Feb 5, 2026 · 43m · no-priors

How AI Will Transform Roblox Games into Photorealistic Worlds | CEO David Baszucki

David Baszucki · 32m spoken Sarah Guo · 4m spoken Elad Gil · 3m spoken
0:00 / 0:00
▶ Watch on YouTube →

gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions

Roblox founder and CEO David Baszucki joins No Priors to discuss the convergence of artificial intelligence, physics simulation, and creator tools in realizing the platform's multi-decade vision of a shared, photorealistic virtual holodeck.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 18.9% of the talking time here. How this is scored →

The hosts as informed peer 4.1 Guest teaching 4.4 Guest disagreement 0.8 The hosts pushing back 0.7
05100:0015:0030:000:00–5:28 · The hosts as informed peer 3/10 Episode Hook: Human-AI Relationships and Long-Term Vision The hosts set up a broad opening question exploring the 10-year horizon of AI in gaming. David Baszucki outlines Roblox's 20-year-old original business plan, contrasting real-time dreaming with high-fidelity social simulation.5:29–8:23 · The hosts as informed peer 3/10 Immersive 4D Simulation and Future Virtual Conferencing Elad asks about enterprise applications of virtual spaces versus Zoom. Baszucki educates the hosts on the acoustical and synchronization physics challenges of 4D simulations, such as acoustic attenuation and time-forward audio extrapolation for group singing.8:23–14:11 · The hosts as informed peer 5/10 Expanding Physics Engines, Acoustic Simulation, and Platform Diversity Sarah challenges the tension between physics simulation and the stylized low-fidelity reputation of Roblox games. Baszucki explains the underlying vector history format, contrasting it with video raster world models and highlighting their 13 billion monthly hours of 3D interaction data.14:12–18:04 · The hosts as informed peer 6/10 Building Native AI NPCs and Agentic Virtual Doppelgängers Baszucki details how Roblox trains native NPC models and virtual doppelgangers. Elad demonstrates sharp technical domain expertise by drawing a direct parallel between end-to-end deep learning in autonomous driving and modern NPC architecture.18:05–22:17 · The hosts as informed peer 5/10 The Evolving Role of Game Creators and Multi-AI Tech Stacks Sarah inquires about the future role of game designers and pure video-generative world models without physics engines. Baszucki defends a hybrid multi-AI tech stack and argues human creators will achieve vastly higher leverage rather than being replaced by AI slop.22:18–27:26 · The hosts as informed peer 4/10 AI Companions, Social Simulation, and Relationship Boundaries Elad and Sarah bring up AI companions and dinner-party questions about human-AI relationships. Baszucki dismisses the near-term likelihood of crossing the human-AI romantic threshold while exploring agentic social simulations and AI coaching.27:27–31:43 · The hosts as informed peer 4/10 Generative Asset Creation, Dynamic LOD, and Roblox Studio AI Sarah asks why cheaper generative assets haven't yet disrupted gaming economics. Baszucki explains consumer quality expectations rise at equal velocity and details Roblox's dynamic LOD and cloud-based 3D procedural upsampling pipeline.31:43–33:57 · The hosts as informed peer 3/10 Creator Economics, Live Ops, and Transparent Discovery Sarah asks about creator lifecycles on the platform. Baszucki breaks down creator revenue distribution, the necessity of continuous live-ops updates, and the importance of algorithmic transparency in discovery.33:58–42:01 · The hosts as informed peer 4/10 Long-Term Vision, Goal Setting, and Gamified Meritocratic Hiring Sarah and Elad ask about long-term vision execution and meritocratic hiring. Baszucki describes acquiring Embellus to run simulation-based hiring assessments on Roblox, revealing that their data shows little correlation between traditional elite university credentials and practical problem-solving capability.0:00–5:28 · Guest teaching 4/10 Episode Hook: Human-AI Relationships and Long-Term Vision The hosts set up a broad opening question exploring the 10-year horizon of AI in gaming. David Baszucki outlines Roblox's 20-year-old original business plan, contrasting real-time dreaming with high-fidelity social simulation.5:29–8:23 · Guest teaching 5/10 Immersive 4D Simulation and Future Virtual Conferencing Elad asks about enterprise applications of virtual spaces versus Zoom. Baszucki educates the hosts on the