Practical writing on LLMs, agents, coding tools, and real workflows.https://flowhat-blog.pages.dev/https://flowhat-blog.pages.dev/blog/2026-03-31-why-anthropic-added-auto-mode-to-claude-code/https://flowhat-blog.pages.dev/blog/2026-03-31-why-anthropic-added-auto-mode-to-claude-code/Anthropic's Auto Mode matters because coding agents are now competing on autonomy controls, approval design, and harness quality, not just raw model ability.Tue, 31 Mar 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-03-31-why-anthropic-added-auto-mode-to-claude-code-ko/https://flowhat-blog.pages.dev/blog/2026-03-31-why-anthropic-added-auto-mode-to-claude-code-ko/Anthropic의 Auto Mode는 단순한 편의 기능이 아니다. 코딩 에이전트 경쟁이 모델 성능뿐 아니라 자율성 제어, 승인 설계, 하네스 품질로 이동하고 있다는 신호다.Tue, 31 Mar 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-03-31-why-anthropic-added-auto-mode-to-claude-code-zh/https://flowhat-blog.pages.dev/blog/2026-03-31-why-anthropic-added-auto-mode-to-claude-code-zh/Anthropic 的 Auto Mode 不只是一个便利功能。它说明编码代理的竞争，正在从模型能力扩展到自治控制、审批设计和 harness 质量。Tue, 31 Mar 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-14-users-say-claude-opus-got-worse/https://flowhat-blog.pages.dev/blog/2026-04-14-users-say-claude-opus-got-worse/Complaints about Claude Opus and Claude Code are turning into a bigger question about model consistency, version trust, and production reliability.Tue, 14 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-14-users-say-claude-opus-got-worse-ko/https://flowhat-blog.pages.dev/blog/2026-04-14-users-say-claude-opus-got-worse-ko/Claude Opus와 Claude Code에 대한 불만은 단순 성능 논란을 넘어 모델 일관성, 버전 신뢰, 프로덕션 안정성 문제로 번지고 있다.Tue, 14 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-14-users-say-claude-opus-got-worse-zh/https://flowhat-blog.pages.dev/blog/2026-04-14-users-say-claude-opus-got-worse-zh/围绕 Claude Opus 和 Claude Code 的抱怨，正在从单纯的性能争议，演变成关于模型一致性、版本信任和生产可靠性的问题。Tue, 14 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-16-openai-cloudflare-agent-cloud/https://flowhat-blog.pages.dev/blog/2026-04-16-openai-cloudflare-agent-cloud/OpenAI and Cloudflare's latest announcement is less about a simple partnership and more about where agent runtime infrastructure is heading.Thu, 16 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-16-openai-cloudflare-agent-cloud-ko/https://flowhat-blog.pages.dev/blog/2026-04-16-openai-cloudflare-agent-cloud-ko/OpenAI와 Cloudflare의 최근 발표는 단순한 파트너십 소식이라기보다, 에이전트 런타임 인프라가 어디로 가고 있는지를 보여준다.Thu, 16 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-16-openai-cloudflare-agent-cloud-zh/https://flowhat-blog.pages.dev/blog/2026-04-16-openai-cloudflare-agent-cloud-zh/OpenAI 和 Cloudflare 的最新宣布，与其说是普通合作，不如说是在说明 agent runtime infrastructure 正在往哪里走。Thu, 16 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-17-anthropic-opus-4-7-mythos-trust/https://flowhat-blog.pages.dev/blog/2026-04-17-anthropic-opus-4-7-mythos-trust/Claude Opus 4.7 arrives after recent trust concerns and alongside Mythos Preview, making it look less like a routine model bump and more like a strategic bridge.Fri, 17 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-17-anthropic-opus-4-7-mythos-trust-ko/https://flowhat-blog.pages.dev/blog/2026-04-17-anthropic-opus-4-7-mythos-trust-ko/Claude Opus 4.7은 최근의 신뢰 문제 이후 등장했고, 동시에 Mythos Preview와 함께 포지셔닝되면서 평범한 모델 업그레이드보다 더 전략적인 의미를 갖는다.Fri, 17 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-17-anthropic-opus-4-7-mythos-trust-zh/https://flowhat-blog.pages.dev/blog/2026-04-17-anthropic-opus-4-7-mythos-trust-zh/Claude Opus 4.7 出现在一次信任波动之后，又与 Mythos Preview 同时出现，因此它看起来不像普通的模型升级，而更像一次策略性的过渡。