Entrepreneurs
3 Curated Dialogues
Mark Zuckerberg
Frontier Insights
Frontier Thesis: Biology is fundamentally an information science, solvable by combining AI foundation models (ESMfold, virtual cell simulations) with massive, open-source biological datasets.
Strategic Execution: Deploying long-horizon philanthropic capital ($500M+, 10–15 years) to fund non-traditional infrastructure: scaling compute from 1k to 10k GPUs, wet-lab instrumentation, and expanding the CELLxGENE data flywheel from 125M toward 1B cells to establish open industry standards.
Risks & Warnings: In silico simulations face a severe clinical translation gap; bridging wet-lab validation to actionable therapeutics and navigating biosecurity risks remain unproven bottlenecks.
Key Views & Dialogues
Mark Zuckerberg on how Meta’s Muse AI agent will make people money
Meta is launching Muse as a 24/7 goal-driven personal agent with roughly 100 million free tokens weekly, monetized through a small cut of transactions potentially paid by merchants rather than users. Its confidential-VM privacy architecture, purpose-built models, and eventual fleet learning could differentiate it as distribution scales to billions, while Watermelon’s release and Muse Spark 1.3’s position behind Claude’s Fable 5.1 and Opus 5 remain key tests.
View summary (8)
- Meta is launching Muse, a long-lived personal agent that Zuckerberg positions as the fullest expression of his “personal superintelligence” thesis. Unlike prompt-and-answer chatbots, Muse runs on a cloud virtual machine, takes goals rather than prompts, works “24/7,” “studies overnight” by consolidating reflections into memory, and proactively suggests new projects — with meaningful model updates shipping roughly monthly.
- The business model is deliberately aggressive: roughly 100 million tokens a week free, plus the VM, with monetization via “a very small cut” of transactions (Stripe on payments), potentially paid by merchants rather than users. Zuckerberg’s stated conviction: “we think that this thing is actually going to make you money and save you money and that is how it’s going to pay for itself” — free access being “critical if you want to build this future for everyone.”
- On the model roadmap: Muse Spark 1.3 is based on a relatively smaller pre-training code-named Avocado. Heath cites a recent Artificial Analysis chart and says he thinks Muse Spark was behind Claude’s Fable 5.1 and Opus 5. Zuckerberg says the larger Watermelon model is shipping soon, while the gigawatt Prometheus cluster in Ohio is scaling post-Watermelon models.
- Zuckerberg describes a privacy differentiator: a confidential-VM architecture involving Signal founder Moxie Marlinspike, whom he and Nat personally recruited and who helped build WhatsApp encryption in 2014, so “even Meta cannot see the content” — a commitment he says can be technically verified. Layered on top: Sentinel agents monitoring for prompt injection, human-in-the-loop approval for payments and logins, a secure credential store, and least-privilege connectors. “I’m not aware of anyone having anything close.”
- The manifesto’s core is a rejection of Silicon Valley’s restrict-access consensus: “I personally am much more worried about a small number of labs or people having control of something that is so capable.” His three principles — empowerment drives prosperity, AI’s purpose is invention not automation, and safety comes from checks and balances — and his best evidence is the Hugging Face intrusion, where defenders turned to open-source models because they lacked access to some closed ones.
- The personal-agent market is not winner-take-all, but Zuckerberg guesses fewer than a dozen companies can do state-of-the-art work, with power-law usage accruing to the best. Meta’s claimed edges: ground-up model design (including attention to “discretion”), social DNA, possible fleet-level interaction and learning across agents — “right now most of the industry is thinking about agents as like a single player game” — and distribution to “eventually billions of people.” Zuckerberg says most agent-to-agent interaction is not rolling out in this release.
- On data-center backlash, he draws a dichotomy between decades-long operators and speculators flipping sites, citing Louisiana tax revenue funding $50,000 teacher bonuses and a workforce academy guaranteeing jobs to trained tradespeople. He won’t call it a bubble — “that implies that it’s overvalued… but there’s certainly a boom” — and argues that if AI doesn’t create broad-based prosperity, “it just effectively won’t be able to happen.”
- On governance, he emphasizes close partnership with the U.S. government over relying on rigid frameworks that could be “out of date in a few months,” including ensuring the government knows about important training runs. The youth-safety settlement is framed as a legally binding industry mechanism: Meta unilaterally limits teen usage first, with the next step locking in when YouTube and TikTok sign the same terms — “we’re putting ourselves a little bit out there by going first.”
Mark Zuckerberg & Priscilla Chan: How AI Will Cure All Disease
Biohub is betting that shared scientific tools, not another round of small grants, can accelerate cures through $100 million to $1 billion investments over 10–15 years. CELLxGENE standardized single-cell data and created a network effect: CZI funded 25% of the resource while the broader community contributed 75%, supporting a model-to-experiment flywheel. Biohub plans to expand from roughly 1,000 GPUs toward 10,000, but virtual-cell models remain quite early and must prove that directional predictions can reliably derisk costly wet-lab work.
View summary (7)
- The Biohub thesis is that shared scientific tools—not another round of small grants—are the highest-leverage route to faster cures. Mark Zuckerberg said major breakthroughs usually follow new ways to observe phenomena, yet tools such as imaging systems and virtual-cell models can require $100 million to $1 billion over 10–15 years. “We’re not going to cure all diseases”; the strategy is to help the scientific community do it.