acoustical and synchronization physics challenges of 4D simulations, such as acoustic attenuation and time-forward audio extrapolation for group singing.8:23–14:11 · Guest teaching 6/10 Expanding Physics Engines, Acoustic Simulation, and Platform Diversity Sarah challenges the tension between physics simulation and the stylized low-fidelity reputation of Roblox games. Baszucki explains the underlying vector history format, contrasting it with video raster world models and highlighting their 13 billion monthly hours of 3D interaction data.14:12–18:04 · Guest teaching 4/10 Building Native AI NPCs and Agentic Virtual Doppelgängers Baszucki details how Roblox trains native NPC models and virtual doppelgangers. Elad demonstrates sharp technical domain expertise by drawing a direct parallel between end-to-end deep learning in autonomous driving and modern NPC architecture.18:05–22:17 · Guest teaching 4/10 The Evolving Role of Game Creators and Multi-AI Tech Stacks Sarah inquires about the future role of game designers and pure video-generative world models without physics engines. Baszucki defends a hybrid multi-AI tech stack and argues human creators will achieve vastly higher leverage rather than being replaced by AI slop.22:18–27:26 · Guest teaching 3/10 AI Companions, Social Simulation, and Relationship Boundaries Elad and Sarah bring up AI companions and dinner-party questions about human-AI relationships. Baszucki dismisses the near-term likelihood of crossing the human-AI romantic threshold while exploring agentic social simulations and AI coaching.27:27–31:43 · Guest teaching 5/10 Generative Asset Creation, Dynamic LOD, and Roblox Studio AI Sarah asks why cheaper generative assets haven't yet disrupted gaming economics. Baszucki explains consumer quality expectations rise at equal velocity and details Roblox's dynamic LOD and cloud-based 3D procedural upsampling pipeline.31:43–33:57 · Guest teaching 4/10 Creator Economics, Live Ops, and Transparent Discovery Sarah asks about creator lifecycles on the platform. Baszucki breaks down creator revenue distribution, the necessity of continuous live-ops updates, and the importance of algorithmic transparency in discovery.33:58–42:01 · Guest teaching 5/10 Long-Term Vision, Goal Setting, and Gamified Meritocratic Hiring Sarah and Elad ask about long-term vision execution and meritocratic hiring. Baszucki describes acquiring Embellus to run simulation-based hiring assessments on Roblox, revealing that their data shows little correlation between traditional elite university credentials and practical problem-solving capability.0:00–5:28 · Guest disagreement 1/10 Episode Hook: Human-AI Relationships and Long-Term Vision The hosts set up a broad opening question exploring the 10-year horizon of AI in gaming. David Baszucki outlines Roblox's 20-year-old original business plan, contrasting real-time dreaming with high-fidelity social simulation.5:29–8:23 · Guest disagreement 0/10 Immersive 4D Simulation and Future Virtual Conferencing Elad asks about enterprise applications of virtual spaces versus Zoom. Baszucki educates the hosts on the acoustical and synchronization physics challenges of 4D simulations, such as acoustic attenuation and time-forward audio extrapolation for group singing.8:23–14:11 · Guest disagreement 1/10 Expanding Physics Engines, Acoustic Simulation, and Platform Diversity Sarah challenges the tension between physics simulation and the stylized low-fidelity reputation of Roblox games. Baszucki explains the underlying vector history format, contrasting it with video raster world models and highlighting their 13 billion monthly hours of 3D interaction data.14:12–18:04 · Guest disagreement 0/10 Building Native AI NPCs and Agentic Virtual Doppelgängers Baszucki details how Roblox trains native NPC models and virtual doppelgangers. Elad demonstrates sharp technical domain expertise by drawing a direct parallel between end-to-end deep learning in autonomous driving and modern NPC architecture.18:05–22:17 · Guest disagreement 1/10 The Evolving Role of Game Creators and Multi-AI Tech Stacks Sarah inquires about the future role of game designers and pure video-generative world models without physics engines. Baszucki defends a hybrid multi-AI tech stack and argues human creators will achieve vastly higher leverage rather than being replaced by AI slop.22:18–27:26 · Guest disagreement 2/10 AI Companions, Social Simulation, and Relationship Boundaries Elad