diff --git a/public/img/sim/action1.mp4 b/public/img/sim/action1.mp4
new file mode 100644
index 0000000..263f8f5
Binary files /dev/null and b/public/img/sim/action1.mp4 differ
diff --git a/public/img/sim/action2.mp4 b/public/img/sim/action2.mp4
new file mode 100644
index 0000000..e18c251
Binary files /dev/null and b/public/img/sim/action2.mp4 differ
diff --git a/public/img/sim/image-20260103135227003.png b/public/img/sim/image-20260103135227003.png
new file mode 100644
index 0000000..27f8069
Binary files /dev/null and b/public/img/sim/image-20260103135227003.png differ
diff --git a/public/img/sim/image-20260103135535936.png b/public/img/sim/image-20260103135535936.png
new file mode 100644
index 0000000..c3119c5
Binary files /dev/null and b/public/img/sim/image-20260103135535936.png differ
diff --git a/public/img/sim/image-20260103135553368.png b/public/img/sim/image-20260103135553368.png
new file mode 100644
index 0000000..3548b8e
Binary files /dev/null and b/public/img/sim/image-20260103135553368.png differ
diff --git a/public/img/sim/image-20260103143108173.png b/public/img/sim/image-20260103143108173.png
new file mode 100644
index 0000000..4244b72
Binary files /dev/null and b/public/img/sim/image-20260103143108173.png differ
diff --git a/public/img/sim/image-20260103143506983.png b/public/img/sim/image-20260103143506983.png
new file mode 100644
index 0000000..75b5a89
Binary files /dev/null and b/public/img/sim/image-20260103143506983.png differ
diff --git a/public/img/sim/image-20260103143618886.png b/public/img/sim/image-20260103143618886.png
new file mode 100644
index 0000000..42e4eba
Binary files /dev/null and b/public/img/sim/image-20260103143618886.png differ
diff --git a/public/img/sim/image-20260103144629767.png b/public/img/sim/image-20260103144629767.png
new file mode 100644
index 0000000..1643558
Binary files /dev/null and b/public/img/sim/image-20260103144629767.png differ
diff --git a/public/img/sim/image-20260103144757557.png b/public/img/sim/image-20260103144757557.png
new file mode 100644
index 0000000..ce9b940
Binary files /dev/null and b/public/img/sim/image-20260103144757557.png differ
diff --git a/public/img/sim/image-20260103151610221.png b/public/img/sim/image-20260103151610221.png
new file mode 100644
index 0000000..749fd5f
Binary files /dev/null and b/public/img/sim/image-20260103151610221.png differ
diff --git a/public/img/sim/image-20260103220522907.png b/public/img/sim/image-20260103220522907.png
new file mode 100644
index 0000000..56df33e
Binary files /dev/null and b/public/img/sim/image-20260103220522907.png differ
diff --git a/public/img/sim/image-20260103220724061.png b/public/img/sim/image-20260103220724061.png
new file mode 100644
index 0000000..cc92339
Binary files /dev/null and b/public/img/sim/image-20260103220724061.png differ
diff --git a/public/img/sim/image-20260103221515111.png b/public/img/sim/image-20260103221515111.png
new file mode 100644
index 0000000..beed0ce
Binary files /dev/null and b/public/img/sim/image-20260103221515111.png differ
diff --git a/public/img/sim/image-20260103225328920.png b/public/img/sim/image-20260103225328920.png
new file mode 100644
index 0000000..b548c2b
Binary files /dev/null and b/public/img/sim/image-20260103225328920.png differ
diff --git a/public/img/sim/image-20260103230746221.png b/public/img/sim/image-20260103230746221.png
new file mode 100644
index 0000000..1eabef3
Binary files /dev/null and b/public/img/sim/image-20260103230746221.png differ
diff --git a/public/img/sim/rag1.png b/public/img/sim/rag1.png
new file mode 100644
index 0000000..870b020
Binary files /dev/null and b/public/img/sim/rag1.png differ
diff --git a/public/img/sim/rag2.png b/public/img/sim/rag2.png
new file mode 100644
index 0000000..e617e7e
Binary files /dev/null and b/public/img/sim/rag2.png differ
diff --git a/public/img/sim/test_img.png b/public/img/sim/test_img.png
new file mode 100644
index 0000000..21534d1
Binary files /dev/null and b/public/img/sim/test_img.png differ
diff --git a/public/img/sim/test_img2.png b/public/img/sim/test_img2.png
new file mode 100644
index 0000000..a310161
Binary files /dev/null and b/public/img/sim/test_img2.png differ
diff --git a/public/img/sim/toolbar.png b/public/img/sim/toolbar.png
new file mode 100644
index 0000000..983275f
Binary files /dev/null and b/public/img/sim/toolbar.png differ
diff --git a/scripts/convert-gif.sh b/scripts/convert-gif.sh
index c39f597..db0e6e8 100755
--- a/scripts/convert-gif.sh
+++ b/scripts/convert-gif.sh
@@ -1,19 +1,40 @@
 #!/bin/bash
-set -e
 
-echo "Converting GIFs to WebM..."