Fri, 17 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-18-codex-image-generation-and-the-broader-agent-shift/https://flowhat-blog.pages.dev/blog/2026-04-18-codex-image-generation-and-the-broader-agent-shift/OpenAI's latest Codex update suggests the product is expanding from a coding assistant into a broader agent surface that spans browser work, computer use, memory, automation, and visuals.Sat, 18 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-18-codex-image-generation-and-the-broader-agent-shift-ko/https://flowhat-blog.pages.dev/blog/2026-04-18-codex-image-generation-and-the-broader-agent-shift-ko/OpenAI의 최신 Codex 업데이트는 이 제품이 coding assistant를 넘어 browser work, computer use, memory, automation, visuals까지 포괄하는 더 넓은 agent surface로 확장되고 있음을 보여준다.Sat, 18 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-18-codex-image-generation-and-the-broader-agent-shift-zh/https://flowhat-blog.pages.dev/blog/2026-04-18-codex-image-generation-and-the-broader-agent-shift-zh/OpenAI 最新的 Codex 更新说明，这个产品正在从 coding assistant 扩展成一个更大的 agent surface，覆盖 browser work、computer use、memory、automation 和 visuals。Sat, 18 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-19-what-is-hermes-agent-and-how-it-differs-from-openclaw-claude-code-and-codex/https://flowhat-blog.pages.dev/blog/2026-04-19-what-is-hermes-agent-and-how-it-differs-from-openclaw-claude-code-and-codex/Hermes Agent is not just another coding copilot. It is positioning itself as a persistent, self-improving, multi-surface personal agent runtime — and that makes its tradeoffs different from OpenClaw, Claude Code, and Codex.Sun, 19 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-19-what-is-hermes-agent-and-how-it-differs-from-openclaw-claude-code-and-codex-ko/https://flowhat-blog.pages.dev/blog/2026-04-19-what-is-hermes-agent-and-how-it-differs-from-openclaw-claude-code-and-codex-ko/Hermes Agent는 단순한 coding copilot이 아니다. 이 프로젝트는 persistent하고 self-improving한 multi-surface personal agent runtime을 지향하며, 그 점이 OpenClaw, Claude Code, Codex와의 차이를 만든다.Sun, 19 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-19-what-is-hermes-agent-and-how-it-differs-from-openclaw-claude-code-and-codex-zh/https://flowhat-blog.pages.dev/blog/2026-04-19-what-is-hermes-agent-and-how-it-differs-from-openclaw-claude-code-and-codex-zh/Hermes Agent 不只是另一个 coding copilot。它把自己定位成一个 persistent、self-improving、multi-surface personal agent runtime，而这正是它与 OpenClaw、Claude Code、Codex 的差异所在。