Jinu Kim and the sentence worth saving

At Philz Coffee in San Mateo in 2015, Jinu Kim had something unusually specific to tell a prospective adviser: your blog helped start our company. Sung Moon Cho later recalled meeting Kim and his co-founder while they were still college students. Kim wanted to share Cho's writing with passages highlighted, and had built an early version of the tool within weeks. The internet already offered plenty of ways to send somebody an article. Kim cared about sending the part that mattered.

There is a small social courtesy inside that idea. A link says, here is something to read. A highlight says, here is why I thought of you. Anyone who has received a thirty-page document accompanied by “interesting!” can appreciate the difference. The recipient is spared a minor archaeological expedition.

Cho helped refine the mobile app, supported its Product Hunt launch and angel fundraising, and remained a shareholder. The encounter also supplies a useful opening to Kim's career. Before the funding announcements and research agents, there was a reader trying to make a good sentence easier to find.

An art gallery takes him abroad

Kim, who also uses the name Luke Jinu Kim, studied computer science at Yonsei University. His entrepreneurial interest reached back further. As a schoolchild living in Boston during the dot-com era, he heard adults talk about startups; his parents encouraged the ambition. Programming and starting a business became linked in his imagination. Years later, he could describe his entry into working life with admirable economy: “becoming CEO is my first job title.”

At university, he formed a startup club and met Brian Woo, also known as Chanmin Woo. Their early venture, Ainno Gallery, offered online exhibitions and art distribution. It generated revenue, and recognition brought Kim to Hong Kong, where meeting companies serving much larger audiences changed his sense of what a startup could attempt.

The pair took money earned from the art venture to Silicon Valley. Kim describes a team of four working from an Airbnb, trying to produce an app each week. There were eight apps over two months, including a compliment-relay app and one that turned spare smartphone storage into a USB-like device. Liner was among that batch. His habit of underlining books had acquired a browser.

The experiments had a practical purpose. A product that worked in Korea did not automatically become a global product once its interface was translated. Kim wanted to observe people elsewhere using what he built. San Jose gave the team a place to learn that lesson at close range.

What an underline knows

Liner's early proposition was pleasingly legible: highlight text on the web, save it and find it again. The launch belonged to 2015, years before generative AI became a routine subject at dinner. A person could use the tool without needing a theory of machine intelligence. There was a useful passage. There was a way to keep it.

But the accumulated marks suggested another possibility. A reader's visit to a page reveals interest in a document. The sentence that reader selects reveals something more precise. Liner's 2016 press kit described this distinction explicitly, presenting its collection of highlights as a potential resource for AI systems. The company was thinking about sentence-level information while still explaining desktop-to-mobile syncing and exports to Evernote.

By 2018, Kim and Woo had been recognized on Forbes' 30 Under 30 Asia list. Their stated ambition was to use people's reading activity to filter the excess of material online. The modest utility and the larger search ambition were already sharing a desk.

A highlight is, of course, a human judgment about importance. Importance and truth require different tests. A memorable claim can be false; a dull footnote can be indispensable. The useful part of Kim's approach is the extra signal it supplies: somebody has already stopped here, read this and chosen it. Liner would later combine that signal with retrieval, ranking and references.

In February 2026, Kim's account of early promotion included handing sandwiches to Berkeley students to earn a little attention for the service. A sandwich is an admirably concrete acquisition strategy. It also brings the founder into conversation with someone who may be hungry, curious, indifferent or all three. Those responses can be more instructive than another slide about the size of the market.

The answer needs a way back

As Liner moved into AI search in 2023, Kim kept the same underlying concern: finding dependable information quickly. The arrival of new models changed what the software could do with retrieved material. It could compose an answer, while the company's older work on selecting information helped determine what that answer should draw upon.

For students and researchers, the consequential part was the route back to the original material. Liner attached references to generated statements. Users could inspect quoted text and decide whether the cited passage actually supported the answer. Kim has described integrating years of highlights while keeping that process intuitive as a central challenge of the transition.