and Sarah bring up AI companions and dinner-party questions about human-AI relationships. Baszucki dismisses the near-term likelihood of crossing the human-AI romantic threshold while exploring agentic social simulations and AI coaching.27:27–31:43 · Guest disagreement 1/10 Generative Asset Creation, Dynamic LOD, and Roblox Studio AI Sarah asks why cheaper generative assets haven't yet disrupted gaming economics. Baszucki explains consumer quality expectations rise at equal velocity and details Roblox's dynamic LOD and cloud-based 3D procedural upsampling pipeline.31:43–33:57 · Guest disagreement 0/10 Creator Economics, Live Ops, and Transparent Discovery Sarah asks about creator lifecycles on the platform. Baszucki breaks down creator revenue distribution, the necessity of continuous live-ops updates, and the importance of algorithmic transparency in discovery.33:58–42:01 · Guest disagreement 1/10 Long-Term Vision, Goal Setting, and Gamified Meritocratic Hiring Sarah and Elad ask about long-term vision execution and meritocratic hiring. Baszucki describes acquiring Embellus to run simulation-based hiring assessments on Roblox, revealing that their data shows little correlation between traditional elite university credentials and practical problem-solving capability.0:00–5:28 · The hosts pushing back 0/10 Episode Hook: Human-AI Relationships and Long-Term Vision The hosts set up a broad opening question exploring the 10-year horizon of AI in gaming. David Baszucki outlines Roblox's 20-year-old original business plan, contrasting real-time dreaming with high-fidelity social simulation.5:29–8:23 · The hosts pushing back 0/10 Immersive 4D Simulation and Future Virtual Conferencing Elad asks about enterprise applications of virtual spaces versus Zoom. Baszucki educates the hosts on the acoustical and synchronization physics challenges of 4D simulations, such as acoustic attenuation and time-forward audio extrapolation for group singing.8:23–14:11 · The hosts pushing back 2/10 Expanding Physics Engines, Acoustic Simulation, and Platform Diversity Sarah challenges the tension between physics simulation and the stylized low-fidelity reputation of Roblox games. Baszucki explains the underlying vector history format, contrasting it with video raster world models and highlighting their 13 billion monthly hours of 3D interaction data.14:12–18:04 · The hosts pushing back 0/10 Building Native AI NPCs and Agentic Virtual Doppelgängers Baszucki details how Roblox trains native NPC models and virtual doppelgangers. Elad demonstrates sharp technical domain expertise by drawing a direct parallel between end-to-end deep learning in autonomous driving and modern NPC architecture.18:05–22:17 · The hosts pushing back 1/10 The Evolving Role of Game Creators and Multi-AI Tech Stacks Sarah inquires about the future role of game designers and pure video-generative world models without physics engines. Baszucki defends a hybrid multi-AI tech stack and argues human creators will achieve vastly higher leverage rather than being replaced by AI slop.22:18–27:26 · The hosts pushing back 1/10 AI Companions, Social Simulation, and Relationship Boundaries Elad and Sarah bring up AI companions and dinner-party questions about human-AI relationships. Baszucki dismisses the near-term likelihood of crossing the human-AI romantic threshold while exploring agentic social simulations and AI coaching.27:27–31:43 · The hosts pushing back 1/10 Generative Asset Creation, Dynamic LOD, and Roblox Studio AI Sarah asks why cheaper generative assets haven't yet disrupted gaming economics. Baszucki explains consumer quality expectations rise at equal velocity and details Roblox's dynamic LOD and cloud-based 3D procedural upsampling pipeline.31:43–33:57 · The hosts pushing back 0/10 Creator Economics, Live Ops, and Transparent Discovery Sarah asks about creator lifecycles on the platform. Baszucki breaks down creator revenue distribution, the necessity of continuous live-ops updates, and the importance of algorithmic transparency in discovery.33:58–42:01 · The hosts pushing back 1/10 Long-Term Vision, Goal Setting, and Gamified Meritocratic Hiring Sarah and Elad ask about long-term vision execution and meritocratic hiring. Baszucki describes acquiring Embellus to run simulation-based hiring assessments on Roblox, revealing that their data shows little correlation between traditional elite university credentials and practical problem-solving capability.

