exist it's just they're really hard to talk about and find and the ones i know of you can't say on a podcast

8:30

SPEAKER_04

[SPEAKER_03] yeah it sounds like they're not venture investable so is that helpful to i guess i'm trying to find lanes to help the audience help our audience forget venture capital a lot of entrepreneurs it's go find a secret to the universe you can do this with maybe you need a little money maybe you don't it doesn't really matter type two which i really do believe in is there are all these businesses out there which you can say will obviously be categories right everyone knows what they are we all know that ads are gonna get hyper hyper personalized right there's no question about that so then the other thing is the only way to win is you must own the narrative very loudly i'll give you an example that we can talk about you know i love meme lord like this memetic warfare platform it's i would argue everyone understands if you're smart that the future of communication and marketing whatever is mimetic right and there's an argument that you should be able to build the defining platform at scale in ai mimetic warfare right you can make that argument you might disagree with it but it's an argument you have to fully own that narrative as i am the meme warfare company you will buy it from me right and so i think there's this big schism of you either are a secret company or you're the most obvious company in the world but you leave the narrative you can't be in the middle britt what are you looking for if it's not

8:34

SPEAKER_04

[SPEAKER_02] software literally i mean anything that's not software bio life sciences human connection robotics hardware defense space categories where you can't just replicate this stuff overnight or to sam's point it's almost his first point there's a novel discovery there's a breakthrough in science and you know a lot of these researchers or scientists have a discovery but they don't know how to build a business around it and commercialize it and so i'm leaning into that i have a trump market though i mean that's always been the business of drug discovery i mean i'm not necessarily talking about drugs i mean it could be any kind of science yeah any kind of science materials there's all kinds of breakthroughs happening right now historically i'm too lazy to do that stuff it's too hard to diligence now you could make the argument that with ai it becomes easier for generalists to diligence all this stuff and so you can have marginal leverage on it or be in that industry maybe that's good maybe that dave's point about what's good in all this is there's all this that should get funded that generalists can't figure out but if you give them all the right ai tools all of a sudden they can have productive conversations and make realistic decisions with leverage on these harder areas and so you make more progress and fund more of the weird stuff i don't do that stuff because i think it's too hard can i that brought up

8:40

SPEAKER_04

[SPEAKER_01] i wanted to circle back we were talking a little bit of open ai health and i want to circle back to ai and health and some of the startups and opportunities there because i have a big question called business model but dave to your i liked how you decided that our goal was to try and add something to the silicon valley conversation it's a very high bar so have we done that yet on this topic sam i think you one of your most viral tweets of 2025 was about software being dead or something and then you know you got some pushback from that from the software people but it does seem like we've entered a new phase right there's actual examples of contracts being obliterated we've also raised the wild west of data security concern but if software is dead for entrepreneurs then it's also dead for venture capitalists and that has there's this halo effect to that statement being true pretty dead for venture capitalists for i mean talk to any firm about their stats on their books they've been complaining about this for two years

8:45

SPEAKER_04

[SPEAKER_03] i don't know yeah they've been complaining about it but you look at a lot of people's yeah their portfolios and there's a lot of software in the last six months right

8:49

SPEAKER_04

[SPEAKER_02] no but even new companies people are still investing in software yeah just what i'm saying is that if you're investing in something now or you're an entrepreneur starting something you're shooting a rocket that lands five years from now right and so five years from now this is gonna be crazy you know ai is the new ui you don't even need a ui for anything anymore you can just i wanted a new ui for my mural frames and i just instantiated it in 10 minutes that's crazy yeah how far so two quick follow-up questions one how far are we from jessica lesson being able to do that pretty close yeah really close six months six weeks six days of hard work six months okay yeah good to know you could probably do it with claude co-work actually yeah i'm not sure i haven't played with it i've played with claude co-work i did a big project yesterday i mapped the entire seed stage vc market i curled everyone's website and pulled in everyone's key differentiator statement how much their check sizes are categories they invest in all these different things about them and it was just done for me while i was working

8:56

SPEAKER_04

[SPEAKER_03] here's another good example i took all of our transcripts we have 131 transcripts for this podcast and i ripped through all of them and generated essays across 131 transcripts and did that in less than an hour and i saw lenny you know lenny has a great podcast he was doing the same thing yesterday and he took 250 transcripts and did the same thing and then he actually today i saw open source them all so that other people can do stuff with his transcripts right and so these are just massively leveraged ai moves that you can do using this stuff and it doesn't have to necessarily be code and the answer to your question just how soon will you do it i think actually is best answered through our 11 year old who over this weekend he wanted to build a nintendo game he had an idea for a nintendo game for his game boy and so he used cloud code to develop a new nintendo game and had it done in a day

9:02

SPEAKER_04

i look i think that's not true but i'll give you let me give you the counter sam just one last thing which is that it was very hard for him to understand github and there's all these

9:07

SPEAKER_04

[SPEAKER_03] them all so that other people can do stuff with his transcripts right and so these are just massively leveraged AI moves that you can do using this stuff and it doesn't have to necessarily be code and the answer to your question just like how soon will you do it I think actually is best answered through our 11 year old who over this weekend he wanted to build a Nintendo game. He had an idea for a Nintendo game for his Game Boy and so he used Cloud Code to develop a new Nintendo game and had it done like in a day. I look, I think that's not true but I'll give you, let me give you the counter Sam.

9:18

SPEAKER_04

[SPEAKER_03] One last thing which is that it was very hard for him to understand GitHub and there's all these intricate things that are very nerdy about how to interact with these tools but they're not that complicated and will just get eliminated from the process and so you know it's got to be within six months that you can just deploy something onto Vercel and have it in 10 minutes.

9:25

SPEAKER_04

I do think there's going to be weird things that happen. I'll give you my counter example Dave, which is our kids are now deep into Garmin watches and I was like they're messing with watch faces. All the watch faces kind of suck but you know what, this is a perfect AI thing. I would never want to learn their stupid Garmin language to program a watch face but AI knows this. I was like let's just pop open Cursor, we're going to make our own Garmin watch faces right? Seems like an obvious perfect use case for this, all the things you're talking about. You go and you're like oh wait a minute, interesting business thing. Cursor split the store from VS Code because they want to create their own ecosystem that's independent, etc. And the way it all plays out is VS Code is like no, the package you need to install to do Garmin watch faces, we're going to control that. All right, and so all of a sudden you realize it gets pretty hard to actually do it without VS Code so now you're using VS Code instead of Cursor even though it's the same thing right?

9:30

SPEAKER_04

Did you try Cloud Code?

9:37

SPEAKER_04

I haven't tried yet. I don't think it'll work. I'll just do it. I don't, I'll try it. I actually kind of doubt it'll work but we'll see because they have a fork. Because you go back to, you basically get stuck in an app store problem with Microsoft who has access to the packages and then here's the thing: VS Code and their AI, which is basically all the same stuff anyway, just spins for hours trying to get a dev environment set up. It kind of knows what it's doing but it's just not quite there. Now does that all get resolved? Maybe. But I do think there's going to be all sorts of weird business development games that start getting played with lockdown app stores, lockdown code bases, who has access to what. All of a sudden Garmin is going to be like hey guess what, we had a whole business around watch faces. We're not going to just let you push. Now that it's so easy to make them, we're not going to let you just push them up. We're going to change our business model around them. Like the world will reorient to try to extract value whether it's app stores or whether it's access points or whatever. It will be a game of chicken and mouse on this stuff.