This is a demanding product brief. A sentence must be readable, relevant and supportable. A citation must lead somewhere useful. The person writing an essay or preparing a report still has to decide what the evidence means. An answer that arrives quickly but cannot survive that decision has merely moved the work around.

In October 2024, Kim announced a $29 million Series B and said the United States had become Liner's largest and fastest-growing market. He pointed to students at institutions including Berkeley, Michigan, New York University, Texas A&M and USC. His emphasis was on purpose-built tools for academic work and responsible adoption. By March 2026, Liner was announcing its No. 2 position in the education category of Fast Company's annual innovation list.

“I wanted to create a product that can be used globally.”

Jinu Kim

Luke, meet L2-D2

The English name comes with a straightforward explanation: “my English name is Luke, and it's because I like Luke Skywalker.” Kim's internal customer-feedback tool is called L2-D2. The naming department, at least, has an established fictional universe.

His personal working habits are similarly specific. He asks other people questions, listens to business podcasts and sometimes looks to Chinese history for ideas. He marks tasks green when they energize him and black when they drain him, then reviews the pattern and considers what to delegate. For a founder whose original product organized marks on a page, this is an appropriately literal way to edit a working week.

He also describes the CEO's job as choosing strategy and assembling the people who can execute it. Technology changes quickly enough that learning can require unlearning. Liner's Seoul and San Francisco teams work across a considerable time difference; his preference for regular office collaboration reflects the value he places on people thinking together.

Kim's writing is less breezy than the robot's name. In a 2022 post on Jinu Notes, he recalled expecting software uploaded to the internet to attract enormous audiences, then experiencing the opposite. He wrote about choosing global ambition and persisting with a product others dismissed as merely a highlighter. The post gives the later growth a useful grain of friction. The early version of the future was considerably easier to imagine than to persuade people to use.

From finding a paper to finishing the work

During 2026, Kim's focus widened from the answer to the document someone hoped to produce with it. Research usually has a destination: a paper, presentation, report or proposal. Searching, reading and drafting are connected activities, even when the software puts them in separate windows. The repeated journey between those windows is familiar to anyone whose browser tabs have become a form of interior decoration.

Liner introduced Figure Generator on Liner Scholar in March, letting researchers select a passage and request a visual explanation. In April, Kim described a workflow in which AI prepares an initial draft and a person refines it. He calls the broader approach “Vibe Research.” Behind the fashionably relaxed name sits a fairly exacting requirement: the output must remain grounded in material that exists.

The new tools extend the old highlight in an interesting direction. Originally, selecting text helped someone remember it. Now, the selected material can become input to a figure or part of a larger research workflow. The mark has moved from the end of reading toward the beginning of another task.

Kim also discusses evaluating documents that contradict one another. Simply collecting links cannot settle a disagreement. Ranking and reasoning have to help determine which claims deserve confidence. That is where his ambitions become harder, and more useful, than making a tidy summary. A polished paragraph can disguise disagreement remarkably well. A research tool has to leave the reader equipped to notice it.

The next reader may be an AI

On August 25, 2026, Liner announced a $36.1 million Series C led by LB Investment. The company reported more than 14 million registered users. Registered users are a measure of accumulated reach, rather than a count of people using the product every day, but the figure establishes the distance from the first highlighter experiments.

The next stage changes the customer as well as the interface. Kim is expanding enterprise work and making the company's search infrastructure available to other services. Liner's relationship with SK Telecom includes work on reliability and accuracy within its Korean foundation-model consortium. Search expertise developed around readers is becoming a component other organizations can use.

In September, Kim described an ambition for Liner to become the search engine that AI systems themselves rely upon. He also set out a possible public-listing horizon of two to three years, tied to the company's financial performance. These are plans, with work still to do. Their significance is the change in vantage point: another AI product may need Liner even when the end user never opens a Liner window.

That would put Kim several steps away from sending a highlighted blog post to a friend. Yet the continuity is easy to see. Someone asks a question. Somewhere, a passage helps answer it. Someone must decide that the passage matters, locate it and make the connection visible. Kim started with a sentence worth saving. He is still working on what happens next.