Nov 7, 2025 · 1h 2m · a16z
Amjad Masad & Adam D’Angelo: How Far Are We From AGI?
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 the a16z Podcast, host Erik Torenberg speaks with Quora CEO Adam D'Angelo and Replit CEO Amjad Masad about realistic AGI timelines, the limits of current LLMs, and technological paradigm shifts. They analyze how AI automation is reshaping labor markets and software development while empowering a new wave of solo entrepreneurs.
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 host as informed peer 4.4 Guest teaching 3.6 Guest disagreement 1.9 The host pushing back 2.6
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 4:36 Amjad rejects AGI timeline papers as unscientific vibe hype
Amjad forcefully dismisses popular AI safety and AGI timeline papers as unscientific vibes rather than rigorous research, arguing that hype damages public policy.
Hardest push from the host ▶ 27:40 Host challenges Peter Thiel's AI centralization thesisThe host explicitly rejects Peter Thiel's decade-old claim that AI is centralizing or communist, offering a counter-argument on how AI empowers solo entrepreneurs at the edges.
Biggest teaching moment ▶ 21:52 Adam refutes human-experience necessity using recommender systemsAdam reframes Amjad's assertion that human experience is necessary to understand human desires by demonstrating how existing recommender algorithms are already superhuman at predicting taste.
The host holds their own ▶ 27:40 Host articulates nuanced tech decentralization dynamicsThe host demonstrates strong technology ecosystem mastery by challenging established venture tropes and contrasting crypto, Web 2, and AI market dynamics.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The host as informed peer | Guest teaching | Guest disagreement | The host pushing back | Why |
|---|---|---|---|---|---|---|
| Host Welcome and Guest Introductions | 3 | 2 | 3 | 4 | The host opens by raising prevailing market skepticism around LLM timelines and labor automation. Adam dismisses the bearishness, prompting the host to push back by re-stating concerns regarding end-to-end task automation limits. | |
| Defining AGI: The Remote Worker Benchmark | 3 | 5 | 4 | 2 | Adam defines AGI around the remote worker benchmark, while Amjad counters by labeling hype papers on 2027 AGI as unscientific vibe pieces that mislead policymakers. Amjad breaks down functional AGI and the limitations of present LLM architectures. | |
| Brute-Force Compute vs. Evolutionary Human Intelligence | 4 | 3 | 3 | 5 | When Adam and Amjad appear to converge on brute-force compute solutions, the host intervenes directly to press them on where their actual disagreement lies. This forces the guests to distinguish between engineering brute force and basic research into human-like learning. | |
| Thomas Kuhn's Paradigm Shifts and Commercial AI Research | 4 | 4 | 3 | 3 | Amjad invokes philosopher Thomas Kuhn to explain how commercial research creates black-hole bubbles that distract from fundamental science. The host grounds the debate by steering both guests toward concrete GDP growth projections. | |
| The Expert Data Dilemma and Reinforcement Learning | 3 | 3 | 1 | 2 | Amjad points out the paradox of using expert data to replace the experts needed for training, while Adam references AlphaGo's synthetic RL environment. The host guides the scenario by outlining specific future job categories. | |
| Human Experience in the Economy and 'The Sovereign Individual' | 5 | 5 | 3 | 3 | The host draws on Adam's experience building Quora to ask about tacit knowledge and AI caretakers. Adam refutes Amjad's assertion that human experience is necessary for understanding human desires by highlighting superhuman social media recommendation algorithms. | |
| Centralization vs. Decentralization: Power at the Edges | 7 | 2 | 1 | 6 | The host challenges Peter Thiel's thesis that AI is inherently centralizing or communist, offering a counter-argument on how AI empowers solo entrepreneurs at the edges. Both guests strongly agree with the host's reframing. | |
| Market Dynamics: Incumbent Advantage vs. Startup Disruption | 6 | 4 | 1 | 2 | The host frames the market analysis using Clayton Christensen's Innovator's Dilemma to ask whether AI value capture will favor incumbents or new startups. Adam and Amjad detail how widespread familiarity with Christensen's theories altered incumbent behavior. | |
| Monetization Models and Reduced Network Effects in Web 3.0 | 6 | 3 | 1 | 3 | The host reflects on past venture capital missteps from the Web 2 era regarding winner-take-all consolidation, contrasting it with current multi-model application markets. Adam explains why reduced network effects allow for multiple category winners. | |
| Geopolitics, Model Diversity, and Consumer Sophistication | 4 | 3 | 1 | 2 | Amjad notes how geopolitical fragmentation encourages regional foundation models. The host ties this to consumer behavior, while Amjad notes how even non-technical users now routinely juggle multiple AI models for distinct tasks. | |
| Unlocking Tacit Human Knowledge and Quora's AI Integration | 5 | 4 | 1 | 2 | The host probes the internet training data bottleneck, asking how much uncaptured human knowledge remains. Adam outlines the emerging data-labeling supply chain and explains how Quora monetizes human insights. | |
| The Evolution of Replit: From Autocomplete to Autonomous AI Agents | 5 | 6 | 1 | 1 | The host grounds the discussion in Replit's dramatic revenue shift from non-profit ed-tech to high-growth AI company. Amjad gives an in-depth operational breakdown of Replit's evolution from simple code completion to multi-hour autonomous agents using custom verifier loops. | |
| Workplace Cultural Shifts and the Promise of 'Vibe Coding' | 3 | 3 | 1 | 1 | Amjad reflects on workplace culture shifts where engineers interact more with AI agents than colleagues. The host validates these social observations and steers the conversation toward investment opportunities like 'vibe coding'. | |
| Educational Advice: Should Students Still Major in Computer Science? | 3 | 4 | 2 | 1 | The host asks for educational advice for incoming university students. Adam defends majoring in computer science, while Amjad laments San Francisco's modern tech culture as overly get-rich-quick focused compared to historical open-source tinkering. |
Statements from this episode (37)
Adam D'Angelo: Context delivery, not raw intelligence, bottlenecks current AI models
“I think a lot of what's holding back the models these days is not Actually, intelligence. It's getting the right context into the model so that it can Be able to use its intelligence.”
