Aug 21, 2025 · 42m · no-priors
No Priors Ep. 128 | With Andrew Ng, Managing General Partner at AI Fund
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
AI pioneer Andrew Ng joins hosts Sarah Guo and Elad Gil on No Priors to discuss the mechanics of agentic workflows, the shift in startup bottlenecks from engineering execution to product management, and how hyper-compact, AI-enabled teams are reshaping modern enterprise.
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 39.6% of the talking time here. How this is scored →
The hosts as informed peer 5.3 Guest teaching 2.5 Guest disagreement 1.0 The hosts pushing back 1.6
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 9:11 Dismissing vibe coding as trivializing
Andrew firmly pushes back against the popular industry term 'vibe coding', calling AI-assisted software creation an exhausting and disciplined engineering feat rather than casual vibe curation.
Hardest push from the hosts ▶ 28:58 Elad challenging hyper-lean team orthodoxyElad directly challenges the idea that minimal team size is universally virtuous, calling excessive pride in lean headcount a trap that cost companies like Sketch their market lead.
Biggest teaching moment ▶ 3:07 Reframing agent bottlenecks to evals and tacit knowledgeAndrew reframes the hosts' technical questions by demonstrating that systematic error analysis and tacit organizational knowledge are far bigger real-world bottlenecks than model features.
The host holds their own ▶ 13:51 Elad citing early tech and mobile founder historyElad reinforces and expands on technical founder advantages by citing historical parallels from semiconductor pioneers through mobile leaders like WhatsApp and Instacart.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Exploring Multiple Vectors of AI Capability Growth | 4 | 3 | 1 | 1 | Sarah opens with a foundational question about scaling limits versus data work, prompting Andrew to explain multiple vectors of AI progress. Andrew recounts why he originally coined the term 'agentic AI' and how PR/marketing machinery distorted its nuance. | |
| Overcoming Talent and Evaluation Bottlenecks in Agents | 5 | 5 | 1 | 1 | Elad asks about missing technical components for true agents, but Andrew reframes the core bottleneck around human talent, disciplined error analysis, and evals. Andrew explains that automating agentic workflows requires ingesting tacit, proprietary human knowledge. | |
| Autonomous Coding Agents Outperform Computer Use | 5 | 3 | 2 | 2 | Andrew highlights coding agents as the highest-functioning agentic systems today and rejects the casual term 'vibe coding' in favor of rigorous rapid engineering. Sarah probes whether coding success is fundamental research or market incentives, which Andrew connects to economic forces. | |
| Product Management as the Primary Startup Bottleneck | 4 | 4 | 1 | 1 | Andrew details how accelerated coding speeds shift the primary startup bottleneck to product management decisions. Elad asks whether automated market research agents can bridge that gap, with Andrew remaining cautious about current simulation tools. | |
| The Resurgence of the Technical Startup Founder | 6 | 2 | 1 | 2 | Andrew and the hosts align on the necessity of technical founders during platform shifts. Elad brings in historical examples from Silicon Valley history and mobile pioneers (WhatsApp, Instacart, Uber) to substantiate Andrew's thesis. | |
| Work Ethic, Drive, and Startup Conviction | 5 | 1 | 1 | 1 | The conversation explores founder intensity, hard work, and extreme competitiveness. Sarah and Elad discuss how startup conviction and drive are essential for unreasonable outcomes, while Andrew shares his personal experiences balancing intense focus with customer obsession. | |
| Rapid Decision-Making and the Value of User Empathy | 5 | 4 | 1 | 1 | Andrew illustrates startup decision-making as rapid tennis volleys governed by customer empathy rather than slow calculus. Elad asks about the scarcity of top-tier PM talent, prompting Andrew to reflect on his past mistake of trying to convert engineers into PMs. | |
| Building Products from Intuition and Single Persona Focus | 6 | 2 | 1 | 1 | Sarah and Andrew discuss how products like Cursor and Claude Code iterate from internal intuition rather than heavy user panels by targeting a sharp single persona. Andrew and Elad note that non-technical corporate functions are adopting coding to express precise operational intent. | |
| Vertical AI Disruption and Super-Compact Teams | 6 | 2 | 1 | 1 | Elad and Andrew discuss legal AI adoption using Harvey and Casetext/Catalyst as examples of high-value vertical disruption. Andrew and Sarah note how compact AI-enabled teams dramatically lower coordination overhead. | |
