Transcript

Harry Stebbings (0:01)

You're watching TVPN. Today's Wednesday, April 8, 2026. We are live from the TVPN Ultradome

Noah Hirschfeld (0:07)

Temple of technology, the fortress of finance,

Harry Stebbings (0:09)

the capital of capital. We got white suits on. You know what that means. The stock market is booming. The Dow Jones is up 2.68%, the S&P 500 is up 2.46%, the Nasdaq's up 2.9%. And there's a bunch of other stocks that are moving within that. Of course this is on the back of the the very good news that there has been a ceasefire, that the street might be opened. Of course, it's all back and forth. The front page of the Wall Street Journal is covering all of the geopolitical moves. But we're here to talk about tech and business, of course. And the big news today is that Meta Platforms has launched a new AI model. Alex Wang, the chief AI officer at Meta Platforms, announced a new large language model today, its first major new artificial intelligence model in more than a year. The rollout of the model called Muse Spark is a critical moment for Meta, which is up 7.5% already, which has spent billions of dollars hiring AI talent in a bid to catch up to OpenAI, Anthropic and Google DeepMind leading labs have been putting out models at an accelerating pace. In a departure from its previous models which were open source, Muse Spark is a closed model that will power Meta's AI chatbot and AI feature features within it. John Ludig has a very interesting post about open source AI and sort of predicted this. I can pull that up at some point.

Noah Hirschfeld (1:36)

We can find predicted that Meta would eventually bail.

Harry Stebbings (1:38)

Yeah, let me find it. The future of foundation models is closed source. Let me see if we have this here, he said. Given Meta is the primary deep pocketed large open source model builder, open source AI has become synonymous with Meta AI. He wrote this maybe three or four years ago. So the operative question for open source AI is what game is Meta playing? In a recent podcast, Zuckerberg explains Meta's open source strategy. One he was burned by Apple's closeness for the past two decades and doesn't want to suffer the same fate with the next platform shift. It's a safer bet to commoditize your compliments. He likes building cool products and cheap performant AI enhances Facebook and Instagram. That's 100% true. We've seen this in the ads product and the growth there. There's some call option value if AI assistants become the next platform and that makes sense. In Manus and the Meta AI app, he bought hundreds of thousands of H1 hundreds for improve social feed algorithms across products and this seems like a good way to use the extras. That all makes sense and Llama has been great developer marketing for Facebook. But Zuck also suggests several times that there's some point at which open source AI no longer makes sense, either from a cost or safety perspective. When asked whether Meta will open source the future $10 billion cost model, the answer was as long as it's helping us, at some point they'll shift their focus towards profit. And that's what John Ludig wrote. When did he write this? 20 man time flies. Barely just under two years ago, he says unlike the other model providers, Meta is not in the business of selling model access via API. So while they'll open source as long as it's convenient for them, developers are on their own for model improvements thereafter. That begs the question, if Meta is only pursuing open source insofar as it benefits themselves, what is the tipping point at which Meta stops open sourcing their AI sooner than you think? He says. Exponential data Frontier models trained on the corpus of the Internet, but that data is a commodity model. Differentiation over the next decade will come from proprietary data, both via model usage and private sources. Exponential capex. He highlighted this two years ago. A lagging edge model that requires just a few percent of Meta's.40 billion in capex is easy to open source. No one will ask questions. But when you reach $10 billion or more in capex spend for model training, shareholders will want clear ROI on that spend. The metaverse raised some question marks at a certain scale too, diminishing returns on model quality within Meta. There's a large upfront benefit for Meta building an open source AI model, even if it's worse than the frontier closed source counterpart. There are lots of small AI workloads, think feed algorithms, recommendations and image generation where Meta doesn't want to rely on a third party provider like they had to rely on Apple. And so the news has been. Back in December there was a reporting that Alex Wang disclosed an internal company Q and A that his team was working on two new models. One was this text based LLM codenamed Avocado. And then a separate model that was for image and video.

Summary

TBPN Podcast Summary

Episode: Meta Drops New Model, Mythos, RoboLamp | Luther Lowe, Dan Primack, Lior Susan, Feross Aboukhadijeh, Qasim Mithani, Jaleh Rezaei, Jeremy Philip Galen

Date: April 8, 2026
Hosts: John Coogan & Jordi Hays (with regulars Harry Stebbings, Noah Hirschfeld, Martin Casado, and multiple guests)
Theme: The accelerating pace of AI, cybersecurity, prediction markets, defense tech, and the changing regulatory and venture landscapes.
Tone: Caffeinated, playful, highly technical, and insight-heavy—often with snark and inside jokes.