Adam D'Angelo: AI computer use will automate major work within two years
“And then there's some things like computer use that are still not quite there, but I think we'll almost definitely get there in the next year or two. And when you have that, I think we're gonna be able to automate a large portion of what people do.”
Adam D'Angelo: AGI is achieved when AI can do any remote job
“One definition I kind of like is if you say that you have a remote worker, a human, any job that can be done by someone whose job can be done remotely that, that's AGI.”
Adam D'Angelo: Current LLM architectures are not hitting performance limits
“I don't think so. I mean, I think there are certain things like memory and learning, like continuous learning that are not very easy with the current architectures. I think even those you can sort of fake and maybe are, we're going to be able to get them to wo…”
Amjad Masad: Papers predicting AGI by 2027 are vibe-based hype, not science
“So my criticism of the idea of like AGI, 20, 27, you know, that paper that I think it's called Alexander or someone else wrote and then and the situational awareness and all this hype papers that are not really science. They're just vibe.”
Amjad Masad: AI progress now relies on manual human labeling over scaling
“In the true pre-training scaling era, you know, GPT-II, three, 3.5, maybe up to four it felt like you can just put more internet data in there and just, it just got better. Whereas now it feels like there's a lot of labeling work happening, there's a lot of co…”
Amjad Masad: Claude 4.5 was an underappreciated leap over Claude 4
“Cloud 4.5 was a huge jump. I don't think it's appreciated how much of a jump it was over, over four. There's really, really amazing things about cloud 4.5.”
Amjad Masad: Current large language models are not on the path to AGI
“I don't think LLMs as they can understand are on, on the way to AGI and my definition for AGI is I think the old school RL definition, which is a machine that can go into any environment and learn efficiently in the same way that a human could go into you can …”
Adam D'Angelo: Brute-force scaling will yield AI capable of average human jobs
“I think that's going to be more a function of when we can produce something that is as good as human intelligence, even if it takes a lot more compute, a lot more energy, a lot more training data. We could just put in all that energy and still get to software …”
Amjad Masad: Singularity requires discovering true, non-brute-force intelligence algorithms
“I don't think that we'll get to the singularity or I don't think that, I don't think we're gonna get to the next level of human civilization until we crack the true nature of intelligence. Like until we understand and have algorithms that are actually not brut…”
Amjad Masad: LLMs are distracting top researchers from fundamental AI science
“It just does, it does feel like the LLMs in a way are distracting from that because all the talent is going there. And therefore there's less talent that are trying to do basic research on, on intelligence.”
Adam D'Angelo: Top experts can solve fundamental AI challenges within five years
“Nothing seems fundamentally so hard that it couldn't be solved by the smartest people in the world working incredibly hard for the next five years.”
Adam D'Angelo: Current AI paradigm is far from diminishing returns
“I think the current paradigm is pretty good. And I think we're nowhere near the sort of like diminishing returns of continuing to push on it. And I bet. Yeah, I guess I would just bet that you can keep doing different innovations within the paradigm to get the…”
Adam D'Angelo: GDP growth will exceed 5% if AI automates work cheaply
“I think you're gonna get to much more than four to five percent GDP growth in that world.”
Adam D'Angelo: LLMs will perform all human tasks cheaper within 5-15 years
“I do think at some point you get to LMs can, they can do Everything, every single thing a human can do for cheaper. Like, I don't see a reason why we don't eventually get there. That may take five, 1015 years.”
Adam D'Angelo: LLMs are substituting for entry-level computer science jobs
“CS majors graduating from college, there's just not as many jobs as there used to be, and LLMs are a little more substitutable for what they previously would have done, and I'm sure that's contributing to it.”
Amjad Masad: Automating expert labor will deplete human expert AI training data
“Another related problem is that since we're Dependent on, ah, expert data in order to train the alums and the alums start to substitute those workers. But, you know, at some point there's no more experts because they're all out of jobs and they're equivalent t…”
Adam D'Angelo: Human chess participation increased after AI surpassed human players
“There's a data point that the people playing chess is up since computers got better at human than humans at chess.”