speaking balance: gold is the hosts, purple is the guest (3 minute bins)

0:00 · the hosts 37.5% · guest 62.5%0:00 · the hosts 37.5% · guest 62.5%3:00 · the hosts 7.6% · guest 92.4%3:00 · the hosts 7.6% · guest 92.4%6:00 · the hosts 15.4% · guest 84.6%6:00 · the hosts 15.4% · guest 84.6%9:00 · the hosts 6.3% · guest 93.7%9:00 · the hosts 6.3% · guest 93.7%12:00 · the hosts 1.1% · guest 98.9%12:00 · the hosts 1.1% · guest 98.9%15:00 · the hosts 17.7% · guest 82.3%15:00 · the hosts 17.7% · guest 82.3%18:00 · the hosts 13.7% · guest 86.3%18:00 · the hosts 13.7% · guest 86.3%21:00 · the hosts 26.2% · guest 73.8%21:00 · the hosts 26.2% · guest 73.8%24:00 · the hosts 30.8% · guest 69.2%24:00 · the hosts 30.8% · guest 69.2%27:00 · the hosts 21.9% · guest 78.1%27:00 · the hosts 21.9% · guest 78.1%30:00 · the hosts 6% · guest 94%30:00 · the hosts 6% · guest 94%33:00 · the hosts 23.1% · guest 76.9%33:00 · the hosts 23.1% · guest 76.9%36:00 · the hosts 27.3% · guest 72.7%36:00 · the hosts 27.3% · guest 72.7%39:00 · the hosts 21.3% · guest 78.7%39:00 · the hosts 21.3% · guest 78.7%42:00 · the hosts 35.5% · guest 64.5%42:00 · the hosts 35.5% · guest 64.5%
Sharpest disagreement ▶ 26:43 Skepticism on Human-AI Relationships

Baszucki directly rejects the premise that children will form serious relationships with AI in the near future, putting the likelihood at zero and emphasizing the enduring barrier between human and machine.

Hardest push from the hosts ▶ 8:23 Challenging Roblox's Reputation for Realism

Sarah Guo pushes back on Baszucki's emphasis on high-fidelity simulation by noting Roblox has historically been known for simple blocky graphics rather than realistic physics.

Biggest teaching moment ▶ 11:40 Vector-Based Simulation vs. Raster Video Training

Baszucki educates the hosts on the fundamental architectural difference between 2D raster video models and Roblox's 13-billion-hour 3D vector dataset which allows unlimited camera angle replays.

The host holds their own ▶ 17:28 Autonomous Vehicle Deep Learning Analogy

Elad demonstrates sharp AI systems expertise by connecting Roblox's transition to end-to-end NPC models with the historical architecture shift seen in autonomous vehicle perception stacks.