9:42

SPEAKER_04

[SPEAKER_02] Do you think they're going to just start charging for that? Like with Claude co-work, I was connecting all these things into it yesterday. At some point is Asana going to get any additional revenue from Claude because millions of people are integrating with it or is that just net value add to Asana for more use? [SPEAKER_02] It's because it's actually depleting Asana right? Because it's just writing on top of it and pulling scraping it. [SPEAKER_01] The good news is I believe Dustin's a very large Anthropic investor so he'll be just fine.

9:59

SPEAKER_03

[SPEAKER_02] Okay, Notion for instance as another example.

10:03

SPEAKER_04

[SPEAKER_01] Well no, this is the wild garden thing right, and this gets back to when I was saying the SaaS companies are just trying to embed AI as fast as possible but there's a new race to be the default, to be the UX, and I think it's going to be very hard for a single point of SaaS software. I'm just saying there's no business model that exists so far in terms of who's getting paid to be the connector and all these things. I remember talking to Erin Levy about Box about this and you know I think there's expectations that there would be some kind of reciprocal tariff structure or something like that but we're also seeing this play out in shopping and I don't think that the right incentives for that to happen.

10:08

SPEAKER_04

[SPEAKER_01] Last thing on this point. We'll leave us a little time to get to something not non-AI but did you guys see Toby, the CEO of Shopify, also tweeted something very much on this theme which was he took an MRI, I believe, and had Claude Code or some AI coding software just build him a better web interface for it instead of using the legacy.

10:14

SPEAKER_03

[SPEAKER_02] I was super excited about that because I'm getting two MRIs tomorrow so I'm going to be uploading those.

10:19

SPEAKER_01

[SPEAKER_03] Well yeah, he had a USB key or something that was Windows software that he doesn't have Windows and so he's just, I mean this is the story that AI is the new UI. You don't have a UI to read the thing so you just tell AI figure out how to talk to this thing that I have and make me an interface for it. Boom, it's done right? The number of examples of this this week have been insane and this is all possible just because of APIs. Like no one locked this down in a way or it's just that the way the AI operates it doesn't need a certain type. It's just using some kind of predictive system to be able to do that.

10:24

SPEAKER_01

No, it runs through the whole internet. It goes and looks at the internet. It looks at every example of everybody that's ever tried this. It looks at every forum and every open source project and on and on. [SPEAKER_03] But how does it get the underlying data? You've given it the underlying data. Like you've given it the MRI. It's reading or the, yeah, that's on the disc right? There's a USB key that has the MRI on it but you can't read the file format because it requires Windows and some legacy software and so it just digs into there and figures out the data and pulls it into a web interface.

10:36

SPEAKER_03

[SPEAKER_04] Here's the deal. Forever, this is actually a rehash of an old argument that existed in the early days of Meta and whatever else about APIs in general and what you can and can't have access to. There's no. The only rehash, the rehash is this is just a lot faster and more scaled but it's the same story of like oh you remember there used to be RSS feeds for everything and those got shut off. Why do they get shut off? Because once you built readers on top of them, no one can make any money. And so it doesn't work and so the upshot is like in some ways all that AI is doing is rationalizing effectively the world of APIs and the world of programmatic access because when you drop access where it's not just really sophisticated smart people who can kind of do stuff and you're like ah this doesn't quite make sense business wise but it's fine, you just turn the dial to 11 and

10:44

SPEAKER_03

[SPEAKER_04] To there's no, the only rehash is this is just a lot faster and more scaled, but it's the same story of like, oh, you remember there used to be RSS feeds for everything and those got shut off? Why do they get shut off? Because once you built readers on top of them, no one can make any money. And so, well, this doesn't work. And so the upshot is, in some ways, all that AI is doing is rationalizing effectively the world of APIs and the world of programmatic access. Because when you drop access where it's not just really sophisticated smart people who can do stuff and you're like, "Ah, this doesn't quite make sense business-wise, but it's fine," you just turn the dial to 11 and everyone can do it. You're like, "Oh yeah, we can't have APIs to anything." And so to Jess's point, everything's a walled garden, everything's locked down, everything's a war, right? Everything's a fiefdom. And then you get back to this thing which is it's super hard to have a fiefdom and even have access to data when I just sit something on my desktop and give it permission to everything and let it sit there as an end agent and root it, right? So what do you do with that, right? Well, it's going to be a trust thing and a security thing, right? That's the upshot, okay?

10:49

SPEAKER_02

[SPEAKER_01] So I just have so many questions and I am the voice of the everyman here, but I understand that everything your desktop has a lot of stuff, but my files are in the cloud, they're not on my desktop. My photos are in the cloud, they're not... [SPEAKER_03] In the cloud, yeah, they can open the cloud, it can open a browser, open the cloud, literally.

11:02

SPEAKER_03

[SPEAKER_02] It ran my whole browser yesterday. It was crazy. Running your browser, it's running your computer. [SPEAKER_04] Imagine it just, literally it has your keyboard and mouse. Anything you can do, it can do. Hence the security problem. [SPEAKER_01] Is this what OpenAI's browser wanted to do? Anyone using that lately? I tried it and was not impressed. It doesn't really do anything. So what's on your next to build list?

11:19

SPEAKER_03

[SPEAKER_02] That's a good question. I'm mostly doing stuff internally for Offline right now and I'm like, we're working on what hires do we actually need to make and what things could we just build. And I don't know if you guys have been following this whole Boardy thing. Do you know Boardy? Jess has information written about Boardy yet?

11:24

SPEAKER_01

Sure, I don't. Tell me.

11:30

SPEAKER_03

[SPEAKER_02] Boardy is a WhatsApp and email client that is a marketplace of founders and VCs. And so it's trying to map, like in one week it will source 6,000 founders across its marketplace, figure out the types of VCs that it should introduce them to, ask the VC like, "Hey, I got these five founders, are you interested in any of them? Here's what they're working on." I say yes or no.

11:36

SPEAKER_01

[SPEAKER_03] Companies are paying it ten thousand dollars a year to do BD on it. On their companies themselves are the ones paying. [SPEAKER_02] And then on top of that, it raised a 150 million dollar fund because LPs think this is magical and this is a new mechanism for sourcing. You guys, isn't this just slop? What about the AI slop? It's unclear. This is founder slop.

11:53

SPEAKER_04

[SPEAKER_02] Well, I'll tell you, doesn't it look, if it's perfectly efficient, you'll never make any money. I think the notion that I'm saying, I'm talking about it, is it like, are all venture funds going to be building their own versions of this? Tim Draper launched this thing where you can talk to Tim AI, and it's a year ago. It's just you can go to my website and do it. Is this what the future of matching with founders pitching is? Are there any secrets anymore? Is this all just a marketplace? And, what are the internal tools everyone's building at their venture funds to do all this stuff and experiment with all this stuff? I'm so curious how much time people are spending, how much they're paying or not paying.

11:57

SPEAKER_04

[SPEAKER_01] How hiring plans have changed? Like I'm old enough to remember when Google Ventures was going to automate VC investing. I mean, do you remember this? It's probably like 2006 and it was all going to be an algorithm or whatever, and then it wasn't.

12:02

SPEAKER_04

Look, in the end of the day I think that with seed stage investing, it is the most defensible part of the entire investing world. Everything should get more efficient. ETFs, algorithmic trading, does that come down market? Of course it comes down market and it expands. But any idea that a bot can route to the six right firms and everyone bids and everyone top ticks it, but no one will make any money, which is fine. That's actually what the whole point is. That's why capitalism works, is it creates efficiency, right? So that's fine, but that's not what we do. I'll give you an example. Candidly, we don't do outbound sourcing at all. Like, there are companies, there are venture firms inside, wherever they're built, machines for outbound and market mapping, all stuff. I don't do any of that. We say it on the internet and then people email me, right? And I only want to talk to a certain percentage of them. We don't. And so the idea I think the answer is for a lot of types of investing, yeah, all this data, if you're doing something an associate can do in Excel, that's going to be overly efficient and you'll have no margin in it, but that's not really what we do. I don't think it's what you guys do either.

