Chen says experience helped him set realistic expectations about how difficult and slow startups can be. Founders also need to connect a long-term vision to what customers need now, and test their assumptions against reality. He argues that a founder’s self-awareness can set a limit on how well the company performs.
AI Hacks from a 3x Founder to Scale Your Business and Life
Speaker
Siqi Chen, co-founder and CEO of Runway, discusses company building, practical uses of AI, and startup persistence with Peter Yang on Peter Yang’s channel.
Runway refreshed its values to reflect language and behaviors already present on the team. Chen highlights “Give a F***” as a way to describe people who care deeply about their work, and “Raise the Bar” as a commitment to hiring stronger teammates and pushing the quality of the work. He sees the company website as part of its marketing and hiring work, not separate from the product effort.
Chen says Runway had just under 50 employees and expected AI to let a small team do more without scaling headcount as quickly. He built a Zapier workflow that checks inbound leads, enriches them with company information, asks ChatGPT to assess fit, and drafts personalized emails for qualified prospects. For getting started, he recommends using tools like ChatGPT and Perplexity regularly, then trying automations and code-generation tools.
Chen describes Universal Primer GPT, which teaches a complex topic through analogies, identifies missing prerequisites, and adapts lessons to what a learner already knows. He also says his children used ChatGPT to develop a game idea after seeing a Gundam statue in Japan, then began building it with Unreal Engine. He believes self-directed curiosity can help people learn useful skills.
Chen says Runway spent roughly two and a half to three years before seeing meaningful evidence that its idea could work. He advises founders to keep going while they still believe in the mission, while staying flexible about how to reach it and learning from customers. Runway aims to make business finance easier to understand by connecting models to real company data and giving employees context about how their work affects the business.
Takeaways
- Set expectations for startup difficulty so uncertainty is easier to manage.
- Use customer feedback to adjust the path while holding onto the underlying vision.
- Automate repetitive lead qualification with clear criteria and useful company context.
- Ask AI to explain its work and teach missing prerequisites, not just produce an answer.
- Connect financial models to actual results so teams can understand and revise their assumptions.
Mentioned in this talk
Search inside this talk
What was claimed 19
- ceo ofCEO
“co-founder and CEO of Runway”
YouTube Unverified1 source· Talk
- founded
“my guest is Siqi Chen, co-founder and CEO of Runway”
YouTube Unverified1 source· Talk
- metricRunway: Just under 50 people
“We’re just under 50 people right now.”
YouTube Unverified1 source· Talk
- uses
“we built a no-code automation to qualify them”
YouTube Unverified1 source· Talk
- uses
“we’ve since switched to Default”
YouTube Unverified1 source· Talk
- integrates with
“I built a Zapier automation”
YouTube Unverified1 source· Talk
- uses
“Then it calls Clearbit to get information about the person’s role, company, company size, and market.”
YouTube Unverified1 source· Talk
- uses
“I thought this was doable with GPT-4”
YouTube Unverified1 source· Talk
- feature
AISDR: Qualifies inbound leads and drafts customized emails for suitable prospects.“Then it asks ChatGPT to come up with a creative line connecting the two, draft an email, and put it in my Gmail drafts folder.”
YouTube Unverified1 source· Talk
- built
“you have this GPT that you built a while ago called Universal Primer GPT”
YouTube Unverified1 source· Talk
- metricUniversal Primer GPT: Number one GPT in the Education category in the ChatGPT Store
“It's actually the number one GPT in the Education category in the ChatGPT Store.”
YouTube Unverified1 source· Talk
- featureUniversal Primer GPT: Explains a topic with intuitive analogies, identifies prerequisite gaps, and teaches missing concepts recursively.
“It gives you an intuition for how the pieces fit together first”
YouTube Unverified1 source· Talk
- uses
“I built an AI version”
YouTube Unverified1 source· Talk
- featureAI email labeling automation: Labels incoming email according to natural-language rules for categories and content.
“It describes all the possible labels and what kinds of emails should receive each label, and then the automation applies the label.”
YouTube Unverified1 source· Talk
- integrates withAI email labeling automation integrates with Zapier
“I published it online somewhere, along with the Zapier setup.”
YouTube Unverified1 source· Talk
- uses
“I use GPT-4 on all my emails.”
YouTube Unverified1 source· Talk
- metricRunway: Close to $10K across apps
“between all of our apps, we’re probably paying close to $10K at this point.”
YouTube Unverified1 source· Talk
- metricAI email labeling automation: Less than $100 per day
“Oh, it’s less than $100 per day.”
YouTube Unverified1 source· Talk
- metricRunway: Early access; general access expected soon
“We’re still in early access, so you can sign up now, but we’ll be launching general access soon.”
YouTube Unverified1 source· Talk