Amjad Masad: AI will not automate all jobs without human embodiment
“I don't think we're gonna get to a point where you automate every, every job. Definitely not in the current paradigm. I would I would doubt it happening. I, I'm not certain it would ever happen, but definitely not in the current paradigm. Now here's what I thi…”
Adam D'Angelo: Feed recommendation algorithms are already superhuman at predicting interest
“Recommender systems, the system that ranks your Facebook or Instagram or Quora feed, those recommender systems are already superhuman at predicting what you're gonna be interested in, in reading.”
Amjad Masad: Developers are quitting jobs after making money on Replit
“Yeah, I get this, Tweets all the time about people who like quit their jobs because they started making so much money. You're using other tools like Rap Lead and it's really exciting.”
Adam D'Angelo: Hyperscaler market balances fast-falling prices with high R&D investment
“There's enough competition among the hyperscalers that the... As an application level company, you have choice and you have alternatives and the prices are coming down incredibly quickly. But there's also not so much competition that the hyperscalers and the, …”
Amjad Masad: AI supercharges incumbents while enabling counter-positioned startups
“This time it feels like it is an obvious supercharge for the incumbents, for the hyperscalers, for the large internet companies, but it also enables new business models that that is perhaps counter position against the existing, existing ones.”
Adam D'Angelo: AI would be far more disruptive in a 1990s corporate landscape
“I think if you had an environment more like we had in, say, like the nineties, I think this would actually be more disruptive than the current hyper, hyper competitive world that we're in now.”
Adam D'Angelo: Network effects play a smaller role in AI than Web 2.0
“I think network effects are playing much less of a role now than they did in the web two era also, and that, that makes it easier for competitors to get started. There's still a scale advantage because, you know, if you have more users, you can get more data. …”
Adam D'Angelo: Subscriptions allow AI startups to monetize immediately without scale
“Like you couldn't build a good ad business until you got to 1,000,010 of millions of users. And now with subscriptions, you can just charge right away. I think especially thanks to things like Stripe that are making it easier. And so that, that, that's also ma…”
Amjad Masad: Geopolitical fragmentation creates opportunities for regional AI foundation models
“It seems clear that we're not in this globalized era, and perhaps it's gonna get much worse, and so investing in the foundation, in the open AI of Europe might be a good idea, and like, similarly, China being an entirely different, different world, and so ther…”
Adam D'Angelo: Quora built Poe because GPT-3 answers failed to match humans
“The way we got to it was we, in early, we started experimenting with using GPT-III to generate answers. For Quora. And we compared them to the human answers and sort of realized that they weren't as good, but what was really unique was that you could instantly…”
Amjad Masad: Non-technical consumers routinely switch between multiple competing AI models
“Not particularly technical consumers actually do use multiple AIs. Like I didn't expect that, like, you know, people only use Google. They never like looked at Google and then Yahoo or like very few people do it. But now you talk to just average people and the…”
Adam D'Angelo: Data will become the primary bottleneck for AI development
“As you have, you know, as intelligence gets cheaper and cheaper and more and more powerful, the bottleneck, I think, is increasingly going to be on the data and what do you need to create that intelligence? And so that's going to cause this, that's going to ca…”
Adam D'Angelo: Quora has partnerships with major AI labs for data
“We have relationships with some of the AI labs and we're gonna sort of play the role, Quora will play the role that it is meant to play in this ecosystem, which is a, as a source of human knowledge.”
Amjad Masad: Future developer productivity will depend on managing multiple parallel AI agents
“I think the next, ah, boost in productivity is gonna come from sitting in front of a programming environment like Replit and being able to manage, ah, tens of agents, maybe at some .100, but, you know, at least, you know, five, six, seven, eight, nine, 10 agen…”
Adam D'Angelo: AI tools will replace 100-person software engineering teams within years
“And if you imagine that they're going to get there, and I think there's no reason why they wouldn't, it'll take a few years, but then it's like everyone in the world is going to be able to create any things that would have taken a team of a hundred professiona…”
Adam D'Angelo: Computer science fundamentals will remain essential for managing AI agents
“Having these skills to understand the sort of fundamentals of what's possible with algorithms and data structures, I think that actually really helps you in, in managing agents when you're using them. And I, I'm guessing that it will continue to be a valuable …”
Amjad Masad: Replit left San Francisco over its get-rich-quick culture
“I think we're in an era of Silicon Valley where it's like very very get rich driven. And that makes me a little sad. And that's partly why I moved the company out of SF.”
Amjad Masad: Claude 4.5 shows awareness of context limits and red-team testing
“Quad 4.5 seemed to have to become more aware of its context length. So as it gets closer to the end of the context, it starts becoming more economical with tokens. It also, it looks like its awareness when it's being red teamed or in test environment, like jum…”
Amjad Masad: College freshmen should study philosophy of mind and neuroscience
“I would definitely study philosophy of mind. I would probably go into neuroscience. Cause I think those are the core questions that are kind of become very, very important as AI kind of continuously more of jobs and economy and things like that.”