12:08

SPEAKER_04

[SPEAKER_03] Well, Sam, what I think is interesting is like in the seed market, you have the multi-stage firms and 2,000 seed funds competing for the 75th percentile of all of those, just like you said, the mainstream stuff out there in the market. And so that effectively means you have almost perfect pricing and a bunch of synthetic beta in the top quartile of the seed market. And so where in the seed market is there alpha is a question. And it's, you know, I think you and I have always agreed it's in the illegible spaces. But the question is, how to stay there? Ever more you have to be smarter, more original, have access to networks that everyone else doesn't have access to, ideas other people don't, be willing to take risks other people don't, all that stuff, you know?

12:14

SPEAKER_03

[SPEAKER_02] But I think that at least 80% of other funds, seed and beyond, don't think that way. And they're like, they're gonna, they want to build all the stuff to automate a bunch of things. And they're on the AI train. And so I'm really interested to see where this goes this year given by coding has definitely peaked.

12:18

SPEAKER_04

[SPEAKER_01] Each of you, were you more optimistic about seed investing six months ago or today or less optimistic? [SPEAKER_03] Same. Same. Yeah, I'm probably neutral to six months ago. I don't think anything's even, I honestly, despite the fact this stuff is cool, we try to live a few months ahead of reality, right? I'm not that surprised by anything we're seeing.

12:32

SPEAKER_04

[SPEAKER_03] Yeah, I agree with that. That's why I'm saying I'm neutral to it because I feel like we've been talking about it on this pod. And in fact, I've looked at all of our transcripts in many different ways now. We have been talking about this on this pod. So the point is, I just don't find anything we're seeing surprising right now, right? So it's hard to say that anyone's opinions have changed. They're just the same. It's just playing out in real time as we were talking about that it would. [SPEAKER_02] You guys, would it...

12:42

SPEAKER_04

I don't think anything's even cool. I honestly, despite the fact this stuff is cool, again, we try to live a few months ahead of reality, right? I'm not that surprised by anything we're seeing. [SPEAKER_03] Yeah, I agree with that. That's why I'm saying I'm neutral to it because I feel like we've been talking about it on this pod. In fact, I've looked at all of our transcripts in many different ways now. We have been talking about this on this pod. So the point is, I just don't find anything we're seeing surprising right now, right? So it's hard to say that anyone's opinions have changed. They're just the same.

12:53

SPEAKER_04

[SPEAKER_02] It's just playing out in real time as we were talking about that it would. [SPEAKER_01] That would have been a great sound bite. No one gave it to me. So now we'll move on. So what else is happening? [SPEAKER_02] The Golden Globes. Nikki Glazer. I thought it was a worse year than last year. However, the number of people on GLP-1s year over year definitely increased at the Golden Globes. And by the way, a fun stat: grocery sales are down 5.3 percent since the rise of GLP-1s and eight percent among wealthy and affluent people. So people are eating less.

13:05

SPEAKER_04

[SPEAKER_01] But they're also cooking less. I would just want to point out groceries would not be perfectly equated to eating or cooking. Sam, how are you doing having not eaten for four days? I'm fine. You didn't wait. You did do your ProLon thing? [SPEAKER_01] ProLon. I love ProLon. This is your chance to do an ad for ProLon. Tell the people. ProLon's great. It's just you do the modified fast thing.

13:20

SPEAKER_01

[SPEAKER_04] Your stress level goes back to zero. I slept. I got a 97 of my sleep score last night. I mean, it's not something you want to do all the time, but it's great to do. So Sam's mood still up on day four or five of the fast?

13:32

SPEAKER_01

[SPEAKER_02] I tried it, but I'm good without ProLon. Oh, and everyone's moving to Texas. My home state. Your hometown. Welcome to Austin, everyone. I'm happy to tell you where the best case is. Yeah, so with the wealth tax that we talked about last week on the pod, the Californians are all moving away. They're not going to New York. Some are going to Miami. The new thing that you have to move to—you have to move to Austin before Series B. Is that the new path for founders? That's the new talking point. But when we—so the information we surveyed our subscribers, which is always fun because it's a specific crew to say the least. So 67 would not move, right?

13:37

SPEAKER_01

[SPEAKER_04] So wealth tax and all that kind of—that means there are many other. That's the percentage of your subscribers that are billionaires.

13:43

SPEAKER_03

[SPEAKER_01] It's possible that those are. Yeah, I cannot verify people's net worth. You should change it. It'd be like 33 of our subscribers are billionaires. I know. I actually, that's really funny. I do believe we have the highest daily active billionaire readership on the internet. And the way I would frame it, which I'm sure is true, is you have the highest percentage of American wealth that reads you every day, right? Forget the billionaire thing because that's just usually. If you just take the number, the net worth of our readership exceeds any other publication in the world by a mile. That's what I would pitch.

13:48

SPEAKER_04

[SPEAKER_01] Great. These things are all true. Deeply true. But okay, so Nevada. But do they respond to the survey? Texas and Florida edged out Nevada. And poor New York. Why would you think they would go to New York? They have the same problem.

13:52

SPEAKER_04

[SPEAKER_03] Yeah, there's a tax problem. Yeah. Why would you run to the socialists? You know, they're going to take your money too. Because it's New York City. I know. I mean, I pointed all of that out in my tweet. I know people aren't racing towards Montana, but a mere two percent, guys. It's New York City where I would live if I didn't live in California. I mean, it's amazing. But you're an East Coast person. You're an East coaster. And a lot of people, you know, they have a wealth tax on the ballot. I'm still three percent. I wasn't expecting twenty percent, but it's pretty. To Sam's point, it's entirely politics, right? Florida and Texas are leading the charge for the free world. That's like the guys. Is there real estate arbitrage we should be considering for Austin right now?

13:57

SPEAKER_04

This has been my thing. I've been very excited for the mere millionaires who are going to get to buy all the billionaire real estate in California. I know. But should we also buy and then flip stuff for the billionaires to buy in Austin? I would like to get both sides of the market. [SPEAKER_03] I think that trade's already been taken. That trade's over. Also, the stuff they want is really, really, really expensive. Yeah. No, there's something like 240 lots on Lake Austin. They want one of those. Those are the lots that everybody wants.

14:06

SPEAKER_04

[SPEAKER_01] Okay, well time will tell. But yeah, this week also saw the rise of that meme. I mean, what's interesting to me now is their attempts to organize the tech community to fight this more formally. I think Gary Tan at YC has been very vocal about that. But yeah, it's you know, start your company. Won't that backfire? It could. I mean, it can all backfire. We just no one knows anything yet. But how much truth was it? I also saw that there's an effort to create a new city in Greenland, right?

14:14

SPEAKER_04

[SPEAKER_01] Sure. Everything is possible. So we had the Golden Globe. What's happening in the world? Yes. The 49ers. I know. I mean, it was amazing. I was on a plane for part of the game. We're going to win the Super Bowl. But Kittle's out. You think they're going to win the Super Bowl, Dave? I mean, we lost Kittle. By the way, Dave, Claude Code failed at making a Garmin watch face. I just tried it.

14:24

SPEAKER_04

[SPEAKER_01] Really? Claude Bot. Windows binaries. Be damned. Let's end on this note. What percent? So Claude Code. Claude Codework are amazing. You can squint and see the future. What percentage of things you try fail? Eighty percent? No, no. It depends on what is about inventive being. I just. The real problem I have honestly. Okay, now Sam's about to disrobe. I'm going to reveal my shirt. But before I do, at the end. This. Sam is starting to unbutton himself. If anyone is not watching this, here's the thing that I'm actually curious about. This stuff I like. I love tinkering. I build projects all the time. I don't use any of the things I build.

