25-26 Fifth T19

T19 Session is the main session of GDGoC Yonsei.

• In the APT Update Session, presenters share insights on the latest technologies and tech trends.

• In the Tech Session, presenters give talks on technologies of their own interest.

✔️ APT Update

• Backend Member – Sungyun Lim

Pogocache is a lightweight cache server written in C that operates as a single binary while supporting multiple protocols including HTTP, Memcache, RESP, and Postgres wire protocol—ensuring compatibility with a wide range of clients. It can function as a standalone server process or be directly embedded into an application via its source code (pogocache.c), allowing flexible integration.

Unlike Redis or Valkey, which rely on a single-threaded model, Pogocache automatically adjusts thread usage according to CPU core availability and leverages independent event queues to achieve concurrent processing. Furthermore, instead of using Approximate LRU (commonly seen in traditional systems), it applies the 2-Random algorithm, maximizing memory efficiency without metadata overhead.

By utilizing the io_uring interface in Linux, Pogocache supports batched asynchronous I/O operations, showing remarkable performance in I/O optimization. Although still in its early stages of business application, experimental results have been highly promising.

✔️ Tech Session

• ML/AI Member – Taewan Kim

Generative Agents are interactive AI entities that simulate human behavior—free to act within a sandbox-like environment and observed experimentally. In one study, researchers let these agents operate autonomously for 48 hours, during which they spontaneously generated and shared events such as mayoral elections and parties within the simulated world.

Initially, only one agent recognized the event, but by the end of the experiment, eight had become aware and participated, demonstrating the potential of persona-based agents. This experiment highlights how AI can learn and remember through social interaction, providing insights into the future of human-like autonomous systems.

• Front-End Core – Youngjun Yoon

MPA (Multi-Page Application) architecture reloads a complete HTML page for every request, providing fast initial load times and SEO advantages but offering limited user experience.

In contrast, SPA (Single-Page Application) frameworks such as React, Vue, and Angular dynamically render all content within a single HTML file, ensuring smoother transitions and faster interactions—albeit at the cost of slower initial load times and weaker SEO performance.

Modern frameworks like Next.js, which support Server-Side Rendering (SSR) and Hydration, now bridge this gap by combining the SEO benefits of MPA with the UX strengths of SPA. The front-end field is no longer just about drawing interfaces—it has evolved into a strategic discipline balancing user interaction, performance, and search optimization.

• DevRel Core – Seoyeon Choi

Professions can be categorized as market-driven (based on market demand) or mission-driven (based on vocation or calling). Most careers depend heavily on market trends and fluctuate accordingly, while mission-driven professions—such as clergy or academics—are less influenced by external economic shifts.

In the age of AI, humans can remain valuable in the labor market through two strategies:

Achieving exceptional expertise, or

Developing a unique edge—a combination of specialization and integrative thinking (“expertise + 30%”).

When professional mastery is combined with interdisciplinary insight, humans become truly irreplaceable. The new paradigm is not “doing what machines cannot do,” but rather “becoming the kind of human who can work with machines.”