1. Overview:

The episode is a rapid-fire, information-rich tour through pivotal new launches and crises in tech. Highlights include the closed-source AI pivot by Meta with its new Muse Spark model, Anthropic’s high-security Mythos release, major supply chain hacks, evolving startup and investment landscapes, and the real-world impacts of AI in both consumer tools and national security infrastructure. The show features a parade of influential guests, each dropping fresh perspectives on the intersection between tech, policy, and society.

2. Key Discussion Points & Insights

A. Meta’s Muse Spark: A New Era in Big Tech AI (00:09–19:17)

Meta’s Big Pivot: Meta released its first major new AI model in over a year—Muse Spark—a closed-source LLM, signaling a departure from the open-source LLaMA tradition (00:09–01:35).
- “The future of foundation models is closed source.” – Harry Stebbings citing John Ludig [01:38]
Open vs. Closed Source & Benchmark Games:
- John Ludig’s prescient analysis outlined Meta’s incentives for open-sourcing until ROI demands secrecy and profit [01:36–04:34].
- Meta no longer leading with open-source; focus shifting to competitive, proprietary data and monetization.
Benchmark ‘Chart Crimes’: Mixed performance across leaderboard benchmarks—Muse Spark beats some rivals (Gemini, Grok) but underperforms on others (ARC AGI 2) [05:12–11:09].
- “They have at least moved on from the culture of optimizing for the benchmarks. Isn't that a good thing?” – Harry Stebbings [10:14]
Personalization Quirks:
- Strangeness in Muse Spark’s joke personalization, leading to discomfort about privacy.
  - “Why would you think I want Malibu appropriate surf puns?” – Noah Hirschfeld [06:55]
  - “It says, you got me… But then it denies it again.” – Noah [07:36]
- Hosts question whether the model’s personalization is drawing data too closely from users’ histories, raising PR and safety questions.
Economic Rationale: Closed models allow Meta to better justify immense capex and leverage proprietary data for internal features deployed to billions, rather than selling API access [11:19–16:03].
Market Response: Stock jumps 8%, market hungry for clarity on Meta’s AI investment payoffs [15:22].

B. Anthropic’s Mythos Model: Red Teaming AI Security (19:17–26:35)

Release Details: Anthropic unveils Mythos—optimized for finding zero-day bugs and exploits, limited preview to critical infrastructure partners, tight cybersecurity focus [19:17–23:25].
- “Models with obvious national security implications would not be forthcoming. Those people were wrong.” – Quoting Dean Ball [29:37]
Responsible Release, Not Safety Theater:
- Debate over whether Anthropic’s move is about genuine safety or investor hype / anti-competition with China [23:35–26:14].
- Notable skepticism: “Anthropic's marketing strategy is so funny...our models are so good we can't release them. It would be too dangerous.” – Buco Capital Bloke [23:35]
Implications:
- The cybersecurity arms race accelerates; “we are thoroughly in the era of the lab's best models may well not be in public the way they used to.” – Dean Ball [29:37]
- The best AI will be available only to top bidders due to compute and safety constraints.

C. Defense Tech, Data Centers, and Venture Markets (77:54–92:52)

Defense Tech Boom:
- Discussion about lasting impacts of Middle East unrest, supply chain chokepoints, and defense tech as a huge new venture sector.
- “Standard munitions stockpiles are going to have to get refilled.” – Dan Primack [80:03]
Data Center Bans as Next Flashpoint:
- Data center moratoriums more serious (for deal flow) than even rising oil prices; PR battle for the sector [76:49–77:38].
Venture Funding (AI/Defense/Cyber):
- Fundraising for industrial/physical tech surges, especially in defense, supply chain robotics, and manufacturing (Eclipse’s $1.3B raise).
- “AI is not a bubble thing.” – Recurrent motif in discussion with Dan Primack.
- IPO window narrow, only a few massive companies like SpaceX, OpenAI, Anthropic (the “big three”) are prepping to go public.

