This paper, titled "The Demise of SaaS," by Jung Chul-min, CEO of Maedo-wang, details a transformative shift in software development leveraging Artificial Intelligence (AI), specifically large language models (LLMs) like Claude code. The author recounts his experience revamping his company's workflow, transitioning from a manual, error-prone system marred by communication inefficiencies to an automated, AI-driven paradigm.

The core methodology revolves around utilizing AI for rapid application development (RAD). Initially experimenting with Codex, the author found a "new world" with Claude code, describing the development process as "imagination becoming reality." This AI-assisted development allows for new feature deployment within two-hour increments, reaching a velocity where operational staff struggle to keep pace. For instance, the author was able to quickly develop custom attendance and performance management systems.

This high-speed, cost-effective custom development capability leads to the central thesis: the obsolescence of traditional Software as a Service (SaaS) models. The author's observation is that typical SaaS offerings provide a vast array of features (e.g., 100), of which only a minimal subset (e.g., 2) are truly essential for a specific business. Given the recurring per-user costs associated with SaaS, the author argues that in-house development, powered by AI, is significantly more economical and efficient. The paper explicitly states that AI development, citing Claude code, allows for creating the equivalent of "10 SaaS solutions for 22," underscoring an extremely low marginal cost for additional features or applications once the foundational AI development capability is established. This contrasts sharply with SaaS's __BLOCK_FORMULA_0__ (whereN$is the number of users,$M$is the number of months, and$P$ is the per-user monthly cost), making custom AI-driven solutions practically free after initial setup for recurring needs.

The paper draws an analogy to historical skills like typing or driving, once considered specialized professions but now basic competencies. It predicts that "vibe coding" — an intuitive, perhaps AI-assisted, approach to writing code — will become a fundamental literacy, enabling individuals without deep programming expertise to automate administrative tasks. Regarding security, the author posits that AI-driven development can mitigate common novice errors, such as insecure data transmission via GET parameters (e.g., $URL?password=secret\&card\_number=1234), suggesting that AI's design capabilities may surpass those of junior developers.

While acknowledging that specialized professionals might still produce more polished commercial products, the author highlights the dramatic leap in AI capabilities over just two years, likening it to advancing from the Stone Age to the Bronze Age. The paper concludes by contemplating the future of human skills and education in this rapidly evolving, AI-dominated landscape.