Sun, 19 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-20-claude-design-shows-where-ai-product-design-is-going-next/https://flowhat-blog.pages.dev/blog/2026-04-20-claude-design-shows-where-ai-product-design-is-going-next/Claude Design is less interesting as an AI design toy than as Anthropic's attempt to pull design exploration, prototyping, and implementation into one Claude-centered workflow.Mon, 20 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-20-claude-design-shows-where-ai-product-design-is-going-next-ko/https://flowhat-blog.pages.dev/blog/2026-04-20-claude-design-shows-where-ai-product-design-is-going-next-ko/Claude Design의 핵심은 AI 디자인 장난감이 아니라, Anthropic이 design exploration, prototyping, implementation을 하나의 Claude 중심 workflow로 묶으려는 시도라는 데 있다.Mon, 20 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-20-claude-design-shows-where-ai-product-design-is-going-next-zh/https://flowhat-blog.pages.dev/blog/2026-04-20-claude-design-shows-where-ai-product-design-is-going-next-zh/Claude Design 真正有意思的地方，不在于它是一个 AI 设计玩具，而在于 Anthropic 正试图把 design exploration、prototyping 和 implementation 拉进同一个以 Claude 为中心的 workflow。Mon, 20 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-22-chatgpt-images-2-0-is-about-more-than-better-pictures/https://flowhat-blog.pages.dev/blog/2026-04-22-chatgpt-images-2-0-is-about-more-than-better-pictures/OpenAI’s ChatGPT Images 2.0 is not just another image upgrade. It is another step in turning ChatGPT into a broader multimodal workspace where image creation becomes part of a larger workflow.Wed, 22 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-22-chatgpt-images-2-0-is-about-more-than-better-pictures-ko/https://flowhat-blog.pages.dev/blog/2026-04-22-chatgpt-images-2-0-is-about-more-than-better-pictures-ko/OpenAI의 ChatGPT Images 2.0은 단순한 이미지 업그레이드가 아니다. 이미지 생성이 더 큰 workflow의 일부가 되는 방향으로 ChatGPT를 넓히는 또 하나의 움직임이다.Wed, 22 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-22-chatgpt-images-2-0-is-about-more-than-better-pictures-zh/https://flowhat-blog.pages.dev/blog/2026-04-22-chatgpt-images-2-0-is-about-more-than-better-pictures-zh/OpenAI 的 ChatGPT Images 2.0 不只是一次图片升级。它更像是 ChatGPT 继续扩张的一步，让图像生成成为更大 workflow 的一部分。Wed, 22 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-23-what-is-claude-for-chrome-and-how-it-differs-from-claude-desktop/https://flowhat-blog.pages.dev/blog/2026-04-23-what-is-claude-for-chrome-and-how-it-differs-from-claude-desktop/Claude for Chrome is more than a browser extension. It is Anthropic’s attempt to turn Claude into a browser agent that can read pages, click through websites, and work alongside Claude Desktop.Thu, 23 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-23-what-is-claude-for-chrome-and-how-it-differs-from-claude-desktop-ko/https://flowhat-blog.pages.dev/blog/2026-04-23-what-is-claude-for-chrome-and-how-it-differs-from-claude-desktop-ko/Claude for Chrome은 단순한 브라우저 확장이 아니다. Anthropic이 Claude를 브라우저 안에서 읽고 클릭하고 움직이는 에이전트로 확장하려는 시도에 가깝다.Thu, 23 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-23-what-is-claude-for-chrome-and-how-it-differs-from-claude-desktop-zh/https://flowhat-blog.pages.dev/blog/2026-04-23-what-is-claude-for-chrome-and-how-it-differs-from-claude-desktop-zh/Claude for Chrome 不只是一个浏览器扩展。它更像是 Anthropic 把 Claude 推向浏览器代理的一步，让它能够在网页中读取、点击并执行动作。