14:29

SPEAKER_04

[SPEAKER_03] Really, right? And so maybe that's the problem? No. No. So here's the thing. I am not convinced. I don't use Letter. Meme. Letter is the only thing I do use. But I've spent more. That's more than a vibe-coded project at this point. Letter. Name's great, and it's only going to get better. But that's more than a vibe-coded thing at this point. But all the stuff I build, all the little tools and shims, and I've spent weeks building cool stuff, and it's very, very fun. I love our line from the summer: the real thing is AI does replace Instagram because it is literally more entertaining to build toy. It's legos.

14:39

SPEAKER_04

[SPEAKER_03] Really right. And so maybe that's the problem. No, no. So here's the thing: I am not convinced. I don't use letter. Meme the letter is the only thing I do use, but I've spent more. That's more than... [SPEAKER_03] A vibe-coded project at this point. Letter means the main thing you've built that I use. Letter name's...

14:55

SPEAKER_04

Great. And it's only gonna get better, but that's not that's more than a vibe-coded thing at this point. But all the stuff I build, little tools and shims, and I've spent weeks building cool stuff and it's very fun. I love our line from the summer that the real thing is AI does replace Instagram because it is literally more entertaining to build. It's toy. It's legos. It's super fun. It was Netflix. I still like my Netflix. It's really legos and legos are really fun. I don't actually think that almost any of this stuff is actually that useful. Right. And so for me I think the reality is this is gonna be the thing. Take cloud code or any of these things. It's you're gonna build a lot of toys. What percentage of them or what do you actually use? I don't know. I think it's gonna be pretty few and far between. I think that would be an interesting thing where tons and tons of projects get built but almost nothing's useful.

15:00

SPEAKER_03

Maybe you just need better ideas. [SPEAKER_04] I have pretty good ideas, Dave. That's not the issue. It's more just I don't know what is a good idea. Good idea gets used right. I mean by this way I've built, I've spent tons of hours building fun AI things for myself. I've spent extremely few hours using it right. [SPEAKER_04] And I don't think that's because I'm stupid or building the wrong stuff. I think that most people just don't actually know what to build with it or what to do. I mean I'll quote my Slow Ventures newly appointed partner Yoni on this who was, "This is cool but I have no idea what to build."

15:24

SPEAKER_03

[SPEAKER_01] Brings up for me an interesting point which is obviously these tools are still disruptive if you just think about the future of software which we're talking about. But I think the other narrative which we can get into in coming weeks and have continued to talk about is how quickly now the big companies are also rolling out a lot of some useful, some not useful tools. And Google, unencumbered by the DOJ anymore, is basically going big. Google and tying together all your data and all the things you actually really want as a Google user. And that will continue too. And so I think that also explains the tinkering. But that doesn't make what's possible with the tinkering less significant when you think about that playing out across the whole industry.

15:29

SPEAKER_03

[SPEAKER_04] I don't know. But I look, it's just we're in the fart app era of the iPhone. We're building all the cool stuff. Most of it goes away. It's fun though.

15:35

SPEAKER_04

[SPEAKER_01] Yeah, but the fact that those tools existed created and destroyed a lot of businesses. Not as many as you think. The app store did. [SPEAKER_01] What did it destroy? The iPhone made really big business. The taxi business. [SPEAKER_01] Taxi business is probably the best example of what the iPhone did destroy. I mean the news business because of all the ways you could entertain yourself on your phone.

16:06

SPEAKER_02

[SPEAKER_04] I mean, no. It was already gonna be destroyed by the web though. I think this is the thing. What the iPhone did is just more Internet and it made the big companies bigger as we've been saying. I really believe that, you know, in terms of no new information, in general what AI will do is make big companies bigger. I think it totally takes away software as a pricing mechanism which means a bunch of software companies go away. But when you really scope out, I think it's going to be less. I mean, I think it is socially deeply disruptive. I think it's going to screw up our government and how we think about value and employment and also there's a lot of things that we'll do. But in terms of what applications I use day to day, I'm not convinced that AI is going to dramatically change my app stack personally.

16:11

SPEAKER_04

[SPEAKER_01] I wasn't arguing that it was. But okay, so everyone can have a listen to this episode and reach their own conclusions around what we spent the bulk of our time talking about. Whether it matters or not, I think it matters. But not necessarily it can matter by making the big guys bigger. But I also think it's going to have deep implications for SaaS and the whole enterprise software.

16:15

SPEAKER_02

[SPEAKER_04] SaaS is clearly going away. I agree. [SPEAKER_01] I just don't think. I don't really agree on that. I don't really use SaaS. Many businesses do.

16:29

SPEAKER_01

Okay, friends, it's been a treat. Thank you to all our listeners, viewers, followers, subscribers, commenters. I won't be here next week but the gang will be. So stay tuned and we'll be back next week and I'll see you all soon. Bye bye. [SPEAKER_02] If you enjoyed this show please leave us a virtual high five by rating it and reviewing it on Apple Podcasts, Spotify, YouTube, or wherever you get your podcasts. Find more information about each episode in the show notes and follow us on social media by searching for @moreless, @davemorin, @lessonon, and @jlesson. And as for me, I'm @britseeyouguys next time.

16:40

SPEAKER_01

course important to think of the other side of the coin and it's possible to do both but what about the company whose proprietary ecosystem of software is their business like the other thing that comes to mind well like tonal i mean i guess to the point that maybe if tonal's expertise is the hardware in this situation but like my reaction is that there's like swaths and swaths of businesses where some of the mode around their business is this ux and software that are just going to be a bird right this is the

17:10

SPEAKER_02

naps are part of the analogy just because like dave just made a new version of murals app and like at some point so many people are gonna be doing this and it's gonna be open source are you really gonna police all these people like everyone that burned cds back in the day i don't think so well we did

17:26

SPEAKER_01

ultimately i mean again that that there was itunes but this is the real thing that i don't know maybe

17:31

SPEAKER_03

we can add some value to the silicon valley conversation about this today which is i heard a story yesterday that a big sas app we all know won't say which one had i don't know something like a 150 000 contract out um to try to sell their package to a another you know another company and the company was like yeah no too high so they lowered the price to 70k or something like that and then the cto shows up in the meeting and is like we can like vibe code this in three weeks using cloud code um just let me at it he got it done in like a week and then they just like let go of the contract and

18:06

SPEAKER_03

everybody's kind of been talking about that this is a risk but i think the risk is actually here and

18:11

SPEAKER_04

it's real now that's why sas is trading at zero right now in like the public markets because everyone

18:16

SPEAKER_03

gets this what do we do like everyone's saying like look it's happening but like i don't know that anybody's talking about like what's actually software is not a business it's just not a business

18:26

SPEAKER_04

i think that's right that's what it is it's very simple and like that doesn't mean you can't build businesses and that doesn't mean you don't use software to build businesses you use the written word to build companies but like it's just like not a thing right like it's just a well so i i i you

18:42

SPEAKER_01

know in my heart i always have a i guess i'm a legacy media disruptor but i like i like to make sure we air the the point of view of these legacy companies who by the way and maybe i'll be at liberty to say more depending on what my lawyers say but these software companies are just scared shitless right now they are running around like chickens with their head cut off and from my point of view in my day job i see it constantly and it's honestly baffling to me and i think it only makes sense in the context in this broader context so they are they are paranoid and there is a lot of

