<?xml version="1.0" encoding="UTF-8"?>
<urlset xmlns="http://www.sitemaps.org/schemas/sitemap/0.9"
        xmlns:news="http://www.google.com/schemas/sitemap-news/0.9">
  <url>
    <loc>https://samtaelabs.com/blog/tailscale-pricing-v4-unlimited-devices-personal-free</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-28T03:00:28.031Z</news:publication_date>
      <news:title>Tailscale, 개인 사용자에게 무제한 디바이스 무료 지원 — 가격 정책 전면 개편</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/vertical-ai-market-it-agency-opportunity</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-28T03:00:27.976Z</news:publication_date>
      <news:title>버티컬 AI 시장, IT 에이전시가 가장 먼저 뛰어들어야 하는 이유</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/ai-tools-developer-skill-degradation-agency-perspective</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T22:30:02.901Z</news:publication_date>
      <news:title>AI를 쓰면서 실력이 줄어드는 개발팀, 에이전시는 어떻게 다를까</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/mixture-of-experts-deepseek-cost-optimization</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T22:30:02.835Z</news:publication_date>
      <news:title>MoE(Mixture of Experts)란 무엇인가 — DeepSeek이 GPT 대비 1/10 비용으로 운영되는 구조적 이유</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/ai-coding-agent-deleted-production-database-cursor-railway-failure-analysis</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T12:30:35.955Z</news:publication_date>
      <news:title>AI 코딩 에이전트가 프로덕션 DB를 9초 만에 삭제했다: Cursor·Railway 동시 실패 분석</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/ai-era-it-agency-real-value</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T12:30:35.903Z</news:publication_date>
      <news:title>AI 시대에 IT 에이전시가 진짜 가치를 증명하는 방법</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/ai-agent-production-data-deletion-risk-for-dev-agencies</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T12:30:35.855Z</news:publication_date>
      <news:title>AI 에이전트가 운영 서버를 날렸다 — IT 외주 개발사가 반드시 알아야 할 것</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/skillify-ai-agent-failure-to-skill-methodology</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:40:26.781Z</news:publication_date>
      <news:title>AI 에이전트 실패를 재사용 가능한 &apos;Skill&apos;로 만드는 Skillify 방법론 분석</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/gpt-5-5-api-release-developer-guide</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:40:26.701Z</news:publication_date>
      <news:title>GPT-5.5 API 출시: 개발사가 지금 바로 확인해야 할 변경사항</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/openai-codex-major-update-developer-features</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:40:26.661Z</news:publication_date>
      <news:title>OpenAI Codex 대규모 업데이트: 개발팀이 지금 당장 확인해야 할 5가지 변화</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/ai-coding-tools-agency-real-world-usage</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:40:26.621Z</news:publication_date>
      <news:title>AI 코딩 도구, 에이전시는 어떻게 실전에 쓰고 있나</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/vibe-coding-production-leaf-node-principle</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:40:26.581Z</news:publication_date>
      <news:title>프로덕션에서 바이브 코딩을 책임감 있게 쓰는 법: leaf node 원칙과 검증 설계</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/ai-agent-shared-brain-wiki-markdown-git-context</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:40:26.541Z</news:publication_date>
      <news:title>AI 에이전트 팀의 공유 뇌(Wiki) 구축하기: Markdown과 Git으로 컨텍스트 일관성 유지하는 법</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/kimi-vendor-verifier-llm-inference-accuracy</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:40:26.490Z</news:publication_date>
      <news:title>오픈소스 LLM을 배포했다고 끝이 아니다 – Kimi Vendor Verifier가 드러낸 인퍼런스 신뢰성 문제</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/mini-swe-agent-100-lines-python-ai-agent-swe-bench</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:35:31.000Z</news:publication_date>
      <news:title>100줄 Python 코드로 SWE-bench 74% 달성한 AI 에이전트 &apos;mini-swe-agent&apos;</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/claude-code-routines-plaid-mcp-finance-automation-driggsby</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:34:44.982Z</news:publication_date>
      <news:title>Claude Code Routines + Plaid MCP로 개인 재정 자동화 구축하기: Driggsby 실전 사례</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/revive-abandoned-project-with-ai-coding-tool-opensubsonic-api</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-27T02:34:44.903Z</news:publication_date>
      <news:title>방치된 사이드 프로젝트를 AI 코딩 도구로 되살린 실전 사례: OpenSubsonic API 연동기</news:title>
    </news:news>
  </url>
  <url>
    <loc>https://samtaelabs.com/blog/ai-native-startup-pm-disappears-engineer-driven</loc>
    <news:news>
      <news:publication>
        <news:name>삼태연구소</news:name>
        <news:language>ko</news:language>
      </news:publication>
      <news:publication_date>2026-04-26T12:30:36.016Z</news:publication_date>
      <news:title>PM이 사라지고 있다: AI 네이티브 스타트업의 개발팀 구조 변화</news:title>
    </news:news>
  </url>
</urlset>