숲 해설을 위한 영어 공부
오랫만에 Software Architecture에 관한 책을 읽었다. 이 책을 선택한 배경은 Software Architecture에 관해 비교적 최근(2018)에 나온 책이고 저자가 그 유명한 밥아저씨(Uncle Bob, Robert C. Martin)이기 때문이다. 이 책을 읽으면서 과거 한 때 Software Engineering에 심취해서 여러가지 개념들, 그리고 실제 코드로 적용해 본 기억들이 나면서 한 때 사용하지 않아 소실 됐던 근육이 다시 단련되는 기분이다. 책 전반적으로 실무 코드를 기반으로 설명하는 것이 아니라 개념적으로 접근하고 있어 아마도 초보 개발자들에게는 좀 어려운 내용이지 않을까 생각한다. 하지만 한 번 쯤 귓등으로라도 읽어 놓으면…
While decoupling sounds great in theory, it often introduces invisible coupling – dependencies on event contracts, formats, topics, and message semantics. Here is where things get messy: What was meant to simplify scalability ends up complicating development and operations – especially for small and mid-sized teams.
The Model Context Protocol (MCP) is a standard developed by Anthropic that enables AI models to use tools by defining a structured format for tool descriptions, calls, and responses. It defines a standardized way for LLMs to discover, call, and use tools, often hosted on local or remote servers, via HTTP + JSON RPC2.0. Function calling is like giving LLM a calculator app. MCP is like giving LLM an App Store — which can browse, discover, and use any available tool on its…
The company also introduced a new tool that helps computer programmers use chatbots when writing code. In September(2024), OpenAI introduced A.I. technology that could “reason” through tasks involving math, coding and science. Now, this technology can tackle similar tasks that involve images, including sketches, posters, diagrams and graphs. On Wednesday(2025.4.16), the company unveiled two new versions of its reasoning technology called OpenAI o3 and OpenAI o4-mini. Each can handle tasks that involve both images and text. These systems can “manipulate, crop and…
A software architect is a programmer; and continues to be a programmer. Never fall for the lie that suggests that software architects pull back from code to focus on higher-level issues. They do not! Software architects. are the best programmers, and they continue to take programming tasks, while they also guide the rest of the team toward a design that maximizes productivity. Software architects may not write as much code as other programmers do, but they continue to engage in…
Andrew Barto and Richard Sutton developed reinforcement learning, a technique vital to chatbots like ChatGPT. In 1977, Andrew Barto, as a researcher at the University of Massachusetts, Amherst, began exploring a new theory that neurons behaved like hedonists. The basic idea was that the human brain was driven by billions of nerve cells that were each trying to maximize pleasure and minimize pain. A year later, he was joined by another young researcher, Richard Sutton. Together, they worked to explain human…
요즘 Spring을 활용하는 방법에 익숙해 지려고 노력하고 있다. 사실 Spring은 거의 15년여 전 Spring Framework 초기에 잠깐 사용해보고 그 이후는 경험이 전무한 터라 개념 정도는 얘기할 수 있을지라도 요즘 개발자들과 Spring 활용에 대해 말을 썩을 수 있는 처지가 못된다. 이 참에 Spring에 녹아있는 핵심적인 기술들을 하나씩 꺼내와서 이해하고 정리해 보려고 한다. 이 글은 그 시리즈의 첫 번째로 Spring Cache에 대한 내용이다. 인터넷에서 흔히 소개되고 있는 내용들 보다 좀 더 깊숙이 들어가서 어떤 매커니즘에 의해 Caching이 적용되고 동작하는 지 살펴보도록 한다….
DeepSeek challenged a tech industry consensus that to build bigger and better A.I. systems, companies would have to build bigger and more powerful data centers. It set off fears that companies might pullback on their spending with Nvidia. Since then, a new consensus has emerged that Nvidia will continue to benefit because it will become affordable for more companies to develop A.I. systems. An expanded field of A.I. business would create more customers for Nvidia’s expensive chips, not fewer, as…
The current frontier of AI systems is based on two model types(they are almost identical under the hood, but their behavior is notably different in practice, hence the distinction) Pre-trained models, also known as ‘non-reasoning models’ These are the famous ‘Large Language Models’, or LLMs, gigantic AI models trained on as much as data as possible, reaching double digits of trillions of words (for reference, Lama 3.1 405B was trained on 15 trillion tokens ~ 11-12.5 trillion words, and DeepSeek…
The Chinese start-up used several technological tricks, including a method called “mixture of experts” to significantly reduce the cost of building the technology. A.I. companies typically train their chatbots using supercomputers packed with 16,000 specialized chips or more. But DeepSeek said it needed only about 2,000. DeepSeek engineers needed only about $6 million in raw computing power, roughly one-tenth of what Meta spent in building its latest A.I. technology. What exactly did DeepSeek do? How are A.I. technologies built? Companies…