Hey guys! This week we’re covering something super topical: building AI businesses. Liner, Korea’s answer to OpenAI, is not just building a ChatGPT-wrapper, they’re building a whole new LLM itself.

When Luke Jinu Kim's AI search startup Liner went head-to-head with Perplexity's hundreds of millions in funding, the math seemed impossible. Perplexity was giving away premium features for free, ChatGPT had OpenAI's resources, and countless well-funded competitors were flooding the market.

Yet Liner, with a fraction of their budget, achieved something remarkable: 95% accuracy compared to competitors' 90-93%, and 10% penetration at UC Berkeley through pure word-of-mouth.

How do you compete when giants have infinite marketing budgets? Kim's strategy offers a blueprint for any startup facing seemingly insurmountable competition.

While competitors cast wide nets, Kim made a counterintuitive bet: go narrower, not broader. "Others just said we are AI search or we are like a productivity engine," Kim explains. "We said that Liner is AI search for students."

The math is simple but powerful. When a giant spreads $100 million across five market segments, they're effectively spending $20 million per segment. A startup with $20 million focused on one segment suddenly has equal firepower where it matters most. Kim's team "intentionally focused more on these segments: graduate students, researchers. We tried to connect with academic papers they need, patent data they need."

This wasn't about building a smaller product. It was about building a better product for a specific audience.

The key insight: Pick the largest addressable market where you can be number one. If you can't explain your focus in one sentence, you're not focused enough.

Giants optimize for scalable growth metrics, but upstarts should optimize for love. Kim's Berkeley strategy illustrates this perfectly: instead of broad marketing campaigns, they built relationships with specific student groups.

"We try to create a really good relationship with some of the student groups in Berkeley," Kim says. "When they got to kind of love Liner, they tend to introduce Liner to other student groups." The result? 10% penetration at Berkeley through organic word-of-mouth. That's a conversion rate that would cost millions in paid acquisition.

This wasn't random relationship-building. Kim's team identified the smallest influential groups in their target market and designed experiences that made users want to share. When students bookmark and share Liner's research results, they're becoming advocates for the product.

The framework: Can you name 10 users who would be devastated if your product disappeared? If not, you don't have love.

While giants buy growth, smart startups build sustainable advantages through user feedback loops. This is where Kim's 10-year vision paid off. Every user interaction (highlights, bookmarks, source visits, even hallucination reports) feeds back into improving Liner's accuracy.

"All this user feedback goes back to fine-tuning and post-training models and actually makes the models performance better," Kim explains. But it goes deeper than algorithmic improvements. "Getting to know more about the databases and websites that your customers rely on... this kind of customer insight, I don't think can be just copied in a day."

This creates what Kim calls a "data moat with user scale." While competitors can copy features, they can't instantly replicate years of user behavior data and customer insights. Each interaction strengthens Liner's position in ways that pure capital can't replicate.

The implementation: Build multiple ways for users to signal value, both explicit (ratings) and implicit (behavior). Show users how their feedback directly improves their experience. Document how each customer interaction strengthens your competitive position.

Kim discovered something counterintuitive about partnerships: what looks small to a giant can be transformational for a startup. "We are trying to build more top-down partnerships with really large companies or really large universities," he says.

The secret was to focus on partner success, not just distribution. "I don't think their partners really enjoyed the results of the collaboration" with other AI search tools, Kim observes. By ensuring partners see measurable value, Liner created sustainable relationships rather than transactional ones.

The strategy: Target partnerships that are strategically crucial for you but tactically useful for them. Better to have three successful partnerships than thirty mediocre ones. Always measure partner satisfaction alongside user acquisition.

If you’re going up against well-funded, massive peers: you have to start by going for the largest addressable market that you can be #1 in, while being indispensable for your customers.

In short, compete on your terms, not theirs.

If you enjoyed this issue, please share it with fellow founders, operators, & investors and fill out our feedback form below:

See you next week!

Rahul & Aryaman