the scores for every segment, with the reasoning behind each
ChapterTopicThe hosts as informed peerGuest teachingGuest disagreementThe hosts pushing backWhy
Episode Hook: Human-AI Relationships and Long-Term Vision 3410 The hosts set up a broad opening question exploring the 10-year horizon of AI in gaming. David Baszucki outlines Roblox's 20-year-old original business plan, contrasting real-time dreaming with high-fidelity social simulation.
Immersive 4D Simulation and Future Virtual Conferencing 3500 Elad asks about enterprise applications of virtual spaces versus Zoom. Baszucki educates the hosts on the acoustical and synchronization physics challenges of 4D simulations, such as acoustic attenuation and time-forward audio extrapolation for group singing.
Expanding Physics Engines, Acoustic Simulation, and Platform Diversity 5612 Sarah challenges the tension between physics simulation and the stylized low-fidelity reputation of Roblox games. Baszucki explains the underlying vector history format, contrasting it with video raster world models and highlighting their 13 billion monthly hours of 3D interaction data.
Building Native AI NPCs and Agentic Virtual Doppelgängers 6400 Baszucki details how Roblox trains native NPC models and virtual doppelgangers. Elad demonstrates sharp technical domain expertise by drawing a direct parallel between end-to-end deep learning in autonomous driving and modern NPC architecture.
The Evolving Role of Game Creators and Multi-AI Tech Stacks 5411 Sarah inquires about the future role of game designers and pure video-generative world models without physics engines. Baszucki defends a hybrid multi-AI tech stack and argues human creators will achieve vastly higher leverage rather than being replaced by AI slop.
AI Companions, Social Simulation, and Relationship Boundaries 4321 Elad and Sarah bring up AI companions and dinner-party questions about human-AI relationships. Baszucki dismisses the near-term likelihood of crossing the human-AI romantic threshold while exploring agentic social simulations and AI coaching.
Generative Asset Creation, Dynamic LOD, and Roblox Studio AI 4511 Sarah asks why cheaper generative assets haven't yet disrupted gaming economics. Baszucki explains consumer quality expectations rise at equal velocity and details Roblox's dynamic LOD and cloud-based 3D procedural upsampling pipeline.
Creator Economics, Live Ops, and Transparent Discovery 3400 Sarah asks about creator lifecycles on the platform. Baszucki breaks down creator revenue distribution, the necessity of continuous live-ops updates, and the importance of algorithmic transparency in discovery.
Long-Term Vision, Goal Setting, and Gamified Meritocratic Hiring 4511 Sarah and Elad ask about long-term vision execution and meritocratic hiring. Baszucki describes acquiring Embellus to run simulation-based hiring assessments on Roblox, revealing that their data shows little correlation between traditional elite university credentials and practical problem-solving capability.

Statements from this episode (19)

