StarZero All shows Try the podcast clipper →

Apr 8, 2025 · 15m · tbpn

Is AI UNDERVALUED? | Scott Wu on TBPN April 3rd

Scott Wu · 9m 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

Cognition CEO Scott Wu discusses the launch and general availability of Devin's agent-native IDE, sharing insights on AI market valuation, the future of software engineering careers, and the technical challenges of deploying autonomous coding agents in complex enterprise environments.

How this conversation actually went

Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. How this is scored →

The hosts as informed peer 5.6 Guest teaching 2.0 Guest disagreement 0.6 The hosts pushing back 0.6
05100:0010:000:00–4:44 · The hosts as informed peer 7/10 Devin's Agent-Native IDE Launch and Features The host demonstrates strong domain knowledge by quoting Sam Altman's 10-year-old 'Bubble Talk' blog post in precise detail, including specific unicorn names and valuation targets. Scott Wu readily agrees with the host's historical premise and expands on why AI is actually undervalued rather than in a bubble.4:44–7:25 · The hosts as informed peer 6/10 Vibe Coding and the Mass Adoption of AI Programming The host references Karpathy's 'vibe coding' concept, Levels.io, and viral generative AI moments to ask about mass adoption. Scott provides an educational breakdown separating single-use hobbyist creation from deep enterprise software architecture.7:25–11:08 · The hosts as informed peer 4/10 Engineering Advice and the Evolution of Coding Interfaces The co-host poses broad questions using analogies about engineering talent and adoption lag. Scott grounds the discussion by explaining the transition of AI interfaces from text completion to agentic systems and the real-world friction of legacy codebases like COBOL.11:08–13:30 · The hosts as informed peer 6/10 Multimodality and Iterative Debugging in AI Systems The host articulates an insightful hypothesis regarding how cross-modal and multilingual data ingestion boosts model capability. Scott explains how multimodal capabilities mirror practical human developer workflows like visual iteration and log inspection.13:31–15:31 · The hosts as informed peer 5/10 AI Benchmarks: International Mathematical Olympiad and Competitive Programming The hosts bring up competitive programming benchmarks and IMO silver medal progress. Scott shares insider context about legendary competitive programmer Gennady Korotkevich working at Cognition before concluding with Devin's product launch details.0:00–4:44 · Guest teaching 1/10 Devin's Agent-Native IDE Launch and Features The host demonstrates strong domain knowledge by quoting Sam Altman's 10-year-old 'Bubble Talk' blog post in precise detail, including specific unicorn names and valuation targets. Scott Wu readily agrees with the host's historical premise and expands on why AI is actually undervalued rather than in a bubble.4:44–7:25 · Guest teaching 2/10 Vibe Coding and the Mass Adoption of AI Programming The host references Karpathy's 'vibe coding' concept, Levels.io, and viral generative AI moments to ask about mass adoption. Scott provides an educational breakdown separating single-use hobbyist creation from deep enterprise software architecture.7:25–11:08 · Guest teaching 3/10 Engineering Advice and the Evolution of Coding Interfaces The co-host poses broad questions using analogies about engineering talent and adoption lag. Scott grounds the discussion by explaining the transition of AI interfaces from text completion to agentic systems and the real-world friction of legacy codebases like COBOL.11:08–13:30 · Guest teaching 3/10 Multimodality and Iterative Debugging in AI Systems The host articulates an insightful hypothesis regarding how cross-modal and multilingual data ingestion boosts model capability. Scott explains how multimodal capabilities mirror practical human developer workflows like visual iteration and log inspection.13:31–15:31 · Guest teaching 1/10 AI Benchmarks: International Mathematical Olympiad and Competitive Programming The hosts bring up competitive programming benchmarks and IMO silver medal progress. Scott shares insider context about legendary competitive programmer Gennady Korotkevich working at Cognition before concluding with Devin's product launch details.0:00–4:44 · Guest disagreement 1/10 Devin's Agent-Native IDE Launch and Features The host demonstrates strong domain knowledge by quoting Sam Altman's 10-year-old 'Bubble Talk' blog post in precise detail, including specific unicorn names and valuation targets. Scott Wu readily agrees with the host's historical premise and expands on why AI is actually undervalued rather than in a bubble.4:44–7:25 · Guest disagreement 1/10 Vibe Coding and the Mass Adoption of AI Programming The host references Karpathy's 'vibe coding' concept, Levels.io, and viral generative AI moments to ask about mass adoption. Scott provides an educational breakdown separating single-use hobbyist creation from deep enterprise software architecture.7:25–11:08 · Guest disagreement 1/10 Engineering Advice and the Evolution of Coding Interfaces The co-host poses broad questions using analogies about engineering talent and adoption lag. Scott grounds the discussion by explaining the transition of AI interfaces from text completion to agentic systems and the real-world friction of legacy codebases like COBOL.11:08–13:30 · Guest disagreement 0/10 Multimodality and Iterative Debugging in AI Systems The host articulates an insightful hypothesis regarding how cross-modal and multilingual data ingestion boosts model capability. Scott explains how multimodal capabilities mirror practical human developer workflows like visual iteration and log inspection.13:31–15:31 · Guest disagreement 0/10 AI Benchmarks: International Mathematical Olympiad and Competitive Programming The hosts bring up competitive programming benchmarks and IMO silver medal progress. Scott shares insider context about legendary competitive programmer Gennady Korotkevich working at Cognition before concluding with Devin's product launch details.0:00–4:44 · The hosts pushing back 2/10 Devin's Agent-Native IDE Launch and Features The host demonstrates strong domain knowledge by quoting Sam Altman's 10-year-old 'Bubble Talk' blog post in precise detail, including specific unicorn names and valuation targets. Scott Wu readily agrees with the host's historical premise and expands on why AI is actually undervalued rather than in a bubble.4:44–7:25 · The hosts pushing back 1/10 Vibe Coding and the Mass Adoption of AI Programming The host references Karpathy's 'vibe coding' concept, Levels.io, and viral generative AI moments to ask about mass adoption. Scott provides an educational breakdown separating single-use hobbyist creation from deep enterprise software architecture.7:25–11:08 · The hosts pushing back 0/10 Engineering Advice and the Evolution of Coding Interfaces The co-host poses broad questions using analogies about engineering talent and adoption lag. Scott grounds the discussion by explaining the transition of AI interfaces from text completion to agentic systems and the real-world friction of legacy codebases like COBOL.11:08–13:30 · The hosts pushing back 0/10 Multimodality and Iterative Debugging in AI Systems The host articulates an insightful hypothesis regarding how cross-modal and multilingual data ingestion boosts model capability. Scott explains how multimodal capabilities mirror practical human developer workflows like visual iteration and log inspection.13:31–15:31 · The hosts pushing back 0/10 AI Benchmarks: International Mathematical Olympiad and Competitive Programming The hosts bring up competitive programming benchmarks and IMO silver medal progress. Scott shares insider context about legendary competitive programmer Gennady Korotkevich working at Cognition before concluding with Devin's product launch details.

