With GPT-5, Wrtn builds lifestyle AI for millions in Korea

In Korea, millions of people turn to Wrtn’s AI apps to chat, learn, and get things done. What sets the company apart isn’t just better models—it’s how people keep discovering new ways to use its tools.

When Wrtn⁠(opens in a new window) launched, it focused on individual productivity, creating a set of AI-powered writing assistants and note-taking tools to make everyday work faster.

But productivity was always just the beginning. The team’s bigger vision was to make AI a natural language interface for everyday life.

In Korea, where digital culture already embraces character-driven platforms like KakaoTalk and LINE Friends, Wrtn saw an opportunity for what they now call Lifestyle AI: technology designed to be approachable and capable, helping people learn, create, and explore in ways that deepen over time.

To make AI more accessible to everyday users, the team invested in persona-based prompts, memory scaffolding, and careful localization so interactions felt effortless and natural in Korean slang and cadence. That work led to Crack, Wrtn’s open-ended character chat platform, which quickly became Korea’s top-grossing chat app by giving people a familiar, low-friction way to explore storytelling, language learning, and self-expression.

Today, Wrtn’s product suite spans AI Supporter, Labs, and Crack, reaching 6.5 million monthly active users. Retention has doubled, time spent per user has quadrupled, and the audience has expanded from students to professionals and families—all discovering new ways AI can spark creativity and curiosity.

One of the hardest challenges Wrtn faced was making conversations feel natural in Korean. Early GPT‑3 outputs felt like English sentences had just been run through a translation engine. 

With GPT‑4 and GPT‑4.1, the tone shifted: slang, humor, and even new internet expressions started coming through naturally. To keep responses both controllable and human-like, the team layered in system prompt injections for user notes, persona-based role guidance, and temperature tuning to make responses feel more human.

“Most Koreans don’t see any unnatural or non-native outputs now,” said Park. “This structure was exactly applicable to the Japanese market as well. So once we have this kind of localization, it can be driven by the model performance enhancement itself.”

Under the hood, Wrtn runs a router architecture that lets lightweight models like GPT‑4o mini and GPT‑4.1 mini classify and direct traffic, while heavier tasks like counseling or tutoring run on GPT‑4.1 and multimodal TTS. This modularity makes upgrades seamless: every new OpenAI release can be swapped in with minimal disruption, immediately multiplying both performance and ROI.

After a router upgrade, session time rose 15% and month-one retention jumped 10%. Adding audio models drove more than 10,000 hours of tutoring conversations in the first month alone. Users began using Wrtn’s tools for a wider range of tasks, from tutoring to storytelling, showing how approachable design can expand use cases beyond what the team first imagined.

The recent launch of GPT‑5 showed even bigger jumps, when Wrtn went live with GPT‑5 the same day it launched. “We saw instant increases in users and time spent with GPT‑5,” Park says.

Within a week, the company saw an 8% increase in daily active users. GPT‑5 delivered more consistent instruction adherence, stronger context handling, and better personalization. Wrtn is seeing performance gains with GPT‑5 across its product, including intent classification, web search, and more.

By expanding beyond productivity tools, Wrtn has shown how advanced models can be made accessible and useful for millions across learning, creativity, and communication.

Each new OpenAI release pushes that mission forward. Wrtn builds systems that keep pace with model progress and turn AI into a tool for creativity, learning, and exploration.