Siqi Chen

Competitors are free R&D when you can out-execute anyone

Siqi Chen asked Mark Pincus how Zynga could have a moat in social gaming when anyone can clone a game. Pincus said he WANTED more new entrants because they were free R&D — Zynga would just fast-follow and out-grow them, which is how FarmVille and their poker game won despite not being first.

“I was like, no, this is great. I want there to be more new entrants into the space because it's free R&D for me. And I just like, okay, that is, that blows my mind. That is next level because he was just so confident in his ability to execute that like anyone who's going to come in with some new idea, they can just like fast follow it and do a much better job of growing it, which is like what they did, right? Farmville wasn't the first farm game.”

The Slide acquisition meeting from the Silicon Valley hall of fame

A year out of college with a viral Facebook app, Siqi Chen met Max Levchin and Keith Rabois at Slide. They forgot the meeting, told him his $2M ask would make him 'the 18th or 19th most important person,' said they couldn't wait to buy shitty companies cheap in the coming recession, and quizzed each other on PayPal-era body fat.

“And they said verbatim, you know, we bought Superpoke and he's easily the 18th or 19th most important person in the company now. Out of a company of like 60 people. I said, okay, that's a compelling offer. And then I said, and I said, yeah, this is probably not gonna happen. And I said, you know, we're profitable, we were making a bunch of money from ads, so it's cool if you don't want to buy it. And they said, well, you shouldn't confuse revenue for success.”

Siqi Chen's Amplitude angel bet returned 400x

After selling to Zynga, Siqi Chen got into the seed round of an analytics startup as its fourth customer. The company became Amplitude, IPO'd, and returned 400x on his investment — his best angel win.

“I managed to get into the seed round and that IPO that, you know, we ended up being a 400x return on that, which is pretty great.”

Trainable AI email inbox that understands your whole life

Siqi Chen wants email software where you can customize and train the AI prompt to sort mail by context (e.g. 'anything about my kids goes here'), not just keyword rules. He says he's seen 12 companies attempt it and none have nailed it.

“What's missing is there's no way to train it. So I think building some way for you to, for a product to understand all of the content about your life and organize your stuff, starting with inbox, will be really handy for me. And I've seen at least 12 companies do this and no one has done a really great job of it.”

Steal thisBuild email triage where users write and train their own natural-language sorting prompts instead of rigid keyword filters.

A fake dating app where every match is AI and obsessed with you

Siqi Chen's 'spicier' money-printing idea: a Tinder-style app where every profile is AI-generated, attractive, and super into you — and each AI persona even has a real Instagram account you can follow off-app. He notes AI companion apps already generate millions a month.

“the thing I would make is something that is kind of like Tinder. But basically everything is AI generated, so it's not like, oh, you're creating a robot, but it's a fake dating app where everyone is attractive and is super into you. And then you can like, then you can go off of your Tinder app and go on Instagram and there have an Instagram account on actual Instagram. It's owned by the, the, the person that you met on your fake Tinder app.”

Zynga's 'crew' mechanic: require 20 unique helpers, not the same two friends

Zynga found players asking for materials always pinged the same two friends. By requiring ~20 UNIQUE people to unlock something (the 'crew' mechanic), they widened viral distribution and got a large boost in daily active users.

“One of the mechanics that we invented at Zynga is this idea of crew where you, for whatever thing you want to unlock, you have to get at least like 20 people to help you, unique people. Because what that did is, what we saw in the data is that when you do materials, you ask the same 2 people over and over again. But if we have a unique spread, then that increases this distribution. And that ended up being like a pretty large boost in DAUs.”

Steal thisGate a viral reward on help from many UNIQUE users, not repeat asks to the same friends, to widen your invite surface.

Deep Research that logs into your paywalls with your credentials

Doing medical research, Siqi Chen hit the wall that AI agents can only read the open web. He wants a Deep Research tool where you enter your paywall/login credentials so the agent can access full-text journals and other auth-walled data — the more private data it can reach, the more useful it becomes.

“I really wish someone would make a version of Deep Research that lets you enter your paywall credentials so you can get full text access. And this goes further. I think there's just so much data behind auth walls, paywalls that you can't get to. You can only search on the open web and the more private data you can get access to, the more useful these agents become.”

Steal thisBuild a research agent that securely stores user login credentials and reads full-text content behind paywalls, not just the open web.

A 1970s pinworm drug may treat cancer — but it's unpatentable, so no trials

Siqi Chen explains a broken incentive in medicine: an off-patent, non-prescription pinworm drug from the '70s has 20 years of compelling data suggesting it could treat some cancers, but no clinical trials exist because you can't patent it and recoup costs. R&D money flows only to patentable new molecules.