-find public/img -name '*.gif' -print0 | while IFS= read -r -d '' gif; do
-  webm="${gif%.gif}.webm"
-  echo "  $gif -> $webm"
-  ffmpeg -y -i "$gif" -c:v libvpx-vp9 -crf 30 -b:v 0 "$webm" 2>/dev/null
-done
+MAX_DURATION=10
+
+echo "Converting GIFs to MP4..."
+while IFS= read -r -d '' gif; do
+  mp4="${gif%.gif}.mp4"
+  
+  duration=$(ffprobe -v error -select_streams v:0 -show_entries format=duration -of csv=p=0 "$gif" 2>/dev/null || echo "0")
+  
+  if (( $(echo "$duration > $MAX_DURATION" | bc -l 2>/dev/null || echo 0) )); then
+    speed=$(echo "scale=2; $duration / $MAX_DURATION" | bc -l)
+    vf="setpts=PTS/${speed},scale='min(1280,iw)':-2,scale=trunc(iw/2)*2:trunc(ih/2)*2"
+    echo "  $gif -> $mp4 (${duration}s -> ${MAX_DURATION}s, ${speed}x)"
+  else
+    vf="scale='min(1280,iw)':-2,scale=trunc(iw/2)*2:trunc(ih/2)*2"
+    echo "  $gif -> $mp4"
+  fi
+  
+  if ffmpeg -nostdin -y -i "$gif" \
+    -movflags faststart \
+    -pix_fmt yuv420p \
+    -vf "$vf" \
+    -c:v libx264 \
+    -crf 23 \
+    -preset ultrafast \
+    "$mp4"; then
+    rm "$gif"
+    echo "  Done: $mp4"
+  else
+    echo "  FAILED: $gif (keeping original)"
+  fi
+done < <(find public/img -name '*.gif' -print0)
 
 echo "Updating markdown files..."
 find src/content -name '*.md' -exec perl -i -pe '
-  s/!\[([^\]]*)\]\(([^)]*?)\.gif\)/<video autoplay loop muted playsinline style="max-width:100%">\n  <source src="$2.webm" type="video\/webm">\n<\/video>/g
+  s/!\[([^\]]*)\]\(([^)]*?)\.gif\)/<video autoplay loop muted playsinline style="max-width:100%">\n  <source src="$2.mp4" type="video\/mp4">\n<\/video>/g
 ' {} \;
 
-echo "Removing original GIFs..."
-find public/img -name '*.gif' -delete
-
 echo "Done!"
diff --git a/src/content/blog-en/sim.md b/src/content/blog-en/sim.md
new file mode 100644
index 0000000..6657c58
--- /dev/null
+++ b/src/content/blog-en/sim.md
@@ -0,0 +1,225 @@
+---
+description: sim is an open-source no-code agent tool that uses natural language to
+  generate LLM workflows. Supports self-hosted and managed, but relies on Copilot.
