StarZero All shows Try the podcast clipper →

Jul 11, 2024 · 25m · no-priors

No Priors Ep. 71: The Best of 2024 (so far) with Sarah Guo and Elad Gil

Elad Gil · 4m spoken Dylan Field · 4m spoken Alexandr Wang · 3m spoken Sarah Guo · 3m spoken Emily Glassberg Sands · 2m spoken Brett Adcock · 2m spoken Bill Peebles · 1m spoken Scott Wu · 1s 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

Host Sarah Guo presents a mid-year 2024 retrospective of No Priors, featuring insights from leading AI founders and researchers on fintech automation, design workflows, humanoid robotics, generative video, autonomous coding, and model evaluation.

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 35.4% of the talking time here. How this is scored →

The hosts as informed peer 3.4 Guest teaching 2.9 Guest disagreement 0.3 The hosts pushing back 0.6
05100:0010:0020:000:29–4:21 · The hosts as informed peer 4/10 Emily Glassberg Sands on AI Opportunities in FinTech Elad prompts Emily on non-Stripe fintech AI white space. Emily educates the room on merchant identity verification, compliance nuances with card networks, and automated integration layers.4:22–9:02 · The hosts as informed peer 4/10 Dylan Field on AI Augmentation and Design Workflows Dylan explores the iterative nature of design and dismisses fears of short-term designer elimination. Sarah provides an optimistic counter-take that AI expansion will multiply software volume rather than shrink headcount.9:03–12:42 · The hosts as informed peer 3/10 Brett Adcock on Rapid Robotics Engineering at Figure AI Sarah inquires about hardware-software development velocity. Brett methodically breaks down design review gates and contrasts short software sprint cycles with long-timeline CAD actuator trade studies.12:42–17:42 · The hosts as informed peer 5/10 OpenAI Sora Team on Generative Video and World Modeling Elad demonstrates industry context by drawing direct parallels between Sora's world modeling and Pixar's historical graphics milestones as well as OpenAI's early robotics self-play research.17:42–21:05 · The hosts as informed peer 4/10 Scott Wu on Devin and Future Software Engineering Roles Sarah and Elad question the long-term relevance of traditional computer science education. Scott politely defends foundational CS understanding, arguing that deep knowledge of systems like TCP remains critical despite natural language interfaces.21:05–25:21 · The hosts as informed peer 4/10 Alexandr Wang on AI Evaluations, Benchmarking, and System Trust Sarah prompts Alexandr on evaluation difficulty and references Scale's published work. Alexandr breaks down how academic benchmark contamination skews reported performance and explains GSM-1k.25:21–25:53 · The hosts as informed peer 0/10 Episode Conclusion and Listener Subscription Information Standard monologue housekeeping and outro segment detailing subscription channels and episode show notes.0:29–4:21 · Guest teaching 4/10 Emily Glassberg Sands on AI Opportunities in FinTech Elad prompts Emily on non-Stripe fintech AI white space. Emily educates the room on merchant identity verification, compliance nuances with card networks, and automated integration layers.4:22–9:02 · Guest teaching 3/10 Dylan Field on AI Augmentation and Design Workflows Dylan explores the iterative nature of design and dismisses fears of short-term designer elimination. Sarah provides an optimistic counter-take that AI expansion will multiply software volume rather than shrink headcount.9:03–12:42 · Guest teaching 4/10 Brett Adcock on Rapid Robotics Engineering at Figure AI Sarah inquires about hardware-software development velocity. Brett methodically breaks down design review gates and contrasts short software sprint cycles with long-timeline CAD actuator trade studies.12:42–17:42 · Guest teaching 2/10 OpenAI Sora Team on Generative Video and World Modeling Elad demonstrates industry context by drawing direct parallels between Sora's world modeling and Pixar's historical graphics milestones as well as OpenAI's early robotics self-play research.17:42–21:05 · Guest teaching 3/10 Scott Wu on Devin and Future Software Engineering Roles Sarah and Elad question the long-term relevance of traditional computer science education. Scott politely defends foundational CS understanding, arguing that deep knowledge of systems like TCP remains critical despite natural language interfaces.21:05–25:21 · Guest teaching 4/10 Alexandr Wang on