Blog AI & HCI
Local LLM Agent on a GPU-less
Claude(CLI) 연동의 반복 인증과 토큰 소모가 불편해서 로컬 모델로 바꿔봤다가, GPU 없는 하드웨어의 더 근본적인 벽에 부딪혔다
Originally ran OpenClaw on this machine with Claude (CLI) as the model provider, but recurring re-authentication and heavy token usage made it impractical for regular use. Tried swapping in a local model instead — Ollama (local LLM server) + OpenClaw (agent gateway) + Telegram (remote …
Read more →What Is AGI?
"AGI 시대가 왔다"는 말은 결론이 아니라, 어떤 정의를 기준으로 삼느냐에 따라 달라지는 주장이다.
OpenAI's September 2026 launch of GPT-6 Astra came with a bold claim from President Greg Brockman: "we've entered the AGI era." Nvidia CEO Jensen Huang echoed it days later. But the term AGI has no single agreed-upon definition — so the same announcement means different …
Read more →Ollama: Local LLM in Two Proje
클라우드 API 비용 없이, 로컬에서 LLM을 돌려 두 프로젝트의 추론 엔진으로 써봤다.
Ollama is a local runtime for open-source LLMs (Llama, Qwen, Gemma, DeepSeek, Mistral, and more), released in 2023 and now widely used as the de facto standard for running LLMs locally. Pull a model once, then call it over HTTP from localhost with no cloud …
Read more →
Codex vs Claude Code: 2026 Com
같은 문제를 다르게 푸는 두 코딩 에이전트 — 성능, 비용, 안정성, 커뮤니티 여론까지 비교해봤다.
A deep-dive comparison of the two dominant agentic coding tools of 2026 — OpenAI Codex and Anthropic Claude Code — across benchmarks, usability, pricing, stability incidents, and community sentiment, with guidance on when each tool has the edge.
Core framing: the two tools solve …
AgentRead more →
Harness, Loop, and Graph Engin
프롬프트 엔지니어링 다음에 나온 세 용어, 결국 같은 질문을 다른 해상도에서 묻고 있다.
A short explainer piece on three terms that emerged in rapid succession in 2026 — harness engineering, loop engineering, and graph engineering — and how they relate to each other.
The core claim: all three are attempts to achieve predictable, consistently high-quality output for repetitive …
AgentRead more →