19:16

SPEAKER_01

behavior that they're undertaking in what way which software companies are you talking like which realm you know i'm not at liberty to stay at a podcast in the moment but not companies but like what what's examples of running around like what do you mean um just think of what my day job is and think of how i would encounter such people i mean i i think they're they're they're telling narratives and they have their narrative and you know they're fighting really hard to preserve their narratives what

19:44

SPEAKER_03

narrative they have they're like then that the that software is still a business that their ai is

19:49

SPEAKER_01

taking them to the moon instead of what you said dave but right and now that it's defensible but to the point i just want to make a point dave though so i last summer i i any person who worked at like a forward-thinking company i'm like have you killed your sas contracts yet like what are you doing you know and most of the people at that time who i talked to said no we're playing with things but like at the end of the day there's compliance there's security like you can't actually you know you can't do that much on the margin because like you know you might get audited and so and so has all

20:22

SPEAKER_01

this stuff so hey i heard that answer and i was like wait three months because i do think a lot of those pieces can be filled in but but there are like some modes so software is dead but the the death will be long maybe although dave if you're hearing examples of companies that are really shutting down those contracts today that's obviously fascinating but they're but that's why they're nervous and so what are they doing they're all pivoting to try and be ai first or integrating with other ai providers but then you talk to any business out there and they're like i have five sas

20:53

SPEAKER_01

providers who all have agents who all do exactly the same thing which is mostly nothing but sometimes

20:59

SPEAKER_03

something but they all do the same thing and it's not going to be your savior like adding ai to your sas business is not going to save you and suddenly make software a business again well that's what

21:10

SPEAKER_01

they're all doing right and so there's no i mean but that is kind of fascinating look i think the

21:15

SPEAKER_04

purest financial explanation of what's going on if you want to understand is constellation software you watch constellation software recently so constellation for many years has been my absolute favorite stock in the world right it like it is a perfect growth chart from 2006 to 2000 to may 2026 it's a parabolic beautiful curve but what they basically do is roll up sas right and like make it more profitable they have come down right from from a 16 000 x return i'm sorry 16 000 growth and over this kind of lifetime to 5 000 bucks it's trading at 3 000 right now right and like since may of last year wait may have you said

22:00

SPEAKER_04

may 26 did you say may 2020 uh 2025 25 got it again it's trading back to where it was trading in like 2023 or something the basic point is like it is i think a perfect illustration of what's going on in people's actual story on what's happening here there's a few other things that fit in fact but like just that's the story and then like look i there are plenty of good businesses to build in the world like that's not going to change at a high level but this i this idea that software is like defensible or remote or even a thing you can sell like i think we're just at the end of software sales right is the way to think

22:31

SPEAKER_04

about it and like that means you focus on community you focus on network effects there are things that are valuable but like there's no such thing as a software company more it's a real change though sam and i mean i think

22:42

SPEAKER_02

that's like the era we've been part of the internet in like the last 25 years it's been and there's a

22:47

SPEAKER_03

lot of seed funds out there that only know how to invest in this and so how should they think about

22:52

SPEAKER_02

this well they're just going to go away well a lot of people's portfolios are going to get really fucked by this that company that thought they could make 150k per partner in saas sales not so great

23:02

SPEAKER_04

what i've been telling founders for 2026 is there's only two business types i'm interested in right now that touch into technology right one type is you have a secret about the universe and you should tell no one ever what that secret is because it is an incredible idea and insight a way to make money it's like a secret incantation and if you know what you're doing is like a great way to make money with all the technical leverage but don't ever tell anyone and your hope is that no one ever finds out there are lots of cool businesses being built right now that look like what's the example of that

23:34

SPEAKER_03

yeah give me an example no i'm not saying like i'm not saying a stealth one what's one that's already exists that that was like this coca-cola they're all pretty stealth yeah i guess they have a formula but even coca-cola somebody on youtube last week spent a year reverse engineering it and actually

23:48

SPEAKER_04

achieved it nothing lasts forever but there are businesses i keep hearing of people i have a

23:52

SPEAKER_01

business like that but it's not based on one of those it's based on consistently generating those things that aren't known that are important sure but people know your business these are

24:02

SPEAKER_04

companies or it's like someone has a brilliant insight about a way to make money using like a bunch of ai tools and like they're minting cash but they will never tell anyone what they're doing because it's like it's it's infinitely copyable everything's infinitely copyable right that's option one option

24:18

SPEAKER_03

two is there are these things that everyone agrees so there are none of these in option one there are there this is a new phenomenon that there's no such thing as an example that's big that you can point to

24:30

SPEAKER_04

probably are but in some ways you don't know what they like there are a lot of businesses in america that make a ton of money that you've never heard of and will never find right like these exist right

24:39

SPEAKER_01

prevents the copying at launch like the second it launches no one because you don't even know there's

24:44

SPEAKER_04

no there's no website for these things right these are literally business people are minting cash off of very quietly using more and more technical leverage are you saying these are not these are

24:54

SPEAKER_02

venture types of businesses or just some of them might need a little capital some of them need none

24:59

SPEAKER_04

mostly not yeah it's mostly and they're like with ai leverage and tech leverage you can now build things and connect things together with an insight we're like oh my god i can get this type of data here and sell it there i can do this here and do this here i can make this there are all sorts of things going on with tons of leverage that small teams are doing they're making fortunes that's type one there are those things that exist it's just they're really hard to talk about and find and like the ones i know of like you can't say on a podcast yeah it sounds like they're not venture

25:29

SPEAKER_03

investable so is that helpful to like i guess i'm like trying to find lanes to help the audience help

25:34

SPEAKER_04

our audience forget venture capital like a lot of entrepreneurs it's like go find a secret to the universe you can do this with maybe you need a little money maybe you don't it doesn't really matter like whatever type two which i really do believe in is like there are all these businesses out there which you can say will obviously be categories right like everyone knows what they are like we all know that ads are gonna get hyper hyper personalized right like there's no question about that so then the other thing is the only way to win is you must own the narrative very loudly i'll

26:05

SPEAKER_04

give you an example that we can talk about you know i love meme lord like this in like this memetic warfare platform it's like i would argue everyone understands if you're smart that like the future of communication and like marketing whatever is mimetic right and like there's an argument that you should be able to build the defining platform at scale in like ai mimetic warfare right you can make that argument you might disagree with it but it's an argument you have to fully own that narrative as like i am the meme warfare company you will buy it from me right and so i think there's this big schism

26:38

SPEAKER_04

of like you either are a secret company or you're the most obvious company in the world but you leave the narrative you can't be in the middle britt what are you looking for if it's not

26:47

SPEAKER_02

software literally i mean anything that's not software bio life sciences human connection robotics hardware sort of defense space categories where you can't just replicate this stuff overnight or to sam's point it's almost like his first point there's like a novel discovery like there's a breakthrough in science and you know a lot of these researchers or scientists have a discovery but they don't know how to build a business around it and commercialize it and so i'm like i'm leaning into that i have a trump market though i mean that's always been like the business of drug discovery i mean

27:21

SPEAKER_02

i'm not necessarily talking about drugs i mean it could be like any kind of science yeah any kind of science materials like you know there's all kinds of breakthroughs happening right now my historically

27:32

SPEAKER_04

i'm too lazy to do that stuff it's too hard to diligence now you could make the argument that with ai it becomes easier for generalists to diligence all this stuff and so you can kind of have marginal leverage on it or be in that industry maybe that's good maybe that dave's point about like what's good in all this is there's all this shit that should get funded that generalists can't figure out but if you give them all the right ai tools all of a sudden they can have productive conversations and make realistic decisions with leverage on these harder areas and so you make more progress and

28:01

SPEAKER_04

fund more of the weird shit i like don't do that stuff because i think it's too hard can i that brought up