+pubDate: '2026-01-03'
+tags:
+- LLM
+- Agent
+- No-code
+- Workflow
+- Automation
+- RAG
+title: 'sim: Open-Source No-Code Agent Tool for Natural Language Workflows'
+---
+
+## Summary
+
+In the no-code/low-code workflow automation market, n8n and Make have effectively become the standard.  
+In that landscape, an open-source project called sim has emerged, positioning itself around LLM-generated workflows created from natural language prompts.
+
+[sim](https://www.sim.ai/) officially supports Self-hosted deployments as well as Cloud-hosted (Managed), and even provides both Helm Charts and docker-compose.  
+However, its core feature—Copilot (LLM-based workflow generation)—has a structural limitation: even in Self-hosted environments, it requires a paid API Key.
+
+---
+
+## TL;DR
+
+- sim is an open-source no-code agent tool that generates workflows from natural language input.  
+- It supports both Cloud-hosted and Self-hosted deployments, with relatively strong operational convenience.  
+- It provides RAG (Knowledge), MCP, and a wide range of app integrations.  
+- Its core feature, Copilot, requires paid billing even in Self-hosted environments.  
+- Without Copilot, much of sim’s advantage disappears.
+
+---
+
+<video autoplay loop muted playsinline style="max-width:100%">
+  <source src="../../../public/img/sim/action1.mp4" type="video/mp4">
+</video>
+
+
+
+
+
+## What is sim?
+
+These days, the no-code agent builder market is dominated by n8n and Make.  
+sim is an open-source project in that market that puts natural-language-based automatic workflow generation front and center.
+
+When a user enters requirements in natural language, the LLM interprets them and builds a workflow.  
+This approach also reminds me of Google’s Opal, which is still in beta.
+
+sim is available as open source, and it also offers a Managed Cloud-hosted service.
+
+---
+
+## Pricing Model and How LLMs Are Used
+
+On Cloud-hosted, the free plan includes $20 in credits.
+
+![image-20260103230746221](../../../public/img/sim/image-20260103230746221.png)
+
+The LLM used within workflows can be configured in two ways. (Copilot provided by sim does not fall into this category.)
+
+- Use your own LLM API Key (openai, claude, etc.)
+- Use a platform key provided by sim
+
+If you use the platform key instead of your own API Key, the cost increases by about 2x. Judging by the warning message shown when you delete a personal API Key, it feels like even sim recommends using your own LLM API Key for workflows.
+
+![image-20260103144757557](../../../public/img/sim/image-20260103144757557.png)
+![image-20260103144629767](../../../public/img/sim/image-20260103144629767.png)
+
+---
+
+
+
+## Self-hosted Support Is a Clear Advantage
+
+sim officially supports Self-hosted environments in addition to Cloud-hosted.
+
+Because it provides both docker-compose and Helm Charts,  
+it’s relatively straightforward to operate even in Kubernetes environments.
+
+From an operations perspective, this is a major advantage for sim. However, as I’ll explain later, there’s a limitation: because of the Copilot API Key requirement, costs and constraints still exist even in Self-hosted deployments.
+
+
+
+## How It Works (Basic Flow)
+
+sim’s basic flow works like this:
+
+1. Configure the LLM model you want to use.  
+2. Enter the task you want in natural language.  
+3. The LLM interprets it and automatically generates a workflow.
+
+![image-20260103225328920](../../../public/img/sim/image-20260103225328920.png)
+
+The generated workflow isn’t just simple node connections—  
+it can also include Code execution steps as needed.
+
+You can run it immediately using the RUN button at the top of the UI.
+
+![test_img](../../../public/img/sim/test_img.png)
+![test_img2](../../../public/img/sim/test_img2.png)
+
+---
+
+It’s surprising that you can build a workflow—including an Agent—this easily.
+
+
+
+## Deploy and Trigger
+
+Workflows can be deployed, and deployed workflows can be executed via API calls.
+
+Supported triggers include:
+
+- Start (API or manual execution)
+- Schedule
+- Webhook
+- RSS Feed
+
+Multiple triggers can be configured, and execution priority is summarized in [Official Docs - Trigger](https://docs.sim.ai/triggers).
