A startup that began in 2021 as a "chatbot aggregator" bundling GPT APIs just hit ARR $100M in five years. It's the first Korean AI startup to do so. This is the story of Wrtn.
What is this about?
Wrtn Technologies is an AI startup founded in 2021 by CEO Seyoung Lee, who was in his early twenties at the time. It started as an "AI portal" service that bundled multiple AI models, including GPT, into a single interface. It grew to the point where even people in Korea who didn't know ChatGPT knew "Wrtn," surpassing 5 million MAU (monthly active users) and claiming the #1 spot among generative AI platforms in Korea.
CEO Seyoung Lee has an unusual background. As a high school student, he organized youth debate events (KSCY), and the shift to online during COVID made him realize the power of technology. Being introduced to GPT by the founders of Daum (a major Korean tech company) sparked his entrepreneurial journey, and he was named to Forbes Korea 30 Under 30.
The real turning point came in October 2024, when Wrtn launched "Crack," an AI interactive storytelling app. What makes it different from AI chatbots like Character.AI? Crack isn't about "conversation" — it's about "narrative play". Users become the protagonist, and AI generates fiction, images, and audio in real time, with the story branching based on their choices.
"Experience-wise, it feels like there's a writer designated for you... based off your choices, the 'ghostwriter' writes it on the spot, just for you."
— Dongjae Lee, Wrtn CPO (Fortune interview)
The results were explosive. Within one month of Crack's launch, revenue hit 10 billion KRW (~$7M), and 2024 annual revenue reached 30.7 billion KRW — a 1,103% year-over-year growth. Users spend an average of 2 hours per day on the app — on par with Netflix and YouTube. Paid user retention exceeds 70%.
What's actually changing?
Wrtn's strategy is completely different from the typical approach of Korean AI startups. While most Korean AI companies focus on B2B SaaS or enterprise solutions, Wrtn pivoted to B2C AI entertainment and struck gold.
| Typical Korean AI Startup | Wrtn's Approach | |
|---|---|---|
| Target | B2B / Enterprise | B2C mass consumer |
| Revenue model | SaaS subscription / SI projects | In-app purchases + ads + subscription |
| Key metrics | Contract count, ACV | MAU, retention, time spent |
| Model strategy | Build proprietary LLM | Multi-model aggregation (GPT, Claude, etc.) |
| Expansion strategy | Domestic market focus | Korea → Japan → US sequential expansion |
| User engagement | One-off during work hours | 2 hours/day, Netflix-level engagement |
What's especially worth noting is their "multi-model aggregation" strategy. Wrtn doesn't build its own LLM. Instead, they combine the best models — OpenAI's GPT, Anthropic's Claude, and others — tailored to each service purpose. According to Anthropic's official customer case study, Wrtn explained that "in Asian markets, the way local startups beat Big Tech is through aggregation and localization."
Their global expansion pace is also impressive. They established a Japan subsidiary in 2023 and launched "Kyarapu," the Japanese version. It's tracking nearly the same growth curve as Korea, with subscribers growing 10% week-over-week. They've announced a US market entry for June 2026, and are targeting an IPO in Korea or the US by 2028.
Investment History
2021 Seed (Mashup Angels) → 2022 Pre-Series A (~$3M) → 2023 Series A (~$11M) → Pre-Series B (~$18M) → 2025 Series B (~$75M, led by Goodwater Capital). Total funding of ~$90M, making Wrtn the first Korean AI service startup to surpass $70M+ in cumulative funding. With a valuation around ~$240M, unicorn status is within sight.
The strategic alliance with FuriosaAI (valued at ~$700M) is also noteworthy. With the goal of localizing AI inference infrastructure, the partnership applies FuriosaAI's AI semiconductors to Wrtn's massive user traffic. It's being recognized as the first case in Korea's AI ecosystem where the "service + infrastructure" axis has combined.
The essentials: Wrtn's 5-stage growth formula
- User acquisition through the AI portal (2021–2023)
Lowered the entry barrier by bundling multiple AI models into one interface. Reached 1.5M MAU through free access. - Killer app pivot — 'Crack' launch (Oct 2024)
Differentiated through interactive storytelling, not just AI chatbots. Hit 10 billion KRW in revenue within one month of launch. - Non-English-speaking global expansion (2023–)
Chose Japan as the first overseas market. A strategy to preemptively capture non-English AI ecosystems. - Multi-model + localization (ongoing)
Instead of building a proprietary LLM, they combine GPT, Claude, and others by purpose. Investment goes into cultural and linguistic localization. - Entertainment → platform expansion (2026–)
Building on top of Crack/Kyarapu's entertainment foundation, expanding into productivity (AI Supporter) and experimental (Labs) services. Targeting ARR $700M in 2027 and IPO in 2028.
Risks to watch
There were concerns about organizational stability after reports of large-scale layoffs in 2025. Operating losses were also 29.5 billion KRW (~$21M) as of 2024, meaning they're still in the red. Regulatory risks in the AI entertainment market (deepfakes, harmful content, etc.) and competition from global players like Character.AI and NovelAI are also variables.