D. Supply Chain Attacks and AI Cybersecurity

Featuring: Feross Aboukhadijeh (Socket), Kassim Muthani (Depth First), Jeremy Gallen (Charlemagne Labs)

Breadth of Attacks (Axios/NPM):
- “Socket was among the first to detect and report… we built a system that downloads every open-source package in existence within a few seconds.” – Feross [94:00]
- Social engineering sophistication: North Korean actors targeted top maintainers using staged video calls and fake Slack workspaces.
- “Once you run a compromised open source package… you kind of have to rotate all your credentials…”
Industry Response:
- Security moving to board-level concern; organizations signing up for Socket in droves post-incident.
- The new attacker/defender arms race: “In security, the minute you think you’ve got things under control, the attacker evolves.” – Feross [105:04]
Depth First’s AI Offensive–Defensive Approach:
- Training in-house LLMs to find and triage vulnerabilities faster than state-of-the-art (Opus 4.6).
- “We found vulnerabilities in Chrome, in Linux…” – Kassim Muthani [116:24]
- Emphasis on RL loop, responsible disclosure, and the complexity of building generalized bug-find agents.

E. The App Store, Vibe Coding, and Regulatory Fights

Featuring: Luther Lowe (Y Combinator)

Explosion in “Vibe Coding:” One-person, AI-assisted, high-GMV startups are becoming reality, but face growing App Store gatekeeping [43:11–48:51].
- “The Apple App Store is basically like the worst DMV in the world.” – Luther Lowe [43:30]
Policy Initiatives:
- Advocacy around BASE Act, aiming to curb self-preferencing by tech giants and increase competition, especially for alternative app stores and AI assistants.
Siri & the Race for Personal AI:
- Frustration with limitations on iOS for third-party assistants (“Why can’t Siri do it? … until this self-preferencing is addressed...we’re not going to see the fruits of the LLM revolution.”)
- Prediction that Apple will try to build its own vibe-coding tools and retain control [47:14].

F. Startup Spotlights & Industry Trends

Mutiny (w/ Jaleh Rezaei):
- “The mission of Mutiny was all about killing the dependency season… We hire people that are aligned with our values and we just let them be themselves.” [119:53–125:43]
- AI “agent” handles all customer-facing workflows, case study generation, personalized outbound, prospecting—all at scale.
- Mascot origin stories: “His name is Achoo.” [120:24]
Charlemagne Labs (Jeremy Gallen):
- Defensive AI for spear-phishing and scams—building “Agent Charlie,” a watchdog for every employee.
- “In an AI-powered world, all phishing becomes spear phishing… You’re not going to get a Nigerian prince email much anymore.” [127:17]
- Collaboration with Meta on LLM evaluation for social engineering risk.

3. Notable Quotes & Moments

“What is personal superintelligence if it doesn’t even know your name?” – Harry Stebbings [09:17]
“You gave a hyper specific example based on my life, so I have to assume you were looking at my Twitter account for inspiration.” – Noah Hirschfeld [07:49]
“If the attacker has to just find one way in… defenders have a much harder job.” – Feross Aboukhadijeh [101:03]
“In security, the minute you think you’ve got things under control, the attacker evolves.” – Feross [105:04]
“The antidote… we just announced yesterday a coalition of 275 startups and VCs in support of the BASE Act.” – Luther Lowe [50:15]
“There’s nothing to do about it. But you should remember it.” – Quoting Dean Ball on Anthropic Mythos [29:37]

4. Timestamps for Key Segments

| Time | Topic / Segment | |-------------|--------------------------------------------------------------------------------| | 00:09–19:17 | Meta & Muse Spark model; open vs. closed AI | | 19:17–26:35 | Anthropic Mythos, cybersecurity roll-out, industry responses | | 43:11–54:39 | Luther Lowe on App Store, regulation, Vibe coding, and the BASE Act | | 77:54–92:52 | Defense tech, funding, IPO window, and the impact of global conflicts | | 93:18–110:21| Feross/Sockets & Axios supply chain hack, open source threats, AI in security | | 110:34–117:45| Depth First's LLMs for vulnerability detection (Kassim Muthani) | | 118:05–125:43| Mutiny, AI agents for go-to-market, startup lore (Jaleh Rezaei) | | 126:03–134:57| Charlemagne Labs: AI-driven scam/phishing defense (Jeremy Gallen) |

5. Final Thoughts & Episode Impact

The episode is a panoramic, sometimes dizzying look at how rapidly the ground is shifting across AI, cybersecurity, consumer tech, and the venture ecosystem. The hosts and their guests paint a clear picture: The AI race is entering a new, more closed and risk-aware era; security—human and technical—is now existential; and incumbents and startups are vying not just to innovate, but to determine the rules of competition itself. Stories are underscored by real hacks, policy moves, and product launches all happening in real time, with the show offering both technical depth and sharp commentary.

End of Summary