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

0:00 · the hosts 0% · guest 100%0:00 · the hosts 0% · guest 100%3:00 · the hosts 0% · guest 100%3:00 · the hosts 0% · guest 100%6:00 · the hosts 0% · guest 100%6:00 · the hosts 0% · guest 100%9:00 · the hosts 0% · guest 100%9:00 · the hosts 0% · guest 100%12:00 · the hosts 0% · guest 100%12:00 · the hosts 0% · guest 100%15:00 · the hosts 0% · guest 100%15:00 · the hosts 0% · guest 100%
Sharpest disagreement ▶ 2:42 Scott challenges the popular AI bubble narrative

Scott explicitly rejects mainstream bubble talk, arguing that the transformative value being delivered by AI is actually underrepresented in market perception.

 Hardest push from the hosts ▶ 1:16 Host presses on whether AI is overhyped

The host uses a detailed historical parallel from Sam Altman to push Scott to defend whether current AI valuations are overblown or justified.

 Biggest teaching moment ▶ 11:47 Scott educates on how developers and AI actually solve complex code problems

Scott explains that dumping massive text contexts into an LLM fails because real software engineering relies on iterative tool use, visual debugging, and stepping stones.

 The host holds their own ▶ 1:20 Host recites valuation data from Sam Altman's 2015 blog post

The host demonstrates deep tech history knowledge by quoting specific valuations and company names from Sam Altman's 10-year-old essay from memory.
the scores for every segment, with the reasoning behind each
ChapterTopicThe hosts as informed peerGuest teachingGuest disagreementThe hosts pushing backWhy
Devin's Agent-Native IDE Launch and Features 7112 The host demonstrates strong domain knowledge by quoting Sam Altman's 10-year-old 'Bubble Talk' blog post in precise detail, including specific unicorn names and valuation targets. Scott Wu readily agrees with the host's historical premise and expands on why AI is actually undervalued rather than in a bubble.
Vibe Coding and the Mass Adoption of AI Programming 6211 The host references Karpathy's 'vibe coding' concept, Levels.io, and viral generative AI moments to ask about mass adoption. Scott provides an educational breakdown separating single-use hobbyist creation from deep enterprise software architecture.
Engineering Advice and the Evolution of Coding Interfaces 4310 The co-host poses broad questions using analogies about engineering talent and adoption lag. Scott grounds the discussion by explaining the transition of AI interfaces from text completion to agentic systems and the real-world friction of legacy codebases like COBOL.
Multimodality and Iterative Debugging in AI Systems 6300 The host articulates an insightful hypothesis regarding how cross-modal and multilingual data ingestion boosts model capability. Scott explains how multimodal capabilities mirror practical human developer workflows like visual iteration and log inspection.
AI Benchmarks: International Mathematical Olympiad and Competitive Programming 5100 The hosts bring up competitive programming benchmarks and IMO silver medal progress. Scott shares insider context about legendary competitive programmer Gennady Korotkevich working at Cognition before concluding with Devin's product launch details.

Statements from this episode (9)

Disclosure
Scott Wu: Cognition launched an agent-native IDE for Devin
“What we rolled out today is essentially an agent native IDE. And what that means is, you know, I think there are a lot of great products out there that are about really accelerating you in a synchronous workflow. What we've always built Devon for is to be trul…”
Scott Wu Apr 8, 2025 ▶ 0:26
Opinion
Scott Wu: AI's proven value is underrepresented, not a bubble
“I think people talk about bubbles so much in AI and I, it's almost, I feel almost the exact opposite where the value that we are suddenly proving out in AI is almost underrepresented in, in what we're seeing and what is going on?”
Scott Wu Apr 8, 2025 ▶ 2:55
Insight
Scott Wu: Clean feedback loops make coding the first AI-automated domain
“Code is one of the first ones because it's so nicely automatable. There's a clean feedback loop and all of these things that you can really train models to get really good at code.”
Scott Wu Apr 8, 2025 ▶ 3:15
Opinion
Scott Wu: Enterprise engineering still requires humans for architecture reasoning
“And I think on that side, you know, I think for the time being, that's going to be primarily focused on software engineers. Just because, you know, it's while we're still working with these layers of abstraction, you still really need to have all of these Kind…”
Scott Wu Apr 8, 2025 ▶ 6:45
Insight
Scott Wu: AI coding must solve legacy constraints, not just capabilities
“And so I think one of the big things, honestly, that, that that we're solving as a space is, is actually not just the capability itself, but just solving for a lot of these practical constraints and getting the technology to the point where it works for everyo…”
Scott Wu Apr 8, 2025 ▶ 10:49
Insight
Scott Wu: Multimodal input enables AI agents to debug and iterate
“Similarly, I think, you know, being able to use image or being able to interact with these other data sources is, is what really gives AI is that same capability.”
Scott Wu Apr 8, 2025 ▶ 13:24
Assertion Supported
Scott Wu confirms competitive programmer Gennady Korotkevich works at Cognition
“His name is Gennady. He's from Belarus and he works at Cognition actually.”
Scott Wu Apr 8, 2025 ▶ 13:55
Prediction Held up
Scott Wu predicts AI will win IMO gold in 2025 or 2026
“I think it's probably gonna happen this year. And if not this year, then next year.”
Scott Wu Apr 8, 2025 ▶ 14:26
Disclosure
Scott Wu announces Devin general availability starting at a $20 minimum
“We rolled out a whole new experience, and then on top of that, we, you know, we would really love to hear what folks think, and so we're making it a lot easier to try out and so we have a new plan that starts at a 20 dollar minimum and it's all available at ap…”
Scott Wu Apr 8, 2025 ▶ 14:56
Made with StarZero

Turn any episode into a week of clips.

This entire site, over 500 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