Lydia Hallie (@lydiahallie)

`context: fork` 옵션을 추가하면 스킬을 격리된 서브에이전트에서 실행할 수 있다는 개발 팁 공유. 메인 컨텍스트는 중간 도구 호출을 보지 못하고, 서브에이전트는 CLAUDE.md와 스킬을 프롬프트로 새 컨텍스트 창을 얻으며 `agent` 필드로 서브에이전트 유형을 지정할 수 있음. 에이전트 도구 체계 관련 유용한 개발 업데이트.

https://x.com/lydiahallie/status/2033603164398883042

#claude #agents #subagent #aitooling

Lydia Hallie ✨ (@lydiahallie) on X

Btw you can add `context: fork` to run a skill in an isolated subagent. The main context only sees the final result, not the intermediate tool calls It gets a fresh context window with CLAUDE.md + your skill as the prompt. The `agent` field even lets you set the subagent type!

X (formerly Twitter)

Elias Kempf (@elkmf)

새 모델 출시 후 변경된 LLM 동작을 체계적으로 찾아내기 위한 파이프라인을 구축·평가한 내용입니다. 저자들은 다양한 모델 차이 탐지(model diffing) 방법이 동일한 동작 변화를 발견하는 경우가 많지만, 그 변화에 대해 서로 다른 추상화 수준으로 기술한다는 점을 관찰했습니다. changelog에 없는 동작 변화 탐지에 초점을 둡니다.

https://x.com/elkmf/status/2023453592636846268

#llm #modeldiffing #modelevaluation #aitooling

Elias Kempf (@elkmf) on X

New model release? Great. But did the LLM’s behavior change in ways the changelog doesn't mention? We built and evaluated a pipeline to find out! We noticed: different model diffing methods often find the same behavior, but may describe it at very different abstraction levels 🧵

X (formerly Twitter)

Deedy (@deedydas)

세션에서 AI가 따라간 방대한 단계들을 공유하며, 타임스탬프 기준으로 약 2시간이지만 작성자가 자리를 비운 시간이 대부분이었다고 설명합니다. 메타 포인트로 세션 시각화를 만들기 위해 Claude에게 .jsonl CC 로그에서 명령어를 제거하도록 요청해 처리한 작업을 언급합니다. 즉, Claude를 이용한 로그 전처리 및 세션 시각화 사례를 보여줍니다.

https://x.com/deedydas/status/2011678342211387896

#claude #aitooling #logprocessing #sessionvisualization

Deedy (@deedydas) on X

Here's an overwhelming set of steps it followed in the session. Even though it says ~2hr by timestamps most of that is me disappearing from my computer. PS: this is meta but to generate this session visual, I asked Claude to strip the commands from the .jsonl CC log and then

X (formerly Twitter)
Top 7 MCP clients for AI tooling reviewed: Claude Desktop, Cursor AI, Claude Code, Windsurf, Cline, Continue & LibreChat. MCPs bridge AI models with real-world tools like databases & APIs. These clients make setup easy - often just adding a config line. #AI #MachineLearning #MCP #AITooling #TechReview #Coding #Development #ArtificialIntelligence https://www.kdnuggets.com/top-7-mcp-clients-for-ai-tooling
Top 7 MCP Clients for AI Tooling - KDnuggets

The most popular clients that seamlessly and reliably work with MCP servers ranging from IDEs to chatbots and plugins.

KDnuggets
AI tooling must be disclosed for contributions by mitchellh · Pull Request #8289 · ghostty-org/ghostty

I think, at this stage of AI, it is a common courtesy to disclose this. In a perfect world, AI assistance would produce equal or higher quality work than any human. That isn't the world we live...

GitHub
🚀 Just published: "AI-Augmented Development with Cursor, Claude, and Whispr"
After hundreds of hours with these tools, I've compiled my best practices.
-project setup for AI
-prompts
-voice
-mcp
& a lot more
Read here:
https://henry-schober.com/cursor
#DevProductivity #AITooling
AI-Augmented Development with Cursor, Claude, and Whispr | Henry Schober

Learn how to transform your development workflow with Cursor AI, voice dictation tools like MacWhisper and SuperWhisper, YOLO mode, and MCP extensions. A practical guide to becoming an AI assisted developer with cursor.

Henry Schober