Oct 31, 2025 · 18m · no-priors

No Priors Ep. 138 | The Best of 2025 (So Far) with Sarah Guo and Elad Gil

Sarah Guo · 2m spoken Arvind Jain · 2m spoken Noubar Afeyan · 2m spoken Dr. Fei-Fei Li · 1m spoken Dan Hendrycks · 1m spoken Winston Weinberg · 1m spoken Eric Mitchell · 59s spoken Isa Fulford · 42s 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

This special retrospective of No Priors highlights the defining AI breakthroughs of 2025, bringing together leading researchers, founders, and scientists to explore the frontiers of model reasoning, enterprise automation, labor dynamics, and geopolitical strategy.

Chapters · click to jump

2 statements ▾

Foody: AI job displacement will happen rapidly and cause political fallout open →

Foody: Physical world automation will lag far behind digital AI automation open →

1 statement ▾

Hendrycks: U.S. and China may launch preemptive cyberattacks on data centers open →

3 statements ▾

Guo: Reasoning is the biggest AI architecture shift since the transformer open →

McKinzie: Tool use noticeably changes test-time scaling for visual reasoning open →

Mitchell: LLMs allocate compute efficiently by deferring tasks to specialized tools open →

1 statement ▾

Jain: Glean's largest customer holds over 1 billion internal documents open →

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 14.3% of the talking time here. How this is scored →

The hosts as informed peer 1.6 Guest teaching 2.6 Guest disagreement 0.1 The hosts pushing back 0.3

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

Sharpest disagreement ▶ 9:44 Afeyan rejects the 'shots on goal' venture mindset

Afeyan pushes back against the conventional venture capital narrative that high failure rates and random wins are acceptable in critical sectors.

Hardest push from the hosts ▶ 9:34 Sarah Guo presses on Afeyan's terminology

Sarah directly questions Afeyan's framing of entrepreneurship as 'gamey' and attempts to define it as merely random.

Biggest teaching moment ▶ 9:44 Afeyan educates on professionalizing venture creation

Afeyan corrects Sarah's simplistic 'it's random' summary, explaining how systematic engineering must replace gamified trial and error.

The host holds their own ▶ 13:49 Sarah Guo synthesizes market evolution thesis

Sarah articulates a core VC investment principle regarding how new technological waves reopen supposedly dead software categories.

the scores for every segment, with the reasoning behind each

Topic	The hosts as informed peer	Guest teaching	Guest disagreement	The hosts pushing back	Why
Harvey and the Accidental Discovery of Legal AI	2	3	0	0	Sarah provides polished framing for clips from Winston Weinberg and Dr. Fei-Fei Li. The guests detail the genesis of legal prompting and the cognitive evolution of spatial intelligence without any friction.
Brendan Foody on AI Labor Displacement and the Future of Work	1	3	0	0	Brendan Foody outlines labor displacement scenarios and Dan Hendrycks presents a strategic nuclear deterrence comparison for AI superintelligence. The tone is purely analytical and explanatory.
Nubar Afeyan on Transforming Entrepreneurship into a Science	3	4	1	2	Sarah actively probes Afeyan's critique of venture culture by asking what he means by 'gamey' and suggesting 'it's random'. Afeyan corrects her reduction, explaining that it goes beyond randomness to an inappropriate reliance on luck in serious fields like healthcare.
Brandon McKinzie and Eric Mitchell on Reasoning Architectures and Tool Use	1	3	0	0	Brandon McKinzie and Eric Mitchell explain test-time scaling, self-uncertainty estimation, and compute allocation via tool use. Sarah's intro is brief and the segment consists of technical exposition.
Issa Fulford on the Breakthrough of Deep Research	3	3	0	0	Sarah connects Issa Fulford's Deep Research breakthrough with an intuitive analogy, and introduces Arvind Jain's enterprise search history with an investment thesis premise on SaaS enabling graveyard markets.
Shiv Rao on the Human and Clinical Impact of Abridge	1	2	0	0	Shiv Rao describes doctor feedback and emotional retention stories at Abridge. Sarah sets up the clip respectfully with no debate or counter-framing.
Sarah Guo Wraps Up the Best of 2025	0	0	0	0	Sarah delivers an episode wrap-up and housekeeping monologue summarizing the themes of 2025 AI developments.

Statements from this episode (7)

Prediction Not checkable as stated

Foody: AI job displacement will happen rapidly and cause political fallout

“I think displacement in a lot of roles is going to happen very quickly, and it's going to be very painful and a large political problem.”

Brendan Foody Oct 31, 2025 ▶ 4:25

Prediction Not checkable as stated

Foody: Physical world automation will lag far behind digital AI automation

“I think that automation In the physical world is going to happen a lot slower than what's happening in the digital world, just because of so many of the, like, self-reinforcing gains and a lot of, yeah, self-improvement that can happen in, in the virtual world…”

Brendan Foody Oct 31, 2025 ▶ 5:48

Prediction Not checkable as stated

Hendrycks: U.S. and China may launch preemptive cyberattacks on data centers

“I think that later on, it becomes so destabilizing that China just says, we're going to do something preemptive, like do a cyber attack on your data center. And the U.S. Might do that to China.”

Dan Hendrycks Oct 31, 2025 ▶ 7:24

Opinion

Guo: Reasoning is the biggest AI architecture shift since the transformer

“Reasoning is the biggest paradigm shift in AI architecture since the transformer.”

Sarah Guo Oct 31, 2025 ▶ 10:38

Assertion Not checkable as stated

McKinzie: Tool use noticeably changes test-time scaling for visual reasoning

“We've seen exactly that, like the test time scaling slopes for, without tool use and with tool use for visual reasoning specifically are very noticeably different.”

Brandon McKinzie Oct 31, 2025 ▶ 11:31

Insight

Mitchell: LLMs allocate compute efficiently by deferring tasks to specialized tools

“I think, like, part of this is you can just allocate compute a lot more efficiently because you can defer stuff that the model doesn't have comparative advantage to doing to a tool that is, like, really well suited to doing that thing.”

Eric Mitchell Oct 31, 2025 ▶ 12:28

Assertion Not checkable as stated

Jain: Glean's largest customer holds over 1 billion internal documents

“One of our largest customers. They have more than one billion documents inside their company. Now hear this, you know, when Alain and I, you know, when we were working on search, At Google, you know, in 2004, the entire internet was actually one billion docume…”

Arvind Jain Oct 31, 2025 ▶ 15:56

No Priors Ep. 138 | The Best of 2025 (So Far) with Sarah Guo and Elad Gil

How this conversation actually went

Statements from this episode (7)

Turn any episode into a week of clips.