+
+---
+
+
+
+## Toolbar and Integration
+
+sim supports not only LLM-based auto-generation, but also manual configuration.
+
+The Toolbar provides a large number of Tools and Integrations,  
+and it’s already connected to many external services. It looks like it includes pretty much anything you’d want to build.
+
+![toolbar](../../../public/img/sim/toolbar.png)
+
+---
+
+## Logging System
+
+The execution logging system is also well built. It tracks run success/failure, trigger type, and even detailed error messages—overall, the logging is quite solid.
+
+![image-20260103135535936](../../../public/img/sim/image-20260103135535936.png)
+![image-20260103151610221](../../../public/img/sim/image-20260103151610221.png)
+
+It’s at a level that’s fully usable in production operations.
+
+---
+
+
+
+## Template
+
+You can import pre-built workflows in the form of Templates.
+
+This is practical because you don’t have to generate everything via LLM from scratch.
+
+![image-20260103135553368](../../../public/img/sim/image-20260103135553368.png)
+
+---
+
+## Knowledge: RAG Setup
+
+sim supports Vector Store-based RAG via its Knowledge feature.
+
+I tested it by uploading the Transformer paper, and chunking and vectorization completed quickly.
+
+![rag1](../../../public/img/sim/rag1.png)
+![rag2](../../../public/img/sim/rag2.png)
+
+It also supports OCR, so it can process image-based documents as well. The fact that you can set up a RAG system this easily is compelling. I didn’t measure performance, so I’m not sure. Depending on the documents, I suspect you’d need to tune it yourself.
+
+---
+
+
+
+## Fix in Copilot
+
+The debugging experience is pretty impressive.
+
+Even when it’s hard to find what needs to be fixed—a common issue with no-code tools—the Fix in Copilot feature resolves problems automatically. You click a button, and the LLM debugs and repairs the workflow on its own. I find this part especially compelling.
+
+<video autoplay loop muted playsinline style="max-width:100%">
+  <source src="../../../public/img/sim/action2.mp4" type="video/mp4">
+</video>
+
+---
+
+## MCP Support
+
+Basic MCPs are already registered,  
+and if needed, you can add MCPs yourself.
+
+![image-20260103143506983](../../../public/img/sim/image-20260103143506983.png)
+
+---
+
+
+
+## Drawback: Copilot API Key Dependency
+
+Unfortunately, I also want to mention a critical drawback. This won’t matter to people using Cloud-hosted.
+
+But in Self-hosted environments, the biggest drawback isn’t Copilot itself—it’s the structure that *requires* a Copilot API Key.
+
+To use sim’s Copilot, even in Self-hosted deployments, you must use an API Key issued by sim.
+
+For someone like me who wants to build an Agentic AI system using Self-hosted open source plus my own LLM API Key, I don’t like the fact that another SaaS tool gets inserted.  
+This structure becomes a clear constraint for engineers who care about controlling LLM costs and maintaining data sovereignty in Self-hosted environments.
+
+![image-20260103220522907](../../../public/img/sim/image-20260103220522907.png)
+![image-20260103220724061](../../../public/img/sim/image-20260103220724061.png)
+![image-20260103221515111](../../../public/img/sim/image-20260103221515111.png)
+
+Copilot pricing is set higher than using the base LLM directly. This appears to be sim’s core business model.
+
+---
+
+## Wrap-up
+
+sim is a technically polished open-source no-code agent workflow tool.
+
+Self-hosted support, RAG, MCP, and broad integrations are clear strengths.  
+However, it’s disappointing that the core feature—Copilot—is tied to a commercial service.
+
+If you plan to use Cloud-hosted or use sim’s Copilot key in a Self-hosted setup, it’s an attractive option. If not, it seems best to compare it alongside other alternatives.
