Mar 27, 2026 · 53m · american-optimist
How AI Agents Are Creating 12X Productivity Gains · Joe Lonsdale
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
In this episode of American Optimist, host Joe Lonsdale speaks with Cognition co-founders Scott Wu and Russell Kaplan about Devin, the world's first autonomous AI software engineer. They discuss how AI agents are driving massive productivity gains, transforming software engineering workflows, modernizing enterprise and government IT, and reshaping the future of technology talent.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Joe holds 19.7% of the talking time here. How this is scored →
Joe as informed peer 4.5 Guest teaching 3.8 Guest disagreement 1.3 Joe pushing back 1.7
speaking balance: gold is Joe, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 51:48 Scott dismisses Citrini's AI economic doomerism
Scott directly criticizes a prominent economic analysis, labeling it ridiculous for confusing nominal and real deflationary dynamics.
Hardest push from Joe ▶ 5:59 Joe pushes on whether youth offers an inherent AI advantageJoe challenges the guests on whether teenagers learning AI natively have an insurmountable structural edge over veteran engineers.
Biggest teaching moment ▶ 45:30 Scott clarifies METR benchmark metrics versus real-time executionScott corrects Joe's confusion regarding how model autonomy horizons measure equivalent human work durations rather than literal model compute time.
Joe holds their own ▶ 39:03 Joe details Palantir's forward deployed engineer frameworkJoe draws deeply from his experience co-founding Palantir, breaking down the operational mechanics and feedback loops of forward deployed engineering.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Joe as informed peer | Guest teaching | Guest disagreement | Joe pushing back | Why |
|---|---|---|---|---|---|---|
| Guest Backgrounds and Cognition Team Demographics | 4 | 3 | 1 | 2 | Joe establishes his personal history working with Scott and highlights Scale AI's success, framing the technical caliber of competitive programmers. Scott explains how technical execution is the primary competitive moat in AI compared to traditional marketplace or logistics startups. | |
| The Chief Engineer Philosophy and End-to-End AI Systems | 5 | 4 | 2 | 3 | Joe presses whether teenagers possess a unique cognitive advantage growing up with AI tools. Russell explains the Autopilot 'chief engineer' philosophy and how breaking down modular abstraction boundaries requires end-to-end stack comprehension. | |
| Cognition's Exponential Growth and Legacy Codebase Refactoring | 4 | 5 | 1 | 1 | Scott and Russell break down Cognition's growth metrics and share their early product-market fit finding: massive enterprise refactoring projects where automated multi-file changes exceed regex capability but fall within agent scope. | |
| Defining Software Abundance and Quality Log Scales | 3 | 4 | 1 | 1 | Joe prompts Scott on the definition of software abundance. Scott conceptualizes software quality on a logarithmic scale of user reach, explaining how agentic generation democratizes Tier-1 software engineering quality down to long-tail applications. | |
| The AI Engineering Workflow: Prompting, Iteration, and Self-Service | 4 | 4 | 1 | 1 | Scott explains that internal Cognition engineers write natural language instead of syntax, while Russell highlights that product development cycles have collapsed, eliminating the need to gate engineering time behind lengthy design specs. | |
| High Agency, Abundance Mindsets, and Parallel Agent Fleets | 4 | 4 | 1 | 2 | Joe questions whether engineers still verify assembly or raw code under the hood. Russell and Scott describe managing parallel agent fleets like machine learning experiments, requiring comfort with non-determinism alongside end-to-end verification. | |
| Macroeconomic Impact: Software Deflation and Physical World Energy | 5 | 4 | 2 | 2 | Russell contrasts Baumol's cost disease sectors with hyper-deflationary software production, citing Jevons paradox. Russell and Scott hypothesize that manual coding may become a rare, artisanal luxury good in the future. | |
| Cognition for Government and Public Sector IT Modernization | 6 | 4 | 1 | 2 | Joe discusses government procurement inefficiencies and permitting delays. Russell points out historical precedents like the 1890 census Hollerith punch cards and explains how Devin sidesteps SaaS IP ownership restrictions by writing bespoke code directly for agencies. | |
| Competitive Strategy, Forward Deployed Engineers, and Devin Review | 6 | 3 | 1 | 2 | Joe brings up Palantir's invention of Forward Deployed Engineers and ontologies to probe Cognition's competitive playbook. Russell and Scott explain their deployment framework and the development of Devin Review to manage AI code review bottlenecks. | |
| Future Projections: Exponential Autonomy Gains and Small Business Growth | 5 | 4 | 2 | 2 | Scott references METR benchmarks measuring human-task horizon duration, clarifying Joe's confusion over autonomous execution time versus equivalent human-labor hours. Russell predicts an explosion in AI-enabled small businesses. | |
| Hiring Former Founders and Building High-Agency Teams | 5 | 2 | 1 | 1 | Joe recalls Scott's early career at Addepar and asks why Cognition specifically recruits former founders. Scott and Russell explain that high-agency founders thrive when solving massive, open-ended problem spaces without rigid specialization. | |
| Rethinking Technical Interviews in the Age of AI | 3 | 4 | 2 | 1 | Scott critiques legacy anti-AI technical interviews, arguing evaluations should allow full AI usage to test architectural judgment. He concludes by rejecting Citrini-style economic pessimism and likening future software building to Minecraft creative mode. |
Statements from this episode (28)
Wu: Cognition engineering team's average age is about 25
“So, so on engineering, it's about 25. And then obviously on go-to-market's a little bit older”
Kaplan: Cognition hires 17- and 18-year-olds for engineering
“The engineering team, we have seventeen-year-olds, we have eighteen-year-olds, we have really young folks, and then but we'll take anyone at any age as long as they're ready to grind and ready to have a big impact.”