AI Evaluations, Benchmarking, and System Trust Sarah prompts Alexandr on evaluation difficulty and references Scale's published work. Alexandr breaks down how academic benchmark contamination skews reported performance and explains GSM-1k.25:21–25:53 · Guest teaching 0/10 Episode Conclusion and Listener Subscription Information Standard monologue housekeeping and outro segment detailing subscription channels and episode show notes.0:29–4:21 · Guest disagreement 0/10 Emily Glassberg Sands on AI Opportunities in FinTech Elad prompts Emily on non-Stripe fintech AI white space. Emily educates the room on merchant identity verification, compliance nuances with card networks, and automated integration layers.4:22–9:02 · Guest disagreement 1/10 Dylan Field on AI Augmentation and Design Workflows Dylan explores the iterative nature of design and dismisses fears of short-term designer elimination. Sarah provides an optimistic counter-take that AI expansion will multiply software volume rather than shrink headcount.9:03–12:42 · Guest disagreement 0/10 Brett Adcock on Rapid Robotics Engineering at Figure AI Sarah inquires about hardware-software development velocity. Brett methodically breaks down design review gates and contrasts short software sprint cycles with long-timeline CAD actuator trade studies.12:42–17:42 · Guest disagreement 0/10 OpenAI Sora Team on Generative Video and World Modeling Elad demonstrates industry context by drawing direct parallels between Sora's world modeling and Pixar's historical graphics milestones as well as OpenAI's early robotics self-play research.17:42–21:05 · Guest disagreement 1/10 Scott Wu on Devin and Future Software Engineering Roles Sarah and Elad question the long-term relevance of traditional computer science education. Scott politely defends foundational CS understanding, arguing that deep knowledge of systems like TCP remains critical despite natural language interfaces.21:05–25:21 · Guest disagreement 0/10 Alexandr Wang on AI Evaluations, Benchmarking, and System Trust Sarah prompts Alexandr on evaluation difficulty and references Scale's published work. Alexandr breaks down how academic benchmark contamination skews reported performance and explains GSM-1k.25:21–25:53 · Guest disagreement 0/10 Episode Conclusion and Listener Subscription Information Standard monologue housekeeping and outro segment detailing subscription channels and episode show notes.0:29–4:21 · The hosts pushing back 0/10 Emily Glassberg Sands on AI Opportunities in FinTech Elad prompts Emily on non-Stripe fintech AI white space. Emily educates the room on merchant identity verification, compliance nuances with card networks, and automated integration layers.4:22–9:02 · The hosts pushing back 2/10 Dylan Field on AI Augmentation and Design Workflows Dylan explores the iterative nature of design and dismisses fears of short-term designer elimination. Sarah provides an optimistic counter-take that AI expansion will multiply software volume rather than shrink headcount.9:03–12:42 · The hosts pushing back 0/10 Brett Adcock on Rapid Robotics Engineering at Figure AI Sarah inquires about hardware-software development velocity. Brett methodically breaks down design review gates and contrasts short software sprint cycles with long-timeline CAD actuator trade studies.12:42–17:42 · The hosts pushing back 1/10 OpenAI Sora Team on Generative Video and World Modeling Elad demonstrates industry context by drawing direct parallels between Sora's world modeling and Pixar's historical graphics milestones as well as OpenAI's early robotics self-play research.17:42–21:05 · The hosts pushing back 1/10 Scott Wu on Devin and Future Software Engineering Roles Sarah and Elad question the long-term relevance of traditional computer science education. Scott politely defends foundational CS understanding, arguing that deep knowledge of systems like TCP remains critical despite natural language interfaces.21:05–25:21 · The hosts pushing back 0/10 Alexandr Wang on AI Evaluations, Benchmarking, and System Trust Sarah prompts Alexandr on evaluation difficulty and references Scale's published work. Alexandr breaks down how academic benchmark contamination skews reported performance and explains GSM-1k.25:21–25:53 · The hosts pushing back 0/10 Episode Conclusion and Listener Subscription Information Standard monologue housekeeping and outro segment detailing subscription channels and episode show notes.

