Not a Programmer—So Why Can I Build AI Products

During yesterday’s live class, Gary asked me a question: Pinpin, you’re not a programmer—why are you able to build innovative AI products so completely and systematically? I had to genuinely think about it before answering. There’s really no secret. You could even call it lucky. It just happens that my earlier learning and work experience stacked up a few key abilities that give me a competitive edge right now.

I studied art for over a decade as a kid. What it trained most was observation and discrimination—spotting subtle differences across different dimensions of a thing. Making a product is really one choice after another, standing in the customer’s shoes and picking the more fitting solution. That process of choosing is what Sequoia has been calling aesthetics these past two years. But aesthetics alone isn’t enough. You also need taste. Taste is systematic aesthetics: not just picking which one looks better, but articulating why it looks better. It’s precisely because I have that ability that I can make sure the experience actually improves as a product keeps iterating.

Beyond aesthetics and taste, the second factor I’ve benefited from most is a user-centered way of thinking. When I first entered the industry, I was doing information architecture and interaction design—Ogilvy, Renren, then many years at Tencent after that. Whether the work was marketing or product, the first thing I considered was always the audience. What does the user need? What will the user pay for? Like the princess and the pea, I put myself in the user’s place and feel every bit of discomfort in the experience of using a product. Once that habit is in your bones, your first reaction to any product isn’t “can the tech pull this off” but “does this person feel comfortable using it.”

The third factor, I think, is something baked into my personality: a drive for uniqueness and excellence, even at the cost of disruptive innovation. I’m especially willing to accept a Sisyphean total rebuild. With AI technology iterating so fast right now, almost every project has me learning and using the latest tech. But that doesn’t mean nothing accumulates—wheels other people built and lessons from past projects absolutely get reused. It’s just that when something doesn’t quite fit, there’s zero sentimentality. I can start from scratch without hesitation. A lot of people can’t bear to tear down something half-finished. I get that feeling, but at AI’s pace, clinging to an old approach usually costs more than starting over.

Last thing. I actually think my childhood obsession with gaming has been a huge help in product development. A game system is itself a remarkably complex and addictive product model. Playing well takes serious time studying the mechanics. That ability really did seep in without me noticing. The relationships between systems, resource allocation, feedback loops—all that stuff you puzzle over in a game is the same stuff you deal with every day in product design, just under a different name.

Back to reality. Learning to build an AI product today doesn’t require writing code, but it absolutely demands real effort. At minimum, you need to learn all the relevant technical concepts—enough to understand what the AI is saying and give it accurate feedback. You don’t have to master every concept inside out, but you need to know what those terms mean and when they come up. I did a previous video that compiled over a hundred must-know concepts for AI development. Even just downloading it and memorizing it like a glossary can meaningfully raise a beginner’s development ability.

No secrets. No shortcuts. Just a set of seemingly unrelated experiences that happened to connect at the right moment. If you’re also not a programmer and you want to build AI products, then do the homework honestly and train the instincts patiently. Ability grows from the inside. You can’t copy it by learning a framework.

Source: EP0088 (2026-08-27, 5:09) · audio recorded by the host · on-screen presenter is an AI avatar · both text and timecodes come from gemini-2.5-pro ASR of that recording, with three passages re-checked by ear and one passage lost at a chunk boundary recovered on a re-run

[00:00] Yesterday in a livestream class, Gary asked me a question He said, “Pinpin, you’re not a programmer How are you able to so completely and systematically build an innovative AI product?” I really thought about it seriously before I answered him I think there’s no secret, really You could even say there’s some luck involved It’s just that in my past study and work experience I happened to accumulate the relevant key capabilities that give me this competitive edge today I studied art for over ten years as a kid What that trained most was

[00:25] observing and discriminating So what does discriminating mean? It’s seeing the subtle differences across a thing’s different dimensions Building a product is really a process of choosing, over and over It’s standing in the customer’s position and picking the solution that fits better And this process of choosing is the “aesthetics” Sequoia’s been talking about But aesthetic judgement alone isn’t enough You also need to have taste Taste is systematic aesthetic judgement You can’t just pick which one is better You have to be able to say why it’s better

[00:51] It’s precisely because of this capability that I can make sure that as the product iterates the experience actually improves So besides aesthetics and taste the second factor that’s helped me most is a user-centred way of thinking When I first started out, my job was information architecture and interaction design Ogilvy, Renren and later, many years at Tencent Whether doing marketing or product the first thing I consider is the audience I focus on what the user needs what the user will pay for

[01:16] Like the princess and the pea you put yourself into the user’s role and feel, in the experience of using the product every single point of discomfort designing the product from the user outward The third factor is what I think is a bone-deep pursuit of being distinctive and excellent even to the point of making disruptive innovations I’m unusually able to accept a Sisyphean total rebuild. With AI technology iterating this fast almost every project of mine learns and uses the newest technology Of course, that doesn’t mean I don’t accumulate the wheels other people built

[01:42] experience from past projects—that all gets reused But when I hit something that no longer fits there’s not the slightest attachment I’m entirely willing to start from zero And the last thing is I actually think that as a kid being obsessed with video games was also a huge help to my product development A game system is itself an extremely complex and deliberately addictive product model And to play one well also takes a lot of time spent digging into it That really is a capability absorbed by osmosis Today, when we go and learn to build an AI product we don’t have to write code

[02:07] But it still takes some real hard graft At the very minimum, you have to learn all the relevant technical concepts You need to be able to understand what the AI is telling you and give it accurate feedback You don’t necessarily need to understand every concept deeply I did an earlier video that collected over 100 must-learn concepts for AI development Even if you just download it and drill it like a dictionary it will substantially raise a beginner’s development ability So that’s all for today See you tomorrow