\ No newline at end of file
diff --git a/src/content/blog/sim.md b/src/content/blog/sim.md
new file mode 100644
index 0000000..2f04eb4
--- /dev/null
+++ b/src/content/blog/sim.md
@@ -0,0 +1,226 @@
+---
+title: "sim: 자연어로 워크플로우를 생성하는 오픈소스 노코드 에이전트 툴"
+description: "sim은 자연어 입력만으로 LLM이 워크플로우를 생성하는 오픈소스 노코드 에이전트 툴이다. Self-hosted와 Managed를 모두 지원하지만 Copilot 의존 구조라는 한계도 존재한다."
+pubDate: "2026-01-03"
+tags:
+  - LLM
+  - Agent
+  - No-code
+  - Workflow
+  - Automation
+  - RAG
+---
+
+## 요약
+
+노코드·로우코드 워크플로우 자동화 시장은 n8n과 Make가 사실상 표준처럼 자리 잡고 있다.  
+이런 흐름 속에서 자연어 입력만으로 LLM이 워크플로우를 생성해주는 오픈소스 프로젝트 sim이 등장했다.
+
+[sim](https://www.sim.ai/)은 Cloud-hosted(Managed) 방식뿐 아니라 Self-hosted 환경도 공식 지원하며, Helm Chart와 docker-compose까지 제공한다.  
+다만 핵심 기능인 Copilot(LLM 기반 워크플로우 생성)은 Self-hosted 환경에서도 유료 API Key가 필요하다는 구조적 한계를 가진다.
+
+---
+
+## TL;DR
+
+- sim은 자연어 입력으로 워크플로우를 생성하는 오픈소스 노코드 에이전트 툴이다.  
+- Cloud-hosted와 Self-hosted를 모두 지원하며 운영 편의성은 높은 편이다.  
+- RAG(Knowledge), MCP, 다양한 앱 연동 기능을 제공한다.  
+- 핵심 기능인 Copilot은 Self-hosted 환경에서도 유료 결제가 필요하다.  
+- Copilot 없이 사용하면 sim의 장점이 크게 줄어든다.
+
+---
+
+<video autoplay loop muted playsinline style="max-width:100%">
+  <source src="../../../public/img/sim/action1.mp4" type="video/mp4">
+</video>
+
+
+
+
+
+## sim은 무엇인가
+
+요즘 노코드 에이전트 구축 툴 시장은 n8n과 Make가 지배하고 있다.  
+sim은 이런 시장에서 자연어 기반 워크플로우 자동 생성을 전면에 내세운 오픈소스 프로젝트다.
+
+사용자가 자연어로 요구사항을 입력하면, LLM이 이를 해석해 워크플로우를 구성해준다.  
+이 방식은 아직 beta버전인 Google의 Opal을 떠올리게 한다.
+
+sim은 오픈소스로 공개되어 있으며, 동시에 Managed 형태의 Cloud-hosted 서비스도 함께 제공한다.
+
+---
+
+## 요금 구조와 LLM 사용 방식
+
+Cloud-hosted 기준으로 무료 플랜은 20달러 크레딧이 제공된다.
+
+![image-20260103230746221](../../../public/img/sim/image-20260103230746221.png)
+
+
+
+워크플로우에서 사용하는 LLM은 다음 두 가지 방식으로 사용할 수 있다. ( sim에서 제공하는 Copilot은 여기에 속하지 않는다. )
+
+- 개인 LLM(openai, claude 등) API Key 사용
+- sim에서 제공하는 플랫폼 키 사용
+
+개인 API Key를 사용하지 않고 플랫폼 키를 사용할 경우, 비용은 약 2배로 증가한다. 개인 API Key를 삭제할 때, 경고 문구가 나오는거 보니, sim에서도 워크플로우는 개인 LLM API Key를 사용하기를 권장하는 것 처럼 느껴진다.
+
+![image-20260103144757557](../../../public/img/sim/image-20260103144757557.png)
+![image-20260103144629767](../../../public/img/sim/image-20260103144629767.png)
+
+---
+
+
+
+## Self-hosted 지원은 분명한 장점이다
+
+sim은 Cloud-hosted뿐 아니라 Self-hosted 환경도 공식적으로 지원한다.