Thu, 23 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-25-what-gpt-5-5-changes-for-chatgpt-claude-and-gemini/https://flowhat-blog.pages.dev/blog/2026-04-25-what-gpt-5-5-changes-for-chatgpt-claude-and-gemini/GPT-5.5 is not just another performance upgrade. OpenAI is presenting it as a work model for coding, computer use, research, and long-running tasks, which makes the competitive picture with Claude Opus 4.7 and Gemini 3.1 Pro much clearer.Sat, 25 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-25-what-gpt-5-5-changes-for-chatgpt-claude-and-gemini-ko/https://flowhat-blog.pages.dev/blog/2026-04-25-what-gpt-5-5-changes-for-chatgpt-claude-and-gemini-ko/GPT-5.5는 단순한 성능 개선판이 아니다. OpenAI는 이번 모델을 코딩, 컴퓨터 사용, 리서치, 장기 실행 업무를 맡길 수 있는 작업 모델로 밀고 있다. 그래서 이번 출시는 Claude Opus 4.7, Gemini 3.1 Pro와의 경쟁 구도까지 같이 봐야 한다.Sat, 25 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-25-what-gpt-5-5-changes-for-chatgpt-claude-and-gemini-zh/https://flowhat-blog.pages.dev/blog/2026-04-25-what-gpt-5-5-changes-for-chatgpt-claude-and-gemini-zh/GPT-5.5 不只是一次常规性能升级。OpenAI 正在把它定义成一个可用于编码、电脑操作、研究和长时任务的工作模型，这也让它与 Claude Opus 4.7 和 Gemini 3.1 Pro 的竞争关系变得更清晰。Sat, 25 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-27-what-are-workspace-agents-in-chatgpt/https://flowhat-blog.pages.dev/blog/2026-04-27-what-are-workspace-agents-in-chatgpt/Workspace agents are not just another ChatGPT feature. They are OpenAI’s attempt to turn ChatGPT into a shared team workflow system built around triggers, tools, approvals, and cloud execution.Mon, 27 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-27-what-are-workspace-agents-in-chatgpt-ko/https://flowhat-blog.pages.dev/blog/2026-04-27-what-are-workspace-agents-in-chatgpt-ko/Workspace agents는 ChatGPT에 붙은 기능 하나가 아니다. OpenAI는 이 제품으로 ChatGPT를 팀이 공유해서 쓰는 워크플로우 에이전트 레이어로 밀고 있다.Mon, 27 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-27-what-are-workspace-agents-in-chatgpt-zh/https://flowhat-blog.pages.dev/blog/2026-04-27-what-are-workspace-agents-in-chatgpt-zh/Workspace agents 不是 ChatGPT 里又多了一个小功能。OpenAI 更像是在把 ChatGPT 推向一个面向团队、可共享、可治理的工作流代理层。Mon, 27 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-27-why-deepseek-v4-matters-more-for-agents-than-for-long-context/https://flowhat-blog.pages.dev/blog/2026-04-27-why-deepseek-v4-matters-more-for-agents-than-for-long-context/DeepSeek V4 is easy to describe as another 1M-context model. But the more important story is that DeepSeek is packaging V4 as an agent backend: cheaper long-running loops, Anthropic/OpenAI compatibility, tool-call rules, and direct integration into coding-agent tools.Mon, 27 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-27-why-deepseek-v4-matters-more-for-agents-than-for-long-context-ko/https://flowhat-blog.pages.dev/blog/2026-04-27-why-deepseek-v4-matters-more-for-agents-than-for-long-context-ko/DeepSeek V4는 1M 컨텍스트 모델로 설명하기 쉽다. 하지만 더 중요한 포인트는 DeepSeek가 V4를 에이전트 백엔드로 밀고 있다는 점이다. 더 싼 장기 실행 루프, Anthropic/OpenAI 호환, 툴 호출 규칙, 코딩 에이전트 도구 연동이 모두 그 방향을 가리킨다.Mon, 27 Apr 2026 00:00:00 GMThttps://flowhat-blog.pages.dev/blog/2026-04-27-why-deepseek-v4-matters-more-for-agents-than-for-long-context-zh/https://flowhat-blog.pages.dev/blog/2026-04-27-why-deepseek-v4-matters-more-for-agents-than-for-long-context-zh/把 DeepSeek V4 解释成一个 1M 上下文模型很容易，但更重要的点在于 DeepSeek 正在把 V4 打包成一个代理后端：更便宜的长循环、Anthropic/OpenAI 兼容、工具调用规则，以及对编码代理工具的直接接入。Mon, 27 Apr 2026 00:00:00 GMT