| Evaluating the Strategic Risks of Hyper-Lean Headcount | 8 | 1 | 1 | 6 | Elad offers substantial pushback against the dogma of keeping startup headcounts hyper-lean, arguing it can be an excuse for missing explosive market windows (citing Sketch vs Figma and Slack vs Teams). Andrew acknowledges the market dynamic distinction between winner-take-all spaces and niche efficiency. | |
| Screening for Concrete Startup Opportunities at AI Fund | 5 | 3 | 1 | 1 | Sarah asks about upcoming industry transformations and investment screening practices. Andrew explains AI Fund's mandate for concrete bottom-up use cases over broad top-down market studies and outlines the durable human information advantage in relationship-driven evaluations. | |
| Supporting and Scaling First-Time Technical Founders | 7 | 2 | 1 | 3 | Andrew turns the tables to ask the hosts how they support first-time technical founders. Elad explains his hands-off philosophy of surrounding founders with peer groups and complementary executive hires rather than micromanaging non-existential errors. | |
| The Five-Year Vision of Individual AI Empowerment | 3 | 1 | 0 | 0 | Sarah closes with a forward-looking question on the broader impact of AI. Andrew delivers an optimistic summary about massive individual productivity empowerment across diverse professional roles. |
Statements from this episode (31)
Andrew Ng: Gaining AI capability from pure scale is becoming extremely difficult
“So I think there is probably a little bit more juice out of the scalability element to this piece, so hopefully you'll consider making progress there, but it's getting really, really difficult.”
Andrew Ng: Public AI narrative is skewed by tech giants' PR machinery
“Society's perception of AI has been very skewed by the PR machinery of a handful of companies with amazing PR capabilities. And because that number of companies drove scales and narrative, people think of scale first of this vector progress.”
Ng: AI operates on a spectrum of agency rather than a binary
“And I felt there's a lot of good work, and there was a spectrum of degrees of agency. Where there are highly autonomous agents that could plan, take multiple sets of reasoning, do a lot of stuff by themselves, and then things that were lower degrees of agency,…”
Ng: Agentic AI business progress lags behind marketing hype
“I feel like the marketing hype has gone like that insanely fast, but the real business progress has also been, you know, rapidly growing, but maybe not as fast as the marketing hype.”
Ng: Talent is the biggest barrier to implementing agentic AI workflows
“But what I see is the single biggest Barrier to getting more agentic AI workflows implemented is, is actually talent.”
Ng: Systematic error analysis with evals separates top AI agent teams
“The single biggest differentiator that I see in the market is, does the team know how to drive a systematic error analysis process with evals? So you're building the agents by analyzing at any moment in time, what's working, what's not working, what do you imp…”
Ng: Human engineers needed for agentic workflows for 1-2 more years
“Until and unless we built, you know, AI avatars that can interview employees doing the work, and better visual AI that can look at the computer monitor, I think, maybe eventually, you know, but I think at least right now, for the next year or two, I think ther…”
Ng: Data needed for agentic workflows is proprietary, not on internet
“I feel like for a lot of work to be done building agentic workflows, that data set is proprietary. It's just not, it's not general knowledge on the internet.”
Ng: Coding Agents Are Among the Only Autonomous AI Tools That Actually Work
“So this ability to plan a multi-step thing, execute the multiple steps of a plan is one of the most highly autonomous agents out there being used that, that actually works.”
Ng: AI Computer Use Agents Are Nice Demos, Not Yet Production-Ready
“Some of the computer use stuff, like, you know, go shop for something for me and browse online. Some of those things are really nice demos, but not yet production.”
Ng: AI-assisted development is rapid engineering, not vibe coding
“It's like a deeply intellectual exercise, and I think the term vibe coding makes people think it's easier than it is. So frankly, after a day of using AI-assisted coding, like, I'm exhausted mentally, right? So I think of it as rapid engineering, where AI is l…”
Ng: Solo developers now build in a weekend what took months
“So, there's so many things that, you know, would have taken a team of six engineers, like, three months to build, that now, today, one of my friends or I was just building on a weekend.”
