Mar 20, 2024 · 29m · a16z
Bringing AI to the Masses with Adam D'Angelo, CEO of Quora
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 AI Revolution series, David George interviews Adam D'Angelo, founder of Quora and Poe, on the transition from human knowledge networks to multi-model AI platforms. D'Angelo shares insights on creator monetization, organizational pivots, model scaling, and how startups can capture value against tech incumbents.
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 3.5 Guest teaching 3.1 Guest disagreement 0.8 The host pushing back 1.1
speaking balance: gold is the host, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 6:23 Rejecting fixed predictions on AI future trajectory
Adam gently resists the host's structured binary framework by noting that nobody knows how the future will unfold, reframing the bet on Poe as a hedge on overall ecosystem diversity.
Hardest push from the host ▶ 25:01 Host presenting the AI incumbent-advantage counterargumentDavid directly challenges the standard VC assumption that startups win by highlighting that incumbents hold API access, superior distribution, and billions in projected application revenue.
Biggest teaching moment ▶ 22:00 Adam mapping out the bifurcation of AI market structureAdam educates the host on the stark economic divide in AI, explaining why being six to twelve months off the frontier forces companies into intense commodity competition unless they build unique application wrapper features.
The host holds their own ▶ 18:29 David bringing Andrej Karpathy's LLM compression insight to the discussionDavid demonstrates deep technical fluency by invoking Andrej Karpathy's framing of LLMs as lossy compression algorithms, successfully extending Adam's argument regarding uncaptured human knowledge.
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 |
|---|---|---|---|---|---|---|
| Adam D'Angelo's Journey from Early AI to Founding Poe | 1 | 3 | 0 | 0 | David George yields almost the entire segment to Adam D'Angelo's extended monologue detailing his early AI experiments in college and the pivot from human-curated Quora Q&A to LLM chat interfaces. D'Angelo educates the listener on how social networks initially served as a human-driven substitute for AI automation before LLMs made instant answer generation feasible. | |
| Understanding Poe as a Multi-Model AI Platform | 3 | 4 | 1 | 1 | David George articulates two competing industry theories regarding single-model dominance versus multi-model diversity to prompt D'Angelo. D'Angelo explains Poe's vision as a multi-model aggregator, drawing a historical parallel to how early web browsers standardized access across diverse websites. | |
| Consumer Product Distribution versus Raw Model Provision | 3 | 4 | 1 | 1 | David George asks whether model providers will build consumer products directly or rely on distribution aggregators. D'Angelo walks through the heavy operational requirements of building cross-platform consumer apps, cross-border tax compliance, and billing, showing why researchers choose API integration over direct-to-consumer overhead. | |
| Creator Monetization, Image Generation, and the Long-Tail Ecosystem | 4 | 2 | 0 | 0 | David George actively engages by offering Roblox as an industry analogy for long-tail creator monetization and skill progression. D'Angelo fully agrees with the host's framing, elaborating on Poe's revenue-sharing mechanisms designed to offset GPU inference costs for independent bot creators. | |
| Lessons from Mobile Shift and Organizational Pivot to Poe | 3 | 2 | 0 | 1 | David George prompts D'Angelo on organizational lessons learned from leading Facebook during the mobile shift and how those informed Poe's creation. D'Angelo reflects candidly on Quora's delay in adapting to mobile due to a lack of decisive top-down prioritization, explaining why Poe was established as an independent project. | |
| Synergies Between Quora and Poe in Knowledge Sharing | 5 | 3 | 1 | 2 | David George demonstrates strong domain knowledge by quoting Andrej Karpathy's description of LLMs as lossy compression algorithms to discuss the limits of AI knowledge versus human expertise. D'Angelo builds on this point, explaining that proprietary tacit human knowledge remains uncaptured by models, necessitating hybrid human-AI knowledge networks. | |
| Model Scaling Laws and AI Industry Progress | 4 | 4 | 1 | 1 | David George introduces questions around scaling laws and eventual market structure across frontier labs and open-source models. D'Angelo delivers a clear economic breakdown showing that trailing model providers face brutal commoditization unless they build unique application-level differentiation. | |
| Incumbents versus Startups and Brand Fault Tolerance | 5 | 3 | 2 | 3 | David George pushes back on the assumption that startups win in AI by presenting the strong counter-narrative of incumbent distribution and API access. D'Angelo responds by introducing the concept of brand fault tolerance, explaining that startups like Perplexity can capture share because incumbents cannot risk non-zero error rates in their flagship products. |
Statements from this episode (19)
Quora found GPT-3 could not match top human answer quality
“We ran some experiments using GPT-III to generate answers and compare them to the answers that, that humans had written on Quora. And a lot of the time, GPT-III could not write as good of an answer as what the best human answer was that, that, that had been wr…”
D'Angelo: Chat is a better AI paradigm than Quora's publication model
“A chat kind of experience, where you can write a question and then get an answer instantly from AI, Was more likely to be the best paradigm for interacting with AI as opposed to this kind of like publication paradigm that, that Quora had.”