+
+docker-compose와 Helm Chart를 모두 제공하기 때문에,  
+Kubernetes 환경에서도 비교적 수월하게 운영할 수 있다.
+
+운영 관점에서 보면 이는 sim의 큰 장점이다. 다만 이후에 설명하겠지만, Copilot API Key를 요구하는 구조로 인해 Self-hosted 환경에서도 비용과 제약이 발생하는 한계가 존재한다.
+
+
+
+## 기본 동작 방식
+
+sim의 기본 동작 흐름은 다음과 같다.
+
+1. 사용할 LLM 모델을 설정한다.  
+2. 자연어로 원하는 작업을 입력한다.  
+3. LLM이 이를 해석해 워크플로우를 자동 생성한다.
+
+![image-20260103225328920](../../../public/img/sim/image-20260103225328920.png)
+
+생성된 워크플로우는 단순 노드 연결이 아니라,  
+필요에 따라 Code 실행 단계를 포함한다.
+
+UI 상단의 RUN 버튼으로 즉시 실행할 수 있다.
+
+![test_img](../../../public/img/sim/test_img.png)
+![test_img2](../../../public/img/sim/test_img2.png)
+
+---
+
+이렇게 간단하게 Agent를 포함한 워크플로우를 구축할 수 있다니 놀랍다.
+
+
+
+## Deploy와 Trigger
+
+워크플로우는 Deploy가 가능하며, 배포된 워크플로우는 API 호출로 실행할 수 있다.
+
+지원되는 Trigger는 다음과 같다.
+
+- Start (API 또는 수동 실행)
+- Schedule
+- Webhook
+- RSS Feed
+
+복수의 Trigger 설정도 가능하며, 실행 우선순위는 [공식 Docs - Trigger](https://docs.sim.ai/triggers)에 정리되어 있다.
+
+---
+
+
+
+## Toolbar와 Integration
+
+sim은 LLM 기반 자동 생성뿐 아니라 수동 구성도 지원한다.
+
+Toolbar에는 수많은 Tool과 Integration이 제공되며,  
+이미 다양한 외부 서비스와 연동되어 있다. 아마, 구축하고자 하는 웬만한건 다 포함되어 있을 것 처럼 보인다.
+
+![toolbar](../../../public/img/sim/toolbar.png)
+
+---
+
+## Log 시스템
+
+실행 로그 시스템도 잘 구성되어 있다. Run 성공/실패 여부, Trigger 종류, Error 상세 메시지까지 꽤나 로그 시스템이 잘 구축되어 있다.
+
+![image-20260103135535936](../../../public/img/sim/image-20260103135535936.png)
+![image-20260103151610221](../../../public/img/sim/image-20260103151610221.png)
+
+운영 환경에서도 충분히 활용 가능한 수준이다.
+
+---
+
+
+
+## Template
+
+이미 구축된 워크플로우를 Template 형태로 가져올 수 있다.
+
+처음부터 LLM으로 생성하지 않아도 되기 때문에 실용적이다.
+
+![image-20260103135553368](../../../public/img/sim/image-20260103135553368.png)
+
+---
+
+## Knowledge: RAG 구성
+
+sim은 Knowledge 기능을 통해 Vector Store 기반 RAG 구성을 지원한다.
+
+Transformer 논문을 업로드해 테스트했는데, Chunking과 Vector화가 빠르게 완료되었다.
+
+![rag1](../../../public/img/sim/rag1.png)
+![rag2](../../../public/img/sim/rag2.png)
+
+OCR도 지원해 이미지 문서 처리도 가능하다고 한다. RAG 시스템 구축을 이렇게 간단하게 할 수 있다는 점은 매력적으로 다가온다. 성능에 대해서는 측정해보지 못해서, 잘 모르겠다. 아마 문서에 따라 개인적으로 튜닝해야하지 않을까 싶다.
+
+---
+
+
+
+## Fix in Copilot
+
+디버깅 경험은 꽤 인상적이다.