Ng: Startup bottleneck has shifted from coding speed to product management
“So when we go look at this loop, the speed of coding is accelerating, the cost is falling, and so increasingly the bottleneck is actually product management. So, so the product management bottleneck is now we can build what do we want much faster while the bot…”
Ng: AI tools accelerate software engineers far more than product managers
“I don't think those tools are accelerating product managers nearly as much as coding tools are accelerating software engineers. So this does treat more of the bottleneck on the product management side.”
Andrew Ng: Technical product leaders are far more likely to succeed in AI
“Founders that are on top of Gen AI technology, thus, you know tech oriented product leaders, I think are much more likely to succeed than someone that maybe is more business oriented, more business savvy, but it's not, doesn't have a good feel for where AI is …”
Ng: Winning Founders Are Driven Either By Business Victory Or Customer Success
“I've really seen, I feel like I've seen two types. One is they really want their business to win. That's fine. Some do great. Some are, they really want their customers to win. And they're so obsessed with serving the customer that that works out.”
Ng: Early career mistake was attempting to train engineers into product managers
“One of my failures, one of the things I did not do well in the early phase of my career for some dumb reason, I tried to make a bunch of engineers product managers. I gave them product manager training, and I found that I just foolishly made a bunch of really …”
Ng: High human empathy is the primary correlate of great product instincts
“But I found that one correlate for whether someone, you know, would have good product instincts is that Very high human empathy where you can synthesize lots of signals to really put yourself in other person's shoes to then very rapidly make proper decisions o…”
Guo: Cursor team makes product decisions via instinct, not user interviews
“I think it is, like, reasonably well known that the cursor team, like, they make their decisions actually very instinctively versus spending a lot of time talking to users”
Ng: Startups gain an advantage over incumbents by building for a single persona
“So it turns out one advantage that startups have is while you are early, you can serve kind of one user profile. Today, you know, if you're, I don't know, like Google, right? Google serves such a diverse set of user personas, you really have to think about a l…”
Ng: Every AI Fund Employee Knows How to Code
“Everyone in my team at AI Fund knows how to code. Everyone is a good hub account, and I see for a lot of my team members, you know, when my, I don't know assistant general counsel, or my CFO, or my front desk operator, when they learn how to code, they're not …”
Ng: Top Engineers Pair 10+ Years Experience With AI Tools
“The best engineers I work with today are not fresh color strats. They're people with, you know, 10, 15 or more years of experience, but they're also really on top of AI tools, and that, those engineers are just completing a class of their own.”
Ng: AI Transformation in Engineering Previews Other Disciplines
“I think software engineering is a hard bringer of what will happen in other disciplines, because the tools are most advanced in software engineering.”
Guo: Open Evidence reaches 50% of US doctors while keeping headcount minimal
“I work with several teams now one of which is called Open Evidence and has, like, a pretty good penetration, like, 50% of doctors in the U.S. Now, where it's an explicit objective in the company to try to be as small as possible.”
Gil: Startups must scale fast before incumbents kill them with distribution
“Usually I think what happens is in the early stage of a startup life, you're competing with other startups. And if you're way ahead, it feels great, but eventually if they're incumbents in your market, they come in and the faster you capture the market and mov…”
Gil: Many big tech companies could shrink by 70% and improve
“If you look at the big tech companies, for example, right now, many, not all of them, but many of them could probably shrink by 70% and be more effective, right?”
Ng: AI Fund avoids broad top-down ideas, requiring concrete use cases
“And so one of the lessons I've learned is we really like concrete ideas. So someone says, I did a market analysis, AI will transform healthcare. That's true, but I don't know what to do with that. But if someone, a subject matter expert or an engineer comes an…”
Ng: Deep company and competitive research is ripe for automation
“I feel like doing
Deep research on individual companies and competitive research, that seems right for automation.”
Ng: Humans retain huge information advantage where AI lacks context plumbing
“So I find that there are a lot of these tasks where humans have a huge information advantage still, because they've not figured out the plumbing or whatever's needed to get information to the AI model.”
Gil: Late-career professionals hire for skill gaps rather than building them
“I think in general, one of my big learnings is I feel like early in careers, people try to complement or try to build out the skills that they don't have. In late in careers, they lean into what they're really good at and then they hire people to do the rest.”