28:06

SPEAKER_01

i wanted to circle back we were we're talking a little bit of open ai health and like so i want to circle back to ai and health and some of the startups and opportunities there because i have a big question called business model but dave to your i liked how you decided that our goal was to try and add something to the silicon valley conversation it's a very high bar so i have we done that yet on this topic sam i think you one of your most viral tweets of 2025 was about software being dead or something and then you know you got some pushback from that from the software people but it does seem like we've entered

28:37

SPEAKER_01

a new phase right there's actual examples of contracts being obliterated we've also raised the like wild west of

28:44

SPEAKER_02

data security concern but if software is dead for for entrepreneurs then it's also dead for venture capitalists and that has there's like this like halo effect to that statement being true pretty dead for

28:56

SPEAKER_01

venture capitalists for like i mean talk to any firm about their stats on their books they've been

29:00

SPEAKER_03

complaining about this for two years i don't know yeah they've been complaining about it but you look at a lot of people's uh yeah their portfolios and there's a lot of software in the last six months

29:10

SPEAKER_02

right like no but even new companies people are still investing in software yeah just what i'm

29:14

SPEAKER_03

saying is that like if you're investing in something now uh or you're an entrepreneur starting something like you're what you're you're shooting a rocket that lands five years from now right and so five years from now like this is gonna be crazy like you you know like ai is the new ui like you don't even need a ui for anything anymore you can just like i wanted a new ui for my mural frames and i just instantiated it in

29:36

SPEAKER_01

10 minutes like that's crazy yeah how far so two quick follow-up questions one how far are we from jessica lesson being able to do that pretty close yeah really close like six months six weeks six days

29:49

SPEAKER_03

of hard work six months okay yeah good to know you could probably do it with claude co-work actually like yeah i'm not sure i haven't played with it i've played with claude co-work i did a big project

29:59

SPEAKER_02

yesterday i like mapped the entire seed stage vc market i like curled everyone's website and pulled in everyone's like key differentiator statement how much their check sizes are categories they invest in all these different things about them and and it was just like done for me while i was working

30:15

SPEAKER_03

here's another good example i took all of our transcripts we have 131 transcripts for this podcast and i ripped through all of them and generated essays across 131 uh transcripts and did that in less than an hour and i saw lenny you know lenny has a great podcast he was doing the same thing yesterday like and he took 250 transcripts and did the same thing and then he actually today i saw open source them all so that other people can do stuff with his transcripts right and so these are just massively leveraged ai moves that you can do using this stuff and it doesn't have to necessarily be code and the

30:51

SPEAKER_03

the answer to your question just like how soon will you do it i think actually is best answered through our 11 year old who over this weekend he he wanted to build a nintendo game uh he had an idea for a nintendo game for his game boy and so he used cloud code to develop a new nintendo game and had it done

31:11

SPEAKER_04

like in a day i look i think that's not true but i'll give you let me give you the counter sam just

31:15

SPEAKER_03

one last thing which is that it was very hard for him to understand github and there's all these like intricate things that are very nerdy about how to interact with these tools but they're not that complicated and will just get eliminated from the process and so you know it's got to be within six months that you can just deploy something onto versell and like have it in 10 minutes i i do

31:37

SPEAKER_04

think there's gonna be weird things that happen i'll give you i'll give you my counter example that dave which is our kids are now deep garmin watches and i was like i like they're like messing with watch faces all the watch faces kind of suck but you know what this is a perfect ai thing i would never want to learn their stupid garmin like language to like program a watch face but like ai knows this like i was like let's just like pop open cursor we're gonna make our own uh garmin watch faces right seems like a obvious perfect use case for this all the things you're talking about you go and you're like oh wait

32:07

SPEAKER_04

a minute interesting business thing cursor split the store from vs code because they want to like create their own ecosystem that's independent etc and the way it all plays out is vs code is like no like the package you need to install to do garmin watch faces we're going to control that all right and so all of a sudden you realize it gets pretty hard to actually do it without vs code so now you're using vs code instead of cursor even though it's the same thing right did you try cloud code i haven't tried yet i don't think it'll work i'll just do it i don't i'll try it i actually kind of doubt it'll work but

32:40

SPEAKER_04

we'll see because they have a forked because you go back to you basically get stuck in an app store problem with microsoft who has access to the packages and then here's the thing vs code and their ai which is like basically all the same stuff anyway just spins for hours trying to like get a dev environment set up it like kind of knows what it's doing but it's just not quite there now does that all get resolved maybe but i do think there's gonna be all sorts of weird business development games that start getting played with lockdown app stores lockdown code bases who has access to what all of

33:10

SPEAKER_04

a sudden garmin is going to be like hey guess what we had a whole business around watch faces we're not going to just let you push now that it's so easy to make them we're not going to let you just like push them up we're going to change our business model around them like the world will reorient to try to extract value whether it's app stores or whether it's like access points or whatever like it will be a game of chicken and mouse on this stuff do you think they're going to just start charging for that

33:33

SPEAKER_02

like with claude co-work i was connecting all these things into it yesterday at some point is asana gonna get any additional revenue from claude because millions of people are integrating with with it or is that just net value add to asana for more use it's because it's actually depleting asana right because it's just writing on top of it and pulling scraping it the good news is i believe dustin's a very large

33:58

SPEAKER_01

anthropic investor so he'll be just fine okay notion for instance as another example well no this is the wild garden thing right and this gets back to when i was saying the sas companies are just trying to embed ai as fast as possible but there's a new race to be the default to be the ux and i i think it's going to be very hard for like a single pointus software i'm just saying there's no business model

34:20

SPEAKER_02

that exists so far in terms of who's getting paid to be the connector and all these things i remember

34:25

SPEAKER_01

talking to erin levy a box about this and you know i think there's expectations that would be some kind of like reciprocal tariff structure or something like that but um we're also seeing this play out in shopping and i don't i don't think that they're the right incentives for that to happen last thing on this point we'll leave us a little time to get a little not i mean not non-ai but did you guys see toby the ceo of shopify also tweeted something very much on this theme which was like he took an mri i believe and and had claude code just i think it was claude code had some ai coding software just built him

35:02

SPEAKER_01

a better web interface for it instead of using like the legacy i was super excited about that because i'm

35:08

SPEAKER_02

getting two mris tomorrow so i'm going to be uploading those well yeah he had a he had a usb

35:14

SPEAKER_03

key or something that was uh windows software that he doesn't have windows and so he's just i mean this is the this is the story that ai is the new ui you you don't have a ui to read the thing so you just tell ai figure out how to talk to this thing that i have and like make me an interface for it boom it's done right like the number of examples of this this week have been insane and this is all possible just

35:38

SPEAKER_01

because of apis like no one like locked this down in a way or it's just that the way the ai operates it doesn't need a certain type of it's just using some kind of predictive system to be able to do

35:52

SPEAKER_03

that no it runs through the whole internet it goes and looks at the internet it looks at every example of everybody that's ever tried this it looks at every forum and every you know every open source

36:02

SPEAKER_01

project and on and on and on but how does it get the underlying data you've given it the underlying data like you've given it the mri it's reading or the yeah that's on the disc right that's like

36:13

SPEAKER_03

there's a usb key that has the mri on it but you can't read the file format because it requires windows and some legacy software and so just like digs into there and figures out the data and pulls it into a

36:23

SPEAKER_04

web interface here's the deal forever this is actually a rehash of an old argument that existed in the early days of meta and whatever else about apis in general and what you can and can't have access to there's no the only rehash the rehash is this is just a lot faster and more scaled but it's the same story of like oh you remember there used to be rss feeds for everything and those got shut off why do they get shut off because once you built readers on top of them no one can make any money and so like well this doesn't work and so the upshot is like in some ways all that ai is doing is

