The World’s Most Efficient AI Startup is Korean

In the rapidly evolving landscape of Generative AI, a Seoul-based startup is making waves in the research and information retrieval space. Liner, with its innovative approach to AI-powered research, has quickly captured leading market share, giving established players like Perplexity a run for their money. In an exclusive interview with StartupHub.ai, Luke Jinu Kim, the founder and CEO of Liner, shared insights into the company's journey, technology, and future plans.

Liner's mission is clear, yet ambitious: to help people get smarter, faster. There’s no facade or grandiloquent jargon, just simple language reinforced with a clever technology stack and exemplary product execution. "We've been doing a lot of things that can help people get smarter faster," Kim explained.

Back in 2015, the startup’s journey began with a web highlighter pen that users could install on their Chrome browser. From there, they expanded into recommending high-quality information as a feed. However, it was the emergence of Large Language Model (LLM) technology that truly catalyzed Liner's current direction.

To date, the 40-person team is hitting 30 million web visitors per month. By comparison, Anthropic’s web app Claude does 65 million, valued at over $2.5 billion, and Perplexity does 80 million, rumored to be valued around $3 billion. These metrics don’t add up in the Silicon Valley calculus, making Liner an absolute diamond-in-the-rough. Andreessen Horowitz christened their status when they published an analysis of rankings earlier this year, fourth among the top 100 Generative AI web applications by monthly visits, ahead of Quillbot and Poe.

Performance like that is an emblem of ultra efficiency, placing them among startups like Zapier.

Liner's flagship products are Liner AI Copilot, a browser extension, and Liner AI Workspace (Web and App), a ChatGPT-like AI Search Engine. It sets itself apart in the increasingly crowded AI research space by focusing on reliability and trustworthiness. "Liner gives very reliable answers based on trusted sources," Kim stated. What makes Liner unique is its community-driven approach to curating information. Similar to how Pinterest users pin high-quality images for others to consume, Liner users contribute to a collective pool of high-quality, highlighted information from across the web.

This unique approach addresses a critical issue in the AI research space: the reliability of sources. "If you ask about good foods and bad foods for diabetes patients, Liner will actually use articles from Johns Hopkins and Harvard Medical School," Kim illustrated. “Our focus is not just giving answers for a web page, it’s giving reliable answers based on trustworthy web pages,” he said in comparison to Perplexity’s typical outputs. With that element of differentiation, Liner has found a strong foothold in the higher education market, particularly in the United States.

"90% of our paying users are in higher education, and 60% of them are in the United States," Kim revealed. The appeal for students and researchers lies in Liner's ability to provide answers with exact sentences referenced from high-quality web pages, eliminating the need for double-checking on Google and providing reliable sources for academic work.

Aligning perfectly with wary pundits of the potential drawbacks Generative AI may bring to academia, model hallucinations, unverified sourcing and dulling of skills combined with the impending doom arguments analysts harp about, Liner in itself is a new skillset. It could be argued that not using a tool like Liner puts students at risk for the future of work.

The startup’s high demand from the hallways of academia aligns with the growing demand for AI skills in the job market. PwC’s 2024 Global AI Jobs Barometer report notes that for every job posting requiring AI specialist skills in 2012, there are now seven job postings. Moreover, jobs that require AI skills carry up to a 25% wage premium in some markets, highlighting the value of AI proficiency in today's workforce.

It’s in those sectors, with well-versed AI tool usage by workers, that experience almost fivefold gains in productivity, a testament to the effectiveness of tools like Liner.

Behind their flawless product execution, the startup leverages a combination of LLM models. "We use GPT-4o, GPT-3.5 Turbo, and Claude," Kim shared. The company has also fine-tuned Llama 3 for their needs, creating what they call "Liner 7B".

Liner has also introduced new features as of late capitalizing on Agentic AI workflows. Kim sees Liner as an early example of an AI Agent, replacing human tasks like deciding which links to click, reading documents, and extracting information. "We’ve gone further because there are more steps that we can replace using AI Agents," Kim explained surrounding the introduction of tapping task-specific AI Agents across subscription plans, like an Agent for summarization or coding.

Liner is in competitive pocket of the Generative AI market, each vying for consumer usage with slightly different feature releases. The table below puts the competition into perspective, with web visits sourced from Similarweb.com (although none are verified).

As for building their own foundation model, Kim sees it as a long-term goal. "It's on our roadmap, but it's not in the near future because it needs a lot of funding," he explained. In the meantime, Liner is focused on fine-tuning existing foundation models to better serve their users' needs.

Liner's long-term vision extends beyond providing answers. Kim outlined a three-step process in research: ideation, searching and learning about concepts, and applying knowledge to create outcomes. "We will expand to other use cases like this," Kim hinted, suggesting that Liner aims to support users throughout the entire research and learning process.

Liner AI Workplace has mobile apps (iOS, Android), PC apps, and web versions. Their AI Copilot can be installed on web browsers such as Chrome and Edge to serve as AI assistants.

With their last funding round in August 2022, Liner is now gearing up for their next phase of growth. "We are quite ready right now to raise our next round," Kim stated. While specific details weren't disclosed, the startup is actively engaged in investor discussions.