news

Three Tiers of Work at Genesislab, Discretionary Included [StoryPack-Genesislab ⑥]

The sixth StoryPack installment maps Genesislab's three work tiers — remote, flexible, and discretionary — and the "nim" culture that tries to keep growth and flat communication on the same table.

The sixth StoryPack installment is about work arrangements. Genesislab runs three tiers in parallel: remote, flexible, and discretionary. Remote work is negotiated by team consensus. Flexible work sets a core window of 1 to 5 p.m. and lets each person set their own arrival and departure. The newest tier, discretionary work, applies to development and research roles and judges people by the quality of their output rather than the hours they log.

The company says compensation is “at or above industry average.” Layered on top is a flattened title structure and the “nim” honorific applied uniformly across the organization, both meant to ease flat communication. CEO Young-bok Lee framed the goal as “an organization where members push positive influence on each other while growing alongside the company.”

Founded in 2017, Genesislab now has 81 people across 21 teams. What began as an HR solution has expanded into AI video-content platforms and mental-health applications. The article reads as a pattern for holding scale and culture on the same table.

Source: Digital Daily (디지털데일리) — Past Remote and Flexible — “We’re Fine with Discretionary Work Too” [StoryPack-Genesislab ⑥]

Single LLMs Hit a Wall. Here’s What Comes Next

A single LLM quickly hits its limits on complex business problems. This article traces AI architecture evolution — from single prompts to multi-agent systems — and explores what multi-agent scaling laws mean for B2B platform design.

단일 LLM의 한계를 넘어서: Multi-Agent System은 왜 필요한가

단일 LLM으로 복잡한 비즈니스 문제를 해결하는 접근은 현실에서 쉽게 한계에 부딪힌다. 이 글에서는 단일 프롬프트부터 멀티 에이전트 시스템에 이르기까지 AI 아키텍처의 발전 단계를 분석하고, 각 구조가 왜 실패하거나 부족했는지 그 이유를 짚는다. 그리고 그 흐름 속에서 도출되는 멀티 에이전트 스케일링 법칙이 B2B 플랫폼 설계에 어떤 시사점을 주는지 살펴본다.

Real-Time Labeling Interfaces: Turning Clinical Judgment into AI Training Data

A psychiatrist rating "It's tough, but manageable" as 3/3 relies on reasoning that post-hoc labeling often loses. A joint study by Genesis Lab, UNIST, and Seoul National University Hospital, presented at IASDR 2025, shows how interface design can capture that reasoning as training data.

실시간 라벨링 인터페이스: 정신과 임상 판단을 데이터로 구조화한 방법

"힘들지만 버틸 만하다"는 환자의 말을 3점이라 판단하는 임상의의 근거는 어디에 있을까요. 기존 AI는 판단의 결과만 배웠을 뿐 근거를 놓쳤습니다. IASDR 2025에서 발표된 제네시스랩·UNIST·서울대병원 공동 연구는 라벨링 인터페이스 설계로 이 근거를 데이터로 만드는 방법을 보여줍니다.

Read more

Single LLMs Hit a Wall. Here’s What Comes Next

단일 LLM의 한계를 넘어서: Multi-Agent System은 왜 필요한가

Real-Time Labeling Interfaces: Turning Clinical Judgment into AI Training Data

실시간 라벨링 인터페이스: 정신과 임상 판단을 데이터로 구조화한 방법