Video · 2026-03-15 · 1h 4m · 6 moments

ChatGPT – The Super Assistant Era | BG2 Guest Interview

✦ AI generated

timeline · colored by role

Claim◆

Long-term retention is the metric OpenAI prioritizes above all others for ChatGPT, because it signals that the product durably solves people's problems, and revenue follows from that rather than needing to be optimized directly.

Asked to allocate 100 priority points across possible metrics, Turley puts them all on long-term retention, arguing it's the truest signal of product value and that revenue naturally follows.

transcript

Nick Turley: I care a lot about long-term retention and I would put all my points there. Um because I'm really proud of the retention stats we have. Huge, but ultimately the sign of durable value[s], whether or not people are coming back in three months because that means you're really solving their problems. And I think things like revenue, they follow from that.

explains mechanism · 1

Data

ChatGPT's historical growth has come from a roughly even three-way split: removing friction (like eliminating the login wall), core product investments like search and personalization, and step-change model improvements like GPT-4.

Turley attributes ChatGPT's growth to three roughly equal forces: classic UX friction removal, joint research/product investments like search and personalization, and headline model upgrades.

transcript

Nick Turley: when I look backward, it's been roughly sort of one third one third one third between sort of classic friction removal type of work... another third or so is sort of core product investments... and then another third of the growth has been just model improvements like step changes... going from GPD 3.5 back then to GPD4 then going from GPD4 behind a paywall to suddenly available to everyone.

Claim◆

Because delegation is not a natural skill for most people, ChatGPT still functions too much like a raw power tool or computer terminal that users must discover on their own, and it needs to evolve into something with more built-in affordance, more like an operating system, to reach the next billion users.

Turley argues that most people don't naturally know how to delegate problems to AI, so ChatGPT's biggest opportunity for growth is becoming more self-explanatory and proactive rather than requiring users to discover its power on their own.

transcript

Nick Turley: I don't think delegation is a natural skill for most... chatgpt is a pretty it's a power tool right you come to it, it doesn't tell you what it's for, you kind of have to discover it on your own... I still feel like we're a little bit too much like a computer terminal and it needs to feel more like software, like an operating system of software.

Mechanism

Domain-specific AI agents, like Codex for coding, already work well because coding is testable and RL-friendly, and this pattern will likely extend to other quantitative knowledge work before general-purpose agents that can handle any task become viable.

Turley says narrow, domain-specific agents (especially coding agents like Codex) have already hit escape velocity because their outputs are verifiable, while general-purpose agents for arbitrary tasks remain the harder, still-unsolved goal.

transcript

Nick Turley: the thing that's already come first is the domain specific agents, right? If you look at what's happening in code, we're fully there... it's testable, you know, if it worked or not, it's very RL friendly... the domain specific ones already work. I think the thing everyone's working for is general purpose agents that just kind of work for anything.

gives example · 1supports · 1

Anecdote

ChatGPT was originally free because it was meant to be a one-month demo; OpenAI later shipped subscriptions not primarily to generate revenue but to shape demand and gracefully turn away users when the product hit capacity.

Nick Turley explains that ChatGPT's subscription model emerged accidentally, as a way to ration access to a capacity-constrained product rather than as a deliberate monetization strategy.

transcript

Nick Turley: chat GPT originally was entirely free and the reason for that was that it was intended to be a demo and we were going to wind it down after a month we then realized that the demo went viral And people loved the demo and it was actually a product... we realized to be a product you can't take the product down every time you're at capacity. So we ship subscriptions simply because it could shape the demand.

Anecdote◆

The moments that felt like AGI to Nick Turley were GPT-4 writing poetry and producing working, compiled code — capabilities he hadn't believed were possible for a language model — and later a reasoning model spontaneously swearing in its chain-of-thought after catching its own mistake, a behavior that emerged entirely from the RL process.

Turley recounts his personal 'AGI moments': being stunned that GPT-4 could write poetry and working code, and later watching a reasoning model spontaneously swear and self-correct mid-demo, an emergent behavior no one had programmed.

transcript

Nick Turley: there was two things that GBD4 did that felt like AGI to me. One is it could do poetry and I didn't think it was possible for AI model to do poetry... and then the other one was it could produce code that actually worked and compiled... the model swore and said like, 'Oh, damn it, I have to adjust because I realized I had made a mistake in the puzzle.' ...the fact that it did that in a way that was entirely emergent from the RL process completely blew my mind.

Highlight slides

Retention Over Everything✦ from: Long-term retention is the metric OpenAI prioritizes above all others for ChatGPT, because it signals that the product durably solves people's problems, and revenue follows from that rather than needing to be optimized directly.Why Retention Leads✦ from: Long-term retention is the metric OpenAI prioritizes above all others for ChatGPT, because it signals that the product durably solves people's problems, and revenue follows from that rather than needing to be optimized directly.Delegation isn't intuitive✦ from: Because delegation is not a natural skill for most people, ChatGPT still functions too much like a raw power tool or computer terminal that users must discover on their own, and it needs to evolve into something with more built-in affordance, more like an operating system, to reach the next billion users.From power tool to operating system✦ from: Because delegation is not a natural skill for most people, ChatGPT still functions too much like a raw power tool or computer terminal that users must discover on their own, and it needs to evolve into something with more built-in affordance, more like an operating system, to reach the next billion users.Nick Turley's AGI Moments✦ from: The moments that felt like AGI to Nick Turley were GPT-4 writing poetry and producing working, compiled code — capabilities he hadn't believed were possible for a language model — and later a reasoning model spontaneously swearing in its chain-of-thought after catching its own mistake, a behavior that emerged entirely from the RL process.Emergent Behavior in a Reasoning Model✦ from: The moments that felt like AGI to Nick Turley were GPT-4 writing poetry and producing working, compiled code — capabilities he hadn't believed were possible for a language model — and later a reasoning model spontaneously swearing in its chain-of-thought after catching its own mistake, a behavior that emerged entirely from the RL process.

Related episodes

Video · a16z Podcast

Software in the Age of Agents | The a16z Show

→

Audio · All-In Podcast

SpaceX-Cursor Deal, SaaS Debt Bomb, New Apple CEO, SPLC Indictment, Colon Cancer Spike

→

Video · The Cognitive Revolution

Lindy Teammate: Flo Crivello on Multiplayer Agents, Memory & Why He'd Ban the Chinese Models He Uses

→