36:53

SPEAKER_04

rationalizing effectively the world of apis and the world of programmatic access because when you drop access where it's not just like really sophisticated smart people who can kind of do stuff and you're like ah this doesn't quite make sense business wise but it's fine you just like turn the dial to 11 and everyone can do it you're like oh yeah we can't have apis to anything and so to jess's point everything's a wall garden everything's locked down everything's a war right everything's a fiefdom and then you get back to this this thing which is it's super hard to have a fiefdom and even have access to data when i

37:22

SPEAKER_04

just sit something on my desktop and give it permission to everything and like let it sit there as an end agent and root it all right like so what do you do with that right well it's going to be a trust thing and a security thing right uh is the upshot okay so i just have so many questions and i am

37:38

SPEAKER_01

the voice of the every woman here but i understand that everything your desktop has a lot of stuff but like my files are in the cloud they're not on my desktop my photos are in the cloud they're not in

37:49

SPEAKER_03

the cloud yeah they can like go on the cloud it can open a browser open the cloud like literally

37:54

SPEAKER_02

it run it ran my whole browser yesterday it was crazy running your browser it's running your computer

38:00

SPEAKER_04

imagine it just as literally has your keyboard and mouse anything you can do it can do hence the

38:04

SPEAKER_01

security problem is was this what openai's browser wanted to do anyone using that lately i tried it and

38:11

SPEAKER_02

was not impressed it doesn't really do anything so what's on your next to build list oh that's a good question i i'm mostly doing stuff internally for offline right now and i'm like we're working on like what hires do we actually need to make and what things could we just build and i don't know if you guys been following this whole boardy thing do you know boardy jess has information written about boardy yet sure i don't i tell me boardy is a whatsapp and email client that is a marketplace of founders and vcs and so it's like trying to map like in one week it will source 6 000 founders across its

38:48

SPEAKER_02

marketplace figure out the types of vcs that it should introduce them to ask me the vc like hey i got these five founders are you interested in any of them here's what they're working on i say yes or no

39:01

SPEAKER_03

companies are paying it ten thousand dollars a year to do bd on it on their companies themselves are the

39:07

SPEAKER_02

ones paying so and then on top of that it raised like 150 million dollar fund because lps think this is magical and this is like a new mechanism for sourcing you guys like this isn't this just slop

39:20

SPEAKER_01

what about the ai slop it's unclear this is founder slop well i'll tell you doesn't it look if it's

39:26

SPEAKER_02

perfectly efficient you'll never make any money i think it's the the notion that i'm i'm saying i'm talking about it is it like are all venture funds going to be building their own versions of this tim draper launched this thing where like you can talk to tim ai and it's a year ago it's just you can go to my website and do it is this like what is the future of matching with founders pitching are there any secrets anymore is this all just like a marketplace like it and like what are the internal tools everyone's building at their venture funds to like do all this stuff and experiment with all

39:58

SPEAKER_02

stuff i'm so curious how much time people are spending how much they're paying or not paying

40:02

SPEAKER_01

how hiring plans have changed like i'm old enough to remember when google ventures was going to automate vc investing i mean do you remember this it's probably like 2006 and it was all going to be

40:13

SPEAKER_04

an algorithm or whatever and then it wasn't look in the end of the day i think that with the seed stage investing is the most defensible part of the entire investing world everything should get more efficient etfs algorithmic trading does that come down market of course it comes down market and it spans but any idea that a bot can route to the six right firms and everyone bids and everyone top ticks it but no one will make any money which is fine that's actually what the whole point is like that's why capitalism works is it creates efficiency right so that's fine but that's not what we do like

40:44

SPEAKER_04

i'll give you an example like candidly like we don't do outbound sourcing at all like there are companies there are venture firms inside wherever they're built machines for outbound and market mapping all stuff i don't do any of that like we say on the internet and then people email me right and i only want to talk to a certain percentage of them we don't and so like the idea i think the answer is for a lot of types of investing yeah like the all this data if you're doing something like an associate can do in excel that's going to be overly efficient and you'll have no margin in it but that's not

41:14

SPEAKER_04

really what we do i don't think it's what you guys do either well sam what i think is interesting is

41:20

SPEAKER_03

like in the seed market you have the multi-stage firms and 2 000 seed funds competing for the 75th percentile of all of those sort of just like you said the mainstream stuff out there in the market and so that effectively means you have almost perfect pricing and a bunch of synthetic beta in the top quartile of the seed market and so where in the seed market is there alpha is like a question and it's you know i think you and i have always agreed it's in the illegible spaces but the question is like how

41:49

SPEAKER_04

to stay there ever more you have to be smarter more original have access to networks that everyone else doesn't have access to ideas other people don't be willing to take risks other people don't like all

41:59

SPEAKER_02

that stuff you know but i think that at least 80 of other funds seed and beyond are don't think that way and they're like they're gonna they want to build all the stuff to automate a bunch of things and like they're on the ai train and so i'm really interested to see where this goes this year given by coding has

42:18

SPEAKER_01

definitely peaked each of you were you more optimistic about seed investing six months ago

42:25

SPEAKER_03

or today or less optimistic same like this it doesn't same yeah i'm probably neutral to six months

42:33

SPEAKER_04

ago i don't think anything's even like i honestly despite the fact this stuff is cool again again we try to live a few months ahead of reality right like i'm not that surprised by anything we're seeing no

42:44

SPEAKER_03

yeah i agree with that that's why i'm saying like i'm neutral to it because i feel like we've been talking about it on this pod and in fact i've looked at all of our transcripts in many different ways now we have been talking about this on this pod so the point is like i just don't find anything

42:56

SPEAKER_04

we're seeing surprising right now right so it's hard to say that anyone's opinions have changed they're

43:01

SPEAKER_02

just the same it's just playing out in real time as we were talking about that it would you guys it would

43:07

SPEAKER_01

have been a great sound bite no one gave it to me so now we'll move on um so what else is happening

43:12

SPEAKER_02

brit golden globes what would oh well nikki glazer i thought it was a worse year than last year however um number of people on glp ones year over year definitely increased at the golden globes and by the way fun stat grocery sales are down 5.3 percent since since the rise of glp ones and eight percent among wealthy and affluent people so people but just i mean it's coming down across the board

43:40

SPEAKER_01

so people are eating less but they're also cooking less i would just want to point out groceries would be not perfectly equated to eating or cooking sam how are you doing having not eaten for four days i'm fine you didn't wait you did did you do your prolon thing prolon i love prolon this is your chance to do an ad for prolon tell the people prolon's great it's just you do the modified fast thing

44:01

SPEAKER_04

your stress level goes back to zero i slept i got a 90 97 of my sleep score last night i mean it's not something you want to do all the time but it's great to do okay so sam's mood still up on day

44:12

SPEAKER_02

four or five of the fast i tried it but i'm good without prolon oh and everyone's moving to texas my home state your hometown welcome to austin everyone i'm happy to tell you where the best case though is yeah so with the wealth tax that we talked about last week on the pod the you know the californians are all moving away they're not going to new york some are going to miami the new thing that

44:35

SPEAKER_03

you have to move to you have to move move to austin before series b is that like the new path for

44:41

SPEAKER_01

founders that's the new talking point but when we so the information we surveyed our subscribers which is always fun because it's a specific crew to say the least but so 67 would not move right so wealth

44:54

SPEAKER_04

tax all that kind of that means there are many other that's the who's that's the percentage of your

44:59

SPEAKER_01

subscribers that are billionaires it's possible that those are yeah i cannot verify people's you should

45:04

SPEAKER_04

change it'd be like 33 of our subscribers are billionaires i know i actually that's really funny