“there's this, uh, drug that is FDA approved, non-prescription, uh, it's been out since the '70s to treat pedworms. And over the past 20 years, there's a huge amount of compelling data. This might be a pretty good treatment for different kinds of cancer. And there's been no clinical trials for it because there is no money for it because you can't patent a pinworm drug. And so all of the money, hundreds of millions, billions of dollars are going to new molecules that are patentable, even if things are already available.”

A motivated dad out-researched neurosurgeons on his daughter's rare tumor

After his daughter's rare brain tumor diagnosis, Siqi Chen did primary research and proposed a third treatment path to her neurosurgeons and clinicians. A week later they agreed his path made more sense — possible because, with such a rare disease, he knew more about it than specialists who must study 50 cancers.

“I proposed a third path and they were discussing it, and this is like neurosurgeons and clinicians. They came back a week later, it's like, actually your path makes more sense. And the reason I was able to do that is like, because this disease is so rare, I am like more knowledgeable about the disease than anyone in the room because they have to like study 50 different cancers.”

Two tracks, five levels: how to actually build trust with people

From Stanford's 'Touchy Feely' (Interpersonal Dynamics), Siqi Chen lays out a model: every conversation runs a content track (facts) and a relationship track (emotion). You communicate at 5 escalating levels — ritual, extended ritual, content, emotional self-disclosure, and mutual emotional self-disclosure. Only levels 4 and 5 fill the relationship track and deepen trust.

“So level 5 is the deepest one. And level 5 is mutual emotional self-disclosure. And it is when you are expressing the emotion that you have about the other person. I feel angry at you. I feel proud of you. I feel disappointed by you. That's the deepest level of communication you can have with another person. And the content track is only filled by things from level 1 to 3. And the relationship track is only filled by level 4 and 5.”

Steal thisTo deepen a relationship, deliberately use level-4/5 statements ('I feel proud of you') instead of staying in facts-only content mode.

'The story I tell myself' — share emotion as information, not attack

Siqi Chen's nonviolent-communication move: emotion leaks whether you voice it or not, so choose words over passive aggression. You're entitled to share your feelings as facts but not to mind-read others. The script: 'When I see you do this, the story I tell myself is that you don't respect me… I feel angry, and I wanted you to know.'

“But if I were to say, you know, when I see you do this, the story I tell myself is that you don't respect me. And I don't know if that's true, but this is like what I'm thinking in my head. And because of that, I feel angry. And I just want you to know that because I don't, I don't know if you know that you, I know that you probably don't because you can't read my mind, but I'm guessing you probably aren't intending to make me feel that way.”

Steal thisFrame hard feedback as 'the story I tell myself is X' so you share your emotion as information instead of an accusation.

The interview question that breaks the fourth wall

Siqi Chen's favorite interview question after 20 years: 'What is your greatest strength that you're most worried about NOT coming across in this interview setting?' It breaks formality, gets candidates excited to tell real stories, and surfaces strengths a standardized interview would miss.

“if you just ask what's your greatest strengths, it sounds really formal, but what is the insecurity that you're bringing in that you're hoping to, that you're worried about not coming across in an interview setting? It changes the tone quite a bit.”

Steal thisAsk candidates which of their strengths they fear won't show up in the interview — it unlocks honest, energetic answers.

ElevenLabs hit hundreds of millions ARR in ~1.5 years

Siqi Chen, an early investor, recalls ElevenLabs as fewer than 10 ex-Google people building foundational audio models. Roughly a year and a half later they're at hundreds of millions in ARR — he believes they may be growing faster than any top SaaS company, including Cursor.

“And they are now, I don't know what they last announced, but they're in the hundreds of millions AR in like a year, two, a year and a half or so. Um, and what, what blew my mind is I never seen a group of very good technologists that were also so good at commercializing the technology so rapidly.”

Naval named Runway and told him to drop the 'AI'

Siqi Chen wanted to name his finance startup CFO.AI. Naval Ravikant told him not to — 'AI' would be dated in two years when everything is called AI — and suggested 'Runway' with the .com. Naval wrote the first check, and Siqi Chen later paid $250K for runway.com.

“He's like, no, don't call it AI. It's going to be dated in like 2 years. Everything's going to be called AI. You should just call it Runway and you get the .com. And he wrote the first check into Runway.. And that's why we got the .com.”

Steal thisAvoid trend-of-the-moment names like '.AI' that will date your company — pick a name that ages well and lock the .com.