+
+노코드 툴 특성상 수정 포인트를 찾기 어려운 상황에서도 Fix in Copilot 기능이 자동으로 문제를 해결해준다. 버튼을 누르고 LLM이 알아서 디버깅해서 워크플로우를 고친다. 이 부분은 참 매력적이라고 생각한다.
+
+<video autoplay loop muted playsinline style="max-width:100%">
+  <source src="../../../public/img/sim/action2.mp4" type="video/mp4">
+</video>
+
+---
+
+## MCP 지원
+
+기본적인 MCP는 이미 등록되어 있으며,  
+필요한 경우 직접 MCP를 추가할 수 있다.
+
+![image-20260103143506983](../../../public/img/sim/image-20260103143506983.png)
+
+---
+
+
+
+## 단점: Copilot API Key 의존 구조
+
+아쉽지만, 치명적인 단점에 대해서도 언급하고 싶다. Cloud-hosted를 사용하는 사람에게는 상관없는 내용이다.
+
+하지만, Self-hosted 환경에서 가장 치명적인 단점은 Copilot 기능 자체가 아니라, Copilot API Key를 반드시 요구하는 구조다.
+
+sim에서 제공하는 Copilot을 사용하려면 Self-hosted 환경에서도  반드시 sim에서 발급한 API Key가 필요하다.
+
+나같이, Self-hosted 오픈소스 + 개인 LLM API Key로 Agentic AI 시스템을 구축하겠다는 사람에게는, 다른 SaaS툴이 낀다는 부분이 마음에 들지 않는다.
+이런 구조는 Self-hosted 환경에서 LLM 비용 통제와 데이터 주권을 중요하게 생각하는 엔지니어에게는 명확한 제약으로 작용한다.
+
+![image-20260103220522907](../../../public/img/sim/image-20260103220522907.png)
+![image-20260103220724061](../../../public/img/sim/image-20260103220724061.png)
+![image-20260103221515111](../../../public/img/sim/image-20260103221515111.png)
+
+Copilot 비용은 기본 LLM을 직접 사용하는 것보다 비싸게 책정된다. 이는 sim의 핵심 비즈니스 모델로 보인다.
+
+---
+
+## 정리
+
+sim은 기술적으로 완성도가 높은 오픈소스 노코드 에이전트 워크플로우 툴이다.
+
+Self-hosted 지원, RAG, MCP, 다양한 Integration은 분명한 강점이다.  
+그러나 핵심 기능인 Copilot이 상용 서비스에 종속되어 있다는 점은 아쉬운 부분이다.
+
+Cloud hosted 혹은, Self-hosted에서 sim의 Copilot 키를 사용할 계획이라면 충분히 매력적인 선택지다. 그렇지 않다면 다른 대안과 함께 비교하는 것이 좋아 보인다.
diff --git a/src/content/drafts/sim.md b/src/content/drafts/gpu-sharing.md
similarity index 100%
rename from src/content/drafts/sim.md
rename to src/content/drafts/gpu-sharing.md
diff --git a/src/content/drafts/template.md b/src/content/drafts/template.md
index 527aa7e..757552c 100644
--- a/src/content/drafts/template.md
+++ b/src/content/drafts/template.md
@@ -1,25 +1,18 @@
 ---
-description: 설명
+# description: 검색 결과 스니펫 (120-160자)
+# title: 검색 노출 핵심 (50-60자), 키워드 포함
+# tags: 주요 키워드 + 상위 카테고리 + 관련 기술
+description: ''
 pubDate: '2025-12-19'
 tags:
-- LLMOps
-- MCP
-- Agent
-title: template
+  - 
+title: ''
 ---
 
-## TL;DR
-
-> 요약
-
-Intro
+## 요약
 
-## 소제목
+<!-- 본문 첫 단락: 검색엔진이 중요시함. 핵심 내용 + 키워드 자연스럽게 -->
 
-## 장점
-
-## 단점
-
-## 마무리
+## TL;DR
 
-## 참고 링크
+<!-- 3-4문장 핵심 결론. 이것만 읽어도 글의 가치 파악 가능하게 -->