Assertion Not checkable as stated
Baszucki: 20-Year-Old Roblox Deck Predicted Category of Human Co-Existence
“So there is a Roblox business plan PowerPoint deck from almost 20 years ago that we pull out sometimes and we're amazed at the fidelity of it. It pretends a new category. One could call it the category of human co-existence. It's got the sizes of a bunch of co…”
David Baszucki Feb 5, 2026 ▶ 1:47
Insight
Baszucki: AI Worlds Will Span Shared Holodecks to Solo Real-Time Dreaming
“And so if you put those at two extremes, one is communication platform, high fidelity with people, you know, the other is real time dreaming where everyone's an NPC. Everything in between is possible. And I think we see everything in between ultimately as bein…”
David Baszucki Feb 5, 2026 ▶ 4:38
Prediction Not checkable as stated
Baszucki: Photorealistic 4D simulation will replace standard 2D video conferencing
“I think as a technology for multiplayer, four D simulation gets better and more photo realistic. It's almost going to be like video is the down sampling. And it's like, if we were having this for the immersive communication, we'd say go to legacy analog mode a…”
David Baszucki Feb 5, 2026 ▶ 6:19
Disclosure
Baszucki: Roblox is adding full acoustic simulation and echoes to physics engine
“We are just started talking about full acoustic simulation. I just tweeted about it. So actually, you know, approximate, you know, attenuation and all of this kind of things and echoes that's going to start happening on the physics simulator.”
David Baszucki Feb 5, 2026 ▶ 9:06
Prediction Not checkable as stated
Baszucki: Synchronizing state for 10,000 players will unravel in 5–10 years
“The technology for this, I think we're going to start unraveling it over the next five to 10 years is what's the most efficient way To synchronize the state of 10,000 people.”
David Baszucki Feb 5, 2026 ▶ 10:53
Disclosure
Roblox will store its entire platform history as 3D vector data
“I will share that the format we're using part of the vision for us is to ultimately store the history Of everything on Roblox, and to store it, not raster like video, but store it vector, which means the ability to play back anything that's ever happened on Ro…”
David Baszucki Feb 5, 2026 ▶ 11:50
Disclosure
Baszucki: Roblox is training native AI NPCs to navigate and play games
“What we are starting to train is NPCs based on a combination of all of the data we have in a privacy compliant way that are starting to navigate and starting to play games.”
David Baszucki Feb 5, 2026 ▶ 14:23
Opinion
Baszucki: AI NPC platforms must run inference via cloud rather than locally
“I think to make that happen, these platforms have to be very cloud connected platforms rather than running local to spin up inference on all of these things.”
David Baszucki Feb 5, 2026 ▶ 17:02
Opinion
Baszucki: AI slop will not take over game creation from humans
“I think, you know, I'm not to the point where I believe, you know, what we call AI slop today is going to take over. Like, I think the human touch is Got a long runway on it.”
David Baszucki Feb 5, 2026 ▶ 18:40
Prediction Not checkable as stated
Baszucki: Five-person teams will produce astounding game quality with AI
“So, but I do think the role will be more leveraged and the quality coming out of five people will be astounding, you know, but we'll come to expect that quality. So quality will go up.”
David Baszucki Feb 5, 2026 ▶ 19:00
Prediction Not checkable as stated
Baszucki: World-model game architectures will not be all-in-one models
“And I think that the way we would see the ultimate architectures of these platforms is Probably for a while, not an all in one thing. We would probably see some hyper efficient thousand person synchronization state engine.”
David Baszucki Feb 5, 2026 ▶ 20:37
Opinion
Baszucki: Zero percent of today's children will have serious AI relationships
“I think zero right now. I do. I don't think we're close to crossing the human AI barrier.”
David Baszucki Feb 5, 2026 ▶ 26:43
Insight
Baszucki: Cheaper asset creation is matched by rising consumer quality expectations
“One could say no matter how cheap it is to create assets, the expectation of quality from consumers goes up at exactly the same velocity.”
David Baszucki Feb 5, 2026 ▶ 27:49
Opinion
Baszucki: Gaming tech must move to vertically integrated cloud architecture
“That the technology for the gaming industry has to move to cloud vertically integrated and asset management can't just ship on a DVD or be a giant download.”
David Baszucki Feb 5, 2026 ▶ 28:07
Assertion Supported
Baszucki: Top Roblox Creators Make $30M+ Annually
“The top Roblox creators now are well into the tens and tens and thirties plus million a year. Like it's really serious stuff.”
David Baszucki Feb 5, 2026 ▶ 31:46
Assertion Supported
Baszucki: Roblox's 1,000th creator's earnings are growing faster than top earners
“We have seen in the Roblox creator community, I think a healthy sign that creator number 1000 and what is their average yearly revenue is going up faster than creator number one. So that's a bit of a healthy long tail.”
David Baszucki Feb 5, 2026 ▶ 32:00
Disclosure
Roblox targets capturing 10% of global gaming content and 300M DAUs
“So we took a step, a stepping stone and said, we want to get to 10% of global gaming content. On the way to that. It's actually a remarkably good step. It's about three hundred million DAUs. It's about twenty billion. It's about three X where we are.”
David Baszucki Feb 5, 2026 ▶ 37:00
Insight
Baszucki: Software CEOs must maintain both 3X and 10X targets
“I think for a software CEO, Always knowing both three X and 10 X is very helpful, because if you don't know 10 X, then it's hard to sleep at night. If you don't know three X, it's hard to plan forensically how we're going to operationalize that, how quickly we…”
David Baszucki Feb 5, 2026 ▶ 37:29
Assertion Not checkable as stated
Baszucki: Roblox ignores university pedigree as top talent bypasses elite schools
“Well, I think people typically think like I went to Stanford. That's where all the top talent is. We have found that not necessarily to be true. And we have found community college, you know, the small Midwestern engineering school, like, because we're assessi…”
David Baszucki Feb 5, 2026 ▶ 40:19
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 100 episodes transcribed, diarized, checked and made playable, runs on the StarZero media pipeline. Drop in your own episode and the podcast clipper finds the moments worth sharing, cuts them, captions them, and reframes them for every feed.