- AI changes the timetable only if frontier models and frontier biology form a closed data flywheel. Biohub intends to design experiments and instruments around model blind spots, generate purpose-built data, retrain the models, and repeat. That bridges a striking cultural split: biologists considered curing disease “crazy ambitious,” while AI researchers thought it was “kind of boring—that’s just automatically going to happen.”
- CELLxGENE demonstrates how open infrastructure can acquire a network effect larger than its original funder. Built to clear an annotation bottleneck, it standardized formats and metadata across single-cell labs; CZI funded only 25% of the resulting resource, while the broader community contributed 75%. Vineeta Agarwala’s shorthand: “Come for the annotation, stay for the virtual cell model.”
- Virtual cells could expand biotechnology’s risk budget before expensive wet-lab work begins. Vineeta Agarwala argued that even directional predictions could let researchers test bolder hypotheses in silico before costly experiments, without risking years of work, publication, or tenure. Zuckerberg likened the model to “the new fruit fly” and said a perfect simulation is unnecessary; a directional signal could still be useful. The ambition is human-relevant modeling, tempered by the maxim: “All models are wrong. Some are useful.”
- Precision medicine here means treating common diseases as collections of individually rare biology. Chan’s framing was categorical: “Most diseases should be thought of as rare diseases,” because today’s hypertension and depression treatments still rely heavily on trial and error. Connecting a mutation to downstream cells, protein expression, drug targets, and predicted off-target effects could enable more precise diagnostics and therapies.
- CZI is concentrating its philanthropy around a unified operating Biohub that combines AI, data generation, and biological research. EvolutionaryScale researchers who formerly worked at Meta on protein-folding models are joining a Biohub, with its leader set to run the broader science program. Chan said, “The Biohub is really going to be the main thrust of our philanthropy,” while education and local-community work continue. They did not present centralization as a model for all science; decentralized work and outside labs remain important.
- The next laboratory expansion is computational: Biohub plans to move from roughly 1,000 GPUs toward the 10,000 range. Outside scientists can apply to use that capacity for questions individual labs—with only tens of GPUs—cannot tackle. As Zuckerberg put the allocation constraint, “The GPUs are somewhat zero-sum. The data isn’t.”
Mark Zuckerberg — AI will write most Meta code in 18 months
Meta expects AI agents to write most code for its AI efforts within 12–18 months, moving beyond autocomplete into testing and autonomous improvement. Yet compute, energy, permitting, and human testing capacity remain bottlenecks, while Meta AI’s near-1B monthly users are concentrated outside the US. Monetization hinges on premium compute and product value, with open-source adoption, security, and China’s infrastructure lead unresolved.
View summary (8)
- The headline call: within 12–18 months, “most of the code” going toward Meta’s AI efforts will be written by AI — and Zuckerberg is explicit this isn’t autocomplete: agents that take a goal, run tests, find issues, and write “higher quality code than the average very good person on the team already.” Meta is building these coding and AI-research agents (internal tool: MetaMate) for itself, not as an enterprise product.
- But he pushes back on fast takeoff: intelligence explosions collide with physical-world, human-time bottlenecks — gigawatt clusters, NVIDIA stabilizing new systems, permitting, energy supply chains. His best evidence is internal: Meta’s ads team is already “bottlenecked on compute to run tests,” not on ideas, so AI-generated hypotheses aren’t even marginally useful until they beat the best human ideas already above the testing line.
- Distribution reality check: Meta AI is near 1 billion monthly users, but mostly via WhatsApp outside the US — the US (100M WhatsApp users, iMessage-dominated) is why the standalone Meta AI app exists. The personalization loop — feed, profile, social graph, plus your AI conversations — is “the next thing that’s super exciting.”
- On Maverick ranking #35 on Chatbot Arena: benchmarks are “quite easily gameable” — his team could tune a version “way at the top” but shipped the pure model. Meta will index primarily on product value in Meta AI, rather than leaderboards; note Sonnet 3.7 also isn’t near the top.
- DeepSeek validates chip export controls, in his telling: impressive low-level optimizations were forced by “partially nerfed chips.” Llama 4 is “basically in the same ballpark” on text with a smaller model — lower cost-per-intelligence — and leads on multimodal, which DeepSeek lacks entirely. But if China out-builds the US on power and data centers, “we’ll be at a significant disadvantage.”
- Open-source strategy has a warning embedded: rivals now open-sourcing may be doing it because Meta pushed the trend — “would they still be doing open source if we weren’t doing it?” — and Android’s drift from open to closed is the cautionary precedent. Distillation (90–95% of intelligence at 10% of size) is the killer technique, but language models carry embedded values and models tied to another government could embed exploitable vulnerabilities.
- Monetization: ads for the free consumer tier, a premium tier for “arbitrary amounts of compute” — and on jobs, the contrarian take: AI handling 90% of customer support at one-tenth the cost of a would-be $10–20B call center means Meta will “probably hire more customer support people,” not fewer.
- The engagement thesis for Meta’s core business: the average American has fewer than three friends but the average person has demand for ~15, and feed content goes from video to interactive AI you can talk to or “jump into like a game” — “That’s all going to be AI.”