45:10

SPEAKER_01

i do believe we have the highest daily active billionaire readership on the internet and um the aub i use it in sales pitches by saying about it the way i would frame it which i'm sure is true

45:23

SPEAKER_04

is you have the highest percentage of american wealth reads you every day right like forget the billionaire thing because that's just like usually like if you just take the number the net worth of our readership exceeds any other publication in the world by a mile that's what i would pitch great these

45:40

SPEAKER_01

things are all true like deeply true but okay so nevada but do they respond to the survey texas and florida edged out nevada and poor new york why would you think they would go to new york they have the same

45:54

SPEAKER_03

problem yeah there's but there's tax yeah why would you why would you run to the socialists you know

46:02

SPEAKER_01

they're going to take your money too because it's new york city i know i mean i pointed all of that out in my tweet i know people aren't racing towards montana but a mere two percent guys it's new york city where i would live if i didn't live in california i mean it's it's like it's amazing but you're an east

46:19

SPEAKER_03

coast person like you're an you're an east coaster and a lot of people you know they have a wealth

46:24

SPEAKER_01

tax on the ballot i'm still three percent i wasn't expecting twenty percent but like i mean it it's

46:30

SPEAKER_03

pretty it's to sam's point it's entirely politics right like florida and and texas are like leading the

46:36

SPEAKER_02

charge for the free world that's like the guys is there like real estate arbitrage we should be

46:41

SPEAKER_04

considering for austin right now this has been my thing i've been very excited for the mere millionaires who are going to get to buy all the billionaire research real estate in california i know but

46:52

SPEAKER_02

should we also buy and then flip stuff for the billionaires to buy in austin i would like to get

46:56

SPEAKER_04

like both sides of the market i think that trade's already been taken that trade's over also the stuff

47:01

SPEAKER_03

they want is really really really expensive yeah no there's there's like something like two there's something like 240 lots on lake austin and they want one of those those are the those are the those

47:12

SPEAKER_01

are the lots that everybody wants okay well time will tell but yeah this week also saw the rise of that meme i mean what's interesting to me now is like their attempts to organize the tech community to fight fight this more formally i think gary tan at yc has been very vocal about that but yeah it's like you know start your company and won't that backfire it could i mean it can all backfire we just no one

47:34

SPEAKER_03

knows anything yet but how much truth was it uh i also saw that there's an effort to create a new city

47:40

SPEAKER_01

in greenland right sure everything is possible so we had the golden globe like what's happening in the world yes the 49ers i know i i mean it was amazing i was on a plane for part of the game we're gonna win the super bowl but kittles out you think they're gonna win the super bowl dave i mean we lost kittle

47:57

SPEAKER_04

by the way dave claude code failed at making a garmin watch face i just tried it really

48:02

SPEAKER_01

claude bot windows binaries be damned let's end on this note what percent so claude code claude co-work are amazing you can squint squint and see the future what percentage of things you try fail like 80 no no

48:17

SPEAKER_04

it depends on what is about inventive being i just the real problem i have honestly okay now sam's about to disrobe i'm gonna i'll reveal my shirt but before i do at the end this sam is starting to unbutton himself if anyone is not watching this here's the thing that i'm actually curious about this stuff i like i love tinkering i build projects all the time i don't like use any of the things i build

48:42

SPEAKER_03

really right and so maybe that's the problem no no so here's the thing i am not convinced

48:47

SPEAKER_04

i don't use letter meme the letter is the only thing i do use but i've spent more that's more than

48:51

SPEAKER_03

a vibe-coded project at this point letter means the main thing you've built that i use letter name's

48:55

SPEAKER_04

great and it's only gonna get better but like that's not that's more than a vibe-coded thing at this point but like all the stuff i build like little tools and shims and i've i've spent weeks building cool stuff and it's very very fun like i love our line from the summer that the real thing is ai does replace instagram because it is literally more entertaining to build toy it's like legos it's super fun it was netflix i still like my netflix it's like it's really like legos and legos are really fun i don't actually think that almost any of this stuff is actually that useful right and so for me i

49:27

SPEAKER_04

think the reality is this is gonna be the thing is like take cloud code or any of these things it's like you're gonna build a lot of toys what percentage of them or what do you actually use i don't know like i i think it's gonna be pretty few and far between i think that would be an interesting thing where like tons and tons and tons of projects get built but almost nothing's useful maybe you just need better ideas i have pretty good ideas dave that's not the issue it's more just like i don't know what is

49:52

SPEAKER_03

a good idea good idea gets used right like i just i mean by this way i've built i've spent tons and tons

49:59

SPEAKER_04

of hours building fun ai things for myself i've spent extremely few hours using it right and i don't think that's because i'm stupid or building the wrong stuff i think that most people just don't actually know what to build with it or what to do i mean i'll quote my slow ventures newly appointed partner yoni on this who was like this is cool but i have no idea what to build brings up for me an

50:20

SPEAKER_01

interesting point which is like obviously these tools are still disruptive if you just think about

50:25

SPEAKER_02

like the future of software which we're talking about but i think the other narrative which we can

50:29

SPEAKER_01

get into in coming weeks and have continued to talk about is how quickly now the big companies are also rolling out a lot of some useful some not useful tools and google unencumbered by the doj anymore is basically going big google and tying together all your data and all the things you actually really want as a google user and that will continue too and so i think that also kind of explain like you know the tinkering but that doesn't make what's possible with the tinkering less significant when you think

51:02

SPEAKER_04

about that playing out across the whole industry i don't know but i look again it's just like we're in the fart app era of the iphone we're like people gonna build all the cool stuff most of it goes away it's like i i just i do think it's really fun yeah but the fact that those tools existed created

51:19

SPEAKER_01

and destroyed a lot of businesses not as many as you think well the app store did

51:25

SPEAKER_04

what did it destroy like it made the iphone made really big business the taxi business taxi business is probably the best example of what the iphone did destroy i mean it probably the news

51:37

SPEAKER_01

business because of all the ways you could entertain yourself on your phone i mean no it was already gonna

51:41

SPEAKER_04

be destroyed by the web though i just i think this is like the thing is like what the iphone did is just more internet and it made the big companies bigger as we've been saying i really believe that you know again in terms of no no new information that in general what ai will do is make big companies bigger i think it totally takes away software as a pricing mechanism which means a bunch of software companies go away but when you really really scope out i just i still am in the camp of like it's going to be less i mean i think it is socially deeply disruptive i think it's going to like screw up our government

52:15

SPEAKER_04

and how we think about value and employment and also like there's a lot of things that we'll do but i don't know like in terms of like what applications i use day to day i'm not convinced that ai is going to like dramatically change my app stack personally i wasn't arguing that it was but

52:31

SPEAKER_01

okay so everyone can have a listen to this episode and reach their own conclusions around what we spent the bulk of our time talking about whether it matters or not i think it matters but not necessarily it can matter by making the big guys bigger but i also think it's going to have deep implications for sas and the whole enterprise software yeah sas is clearly going away i agree

52:52

SPEAKER_04

that i just don't think like i don't really agree on that i don't really use sas many businesses do

52:58

SPEAKER_01

okay friends it's been a treat thank you to all our listeners viewers followers subscribers commenters i won't be here next week but the gang will be so stay tuned and we'll more or less we'll be back next week and i'll see you all soon bye bye if you enjoyed this show please leave us a virtual

53:19

SPEAKER_02

high five by rating it and reviewing it on apple podcast spotify youtube or wherever you get your podcasts find more information about each episode in the show notes and follow us on social media by searching for at more or less at dave morin at lesson at j lesson and as for me i'm at brit see you guys next time

Reading tools

Type to find a passage

Appearance
Ask this transcript

Notebook

Export Markdown

Add a note