Wu: One 20-person USACO camp produced Wang, Guo, Ziegler, and Ho
“In my year, actually there were a ton of others who all kind of went into AI. And so obviously Steven and Andrew, who started the company you know, who started Cognition with us but also a ton of others. And so Alexander Wang, who started Scale. Demi Guo, who …”
Wu: Technical execution matters more in AI than in past tech waves
“I think in AI, what you see is that really excelling on the technical aspect just matters much more, I think, in AI than some of these other fields, and I think there's been lots and lots of businesses in the past that have been very, I'll say, like very inten…”
Kaplan: Software engineers must reset their AI workflows every three months
“I mean, every three months you have to throw out your previous experience and build new experience because the tools get so much better.”
Wu: Cognition acquired Windsurf and shares engineers across both products
“We bought Windsurf, you know, seven, eight months ago at this point, but like, we don't have a distinction of like, oh, this person's an engineer working on Devon or this person's engineer working on Windsurf. Like a lot of the same people should be people who…”
Wu: Cognition's early 2026 Devin usage surpassed all of 2025
“Today's March ninth, and we actually, at this point, have already done more Devon sessions in our customers in 2026 than we had in 2025. And so basically over the last like two months and change, we've already done more Devon usage in total than we had in all …”
Wu: Cognition works with Citibank and Santander
“We've been working with a lot of the biggest companies in the world, you know, Citibank, Santander and so on.”
Kaplan: Devin became top contributor to own codebase by Summer 2024
“It took us until Summer of 2024 for Devon to become the number one contributor to its own code base, like which was like the first real milestone.”
Scott Wu: Cognition engineers no longer type code
“For us at Cognition, for example, our engineers don't type code anymore.”
Kaplan: AI prompt-to-code shifts software development from scarcity to abundance
“It used to be the case that you have this entire software development life cycle and every step of that cycle, Is oriented around not wasting the super precious time of engineers writing code. And now you have suddenly this overflowing abundance of the ability…”
Wu: AI agents enable software engineers to focus 10x more on architecture
“10% of that job is basically this really fun part of just pure problem solving, thinking about what you want to build, being creative, like understanding the different solutions, deciding, okay, what architecture makes sense here? How exactly am I going to, yo…”
Kaplan: Software engineers should never let AI agents idle overnight
“At Tesla, we had one mantra, which was you know, never go to sleep while the GPUs are idling. You know, if you let your cluster idle overnight, that's just a huge waste of resources. Just kick off some experiments, you know, before you go to bed so you can wak…”
Kaplan: Software will enter hyper-deflation and output will jump 1000x
“We're going to enter this like hyper deflationary cycle of software where it's so easy to build. It's so abundant. That doesn't mean that people are going to, you know, stop producing software. They're actually going to produce thousands of times more software…”
Kaplan: Cognition measured 6x to 12x productivity gains with Devin in 2024
“And, you know, relatively quickly by like late 20, 24, we were measuring You know, somewhere between a six to 12 x productivity gain for those types of projects. Meaning that, you know, one hour of human time spent managing Devin was worth like six to 12 hours…”
Kaplan: Major regulated firm automatically resolves 70% of security alerts with Devin
“One of the largest regulated firms in the world, they do very thorough sort of security vulnerability scanning on their code... And they're remediating 70% of these automatically now in production with Devon.”
Lonsdale: Kushner and Gil automated Qatar building permits to just two hours
“My friend, Jared Kushner, who's obviously, you know, been involved in these administrations, he worked with Qatar, with Elad, and they built something where the permits only take, ah, 120 minutes there. So if you want a permit to build something, it'll get bac…”
Kaplan: US government spends $100B annually on IT modernization
“The government spends a hundred billion dollars a year on just IT modernization.”
Kaplan: Goldman Sachs tuned Devin for proprietary internal programming languages
“Goldman Sachs has invented some of their own programming languages. People don't know this, but Goldman has like a pretty insane internal engineering team and they've literally written their own programming languages and they've been able to sort of customize …”
Kaplan: Cognition has dozens of FedRAMP deployments across military and primes
“Right now we have dozens of kind of fed ramped deployments of cognition both with agencies and with primes. You know, we work with US Army, US Navy, the Treasury we work with folks like Palantir, with Andrel and other primes”
Kaplan: Frontier engineering problems are easier to solve than institutional inertia
“A lot of times it's actually easier to solve like a frontier science or engineering problem than it is to sort of change the molasses of the existing world”
Kaplan: English specifications will eventually replace code as the source of truth
“And eventually, you know, are we going to like have English as a source of truth and people are just going to be collaborating on specs? Yes.”
Wu: AI autonomous task duration grew from 10 seconds to 18 hours
“One of the stats that people talk about a lot is this METR report, which basically says for each different model that comes out roughly how much human work can it do in an automated fashion before you have to go interrupt it and say, oh, that was wrong. Let's …”
Wu: AI autonomous work capacity doubles 4-5 times per year
“And I think for the last few years, it's doubled about four or five times every year, you know, which is insane, which means, you know, you wait two or three months and it's already doing twice as much work.”
Kaplan: AI will cause an explosion in small businesses
“One thing I'll go on the record on is I think there's going to be an explosion in small businesses. I think AI is actually like an extremely small business enabling technology in particular”
Kaplan: Every engineer on Cognition's special projects team is a former founder
“We have one team at the company they're called special projects engineers. And basically every person on that team is a former founder, like every single one.”
Wu: Cognition lets engineering candidates use unlimited AI during interviews
“For us, and this has always been the case for us, you know, our interview process has always been, you can use as much AI as you want to use. You know, it's just like, we're going to give you a few hours, just build, build your whole own product surface, right…”