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

0:00 · the hosts 37% · guest 63%0:00 · the hosts 37% · guest 63%3:00 · the hosts 22.3% · guest 77.7%3:00 · the hosts 22.3% · guest 77.7%6:00 · the hosts 14.9% · guest 85.1%6:00 · the hosts 14.9% · guest 85.1%9:00 · the hosts 22% · guest 78%9:00 · the hosts 22% · guest 78%12:00 · the hosts 20% · guest 80%12:00 · the hosts 20% · guest 80%15:00 · the hosts 49.7% · guest 50.3%15:00 · the hosts 49.7% · guest 50.3%18:00 · the hosts 99.6% · guest 0.4%18:00 · the hosts 99.6% · guest 0.4%21:00 · the hosts 22% · guest 78%21:00 · the hosts 22% · guest 78%24:00 · the hosts 28.1% · guest 71.9%24:00 · the hosts 28.1% · guest 71.9%
Sharpest disagreement ▶ 6:42 Dylan rejects immediate designer replacement narrative

Dylan reframes Sarah's question by rejecting the premise that AI will replace designers in the near term, citing the complex emotional and contextual constraints of design.

 Hardest push from the hosts ▶ 8:05 Sarah asserts optimistic thesis on software volume expansion

Sarah intervenes to offer an alternative thesis that market dynamics will lead to higher volumes of software rather than a decrease in designer headcount.

 Biggest teaching moment ▶ 23:00 Alexandr details model benchmark leakage and GSM-1k findings

Alexandr walks through why mainstream LLM evaluations fail due to dataset contamination and how Scale's held-out math evaluations revealed significant drops in real capability.

 The host holds their own ▶ 16:22 Elad frames Sora through Pixar evolution and robotics history

Elad demonstrates domain grasp by synthesizing Sora's physics simulation with early OpenAI robotic self-play and Pixar's 30-year simulation trajectory.
the scores for every segment, with the reasoning behind each
ChapterTopicThe hosts as informed peerGuest teachingGuest disagreementThe hosts pushing backWhy
Emily Glassberg Sands on AI Opportunities in FinTech 4400 Elad prompts Emily on non-Stripe fintech AI white space. Emily educates the room on merchant identity verification, compliance nuances with card networks, and automated integration layers.
Dylan Field on AI Augmentation and Design Workflows 4312 Dylan explores the iterative nature of design and dismisses fears of short-term designer elimination. Sarah provides an optimistic counter-take that AI expansion will multiply software volume rather than shrink headcount.
Brett Adcock on Rapid Robotics Engineering at Figure AI 3400 Sarah inquires about hardware-software development velocity. Brett methodically breaks down design review gates and contrasts short software sprint cycles with long-timeline CAD actuator trade studies.
OpenAI Sora Team on Generative Video and World Modeling 5201 Elad demonstrates industry context by drawing direct parallels between Sora's world modeling and Pixar's historical graphics milestones as well as OpenAI's early robotics self-play research.
Scott Wu on Devin and Future Software Engineering Roles 4311 Sarah and Elad question the long-term relevance of traditional computer science education. Scott politely defends foundational CS understanding, arguing that deep knowledge of systems like TCP remains critical despite natural language interfaces.
Alexandr Wang on AI Evaluations, Benchmarking, and System Trust 4400 Sarah prompts Alexandr on evaluation difficulty and references Scale's published work. Alexandr breaks down how academic benchmark contamination skews reported performance and explains GSM-1k.
Episode Conclusion and Listener Subscription Information 0000 Standard monologue housekeeping and outro segment detailing subscription channels and episode show notes.

Statements from this episode (7)

Opinion
Sands: AI presents a major opportunity to automate complex financial integrations
“I think there's almost certainly an opportunity to, you know, whether Stripe does it or somebody else does it, to make sort of financial integrations way more seamless.”
Emily Glassberg Sands Jul 11, 2024 ▶ 2:28
Insight
Field: Design requires iterative AI loops because prompts cannot capture full context
“In the design context one thing that really matters a lot is the iterative loop and being able to keep going back and forth to an agent and give more instructions over time. If you just kind of like go to first principles here, there's so much that you're not …”
Dylan Field Jul 11, 2024 ▶ 5:05
Prediction Not checkable as stated
Field: Engineers will spend more time on design than coding tasks
“Probably before you see potential replacement of any part of the design role, you instead see augmentation and you see access. You see efficiency so that designers can get more done. And I think probably a lot of engineers do more of their, put more of their t…”
Dylan Field Jul 11, 2024 ▶ 7:26
Insight
Adcock: Hardware development succeeds through iterative testing over theoretical analysis
“From like a thesis perspective, I strongly believe in like an iterative design approach. We really don't believe on spending a lot of time, like just doing research and analyzing. We spend a lot of time on just testing, building the testing here.”
Brett Adcock Jul 11, 2024 ▶ 9:36
Prediction Not checkable as stated
Peebles: Training AI on raw video is essential for robotics
“There's so much you learn from video, which you don't necessarily get from other modalities, which companies like OpenAI have invested a lot in the past, like language, you know, like the minutia of like how arms and joints move through space, you know, again,…”
Bill Peebles Jul 11, 2024 ▶ 17:15
Prediction Not checkable as stated
Gil: Software engineers in 5–10 years will resemble architects and product managers
“I think the role of A software engineer five or 10 years from now, it looks something like a mix between a technical architect and a product manager today, you know, where, where a lot of what you do is, you know, you take problems that you're facing or that y…”
Elad Gil Jul 11, 2024 ▶ 19:11
Assertion Supported
Wang: Standard academic benchmarks are contaminated by training data overfitting
“Most of the benchmarks that we as a community look at... Academic benchmarks that are what the industry used to measure the performance of these algorithms are fraught with issues. Many of the models are overfit on these benchmarks. They're sort of in the trai…”
Alexandr Wang Jul 11, 2024 ▶ 23:21
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.

Try the podcast clipper Learn more