D'Angelo: Mass-market AI products require traditional consumer internet optimization know-how
“There's actually a lot of this kind of, like, consumer internet know-how that's important in getting a product to mass market. So this is things like building applications across iOS and Android and Windows and Mac, localization of the interface, A-B testing, …”
D'Angelo: Most AI model developers lack capacity for consumer products
“If you're a large model creator and you have You know, you have tens of employees that you can allocate to building a consumer product, and you have the culture to do that, then you can, you know, you can go direct to consumer and you can build a good product.…”
D'Angelo: API distribution is best path for most AI model startups
“And I think different startups will choose different paths here, but I think for a lot of them, the right path is going to be to just set up an API or plug into the PO API and use that to get to a lot of consumers very, very quickly.”
D'Angelo: AI creator bots will become critical for everyday tasks by 2026
“I think what we're going to see over just the next year or two is going to be incredible. This will go from being sort of useful to some people right now to being something that's just critical to many different tasks that, that anyone is going to try to accom…”
D'Angelo: Poe spends millions on inference, mainly to large model providers
“We're spending millions of dollars already on, on inference. It's mostly going to the large model providers right now, but we want to let as much of that as possible go off to these independent creators.”
D'Angelo: Quora relied on scarce human expert time unlike AI
“Too much of the Quora product has been built up around this publication model that is sort of fundamentally premised on the idea that expert time is going to be scarce. And the AI, the LLM time is not scarce in the same way.”
D'Angelo: Scaling AI models will reach human-level quality in many cases
“As these models continue to scale up, the quality is going to go higher and higher to the point where it actually will be as good as human quality in, in a lot of cases.”
D'Angelo: LLMs will never acquire unrecorded human knowledge
“There's knowledge that people have in their heads that is not in, is not on the internet and is not in any book. And so no LLM is going to have that knowledge.”
D'Angelo: AI hallucination rates will fall but never reach zero
“And I think hallucinations are gonna, the rate is gonna go down as the models get better, but it's never gonna get to the point where it's a hundred percent perfect.”
D'Angelo: AI UX will shift to direct source attribution over pure synthesis
“I expect that that is going to lead to some kind of product or some kind of user experience where the LLM is helping you sort through your sources and quoting exact experts or exact sources, as opposed to just synthesizing it all and giving you something where…”
Adam D'Angelo: AI scaling laws will drive exponential progress for many years
“So far they have held. My prediction would be there are some issues that need to be overcome, but there's just this incredible industry, so many talented people right now who are trying to Make this technology advance, and there's so much money behind it. The …”
D'Angelo: Frontier AI model providers will maintain strong profit margins
“And so it's, you know, right now it's OpenAI, Google, maybe Anthropic. Maybe, maybe Meta can be there. Those who can get there, I think it's going to be good business. You'll be able to make a lot of money. You can have good profit margins. You'll have to work…”
D'Angelo: Non-frontier AI models six-plus months behind face brutal commoditization
“I think when you go six months behind the frontier, even definitely one year, it's brutal. There's just way too many people that are able to get the capital and the resources to train those models. And so it's going to be either fully open source or there will…”
D'Angelo: AI companies must choose between scale at frontier or feature differentiation
“So I think there's going to be this kind of choice where you're either competing on scale by being on the frontier, or you're competing on some kind of, like, Feature differentiation, and in that case, you don't need a frontier model, and in some cases, you'll…”
D'Angelo: AI hallucinations favor startups over risk-averse incumbents
“The hallucination problem, that's in some ways a good thing for startups because a lot of the existing products out there have zero tolerance for anything that's going to have a risk of producing something wrong.”
D'Angelo: Perplexity is taking search market share from Google
“I think with perplexity getting share from Google right now.”
D'Angelo: Hands-on experimentation beats top-down market analysis for AI startups
“I think it's very hard to just kind of, like, think top down about where there's demand in the market. I think experimentation is really the way to go to generate ideas and to set up a, you know, a startup that's going to be able to build something really valu…”