I am testing #OpenCode with Regolo.ai (https://regolo.ai/private-ai-coding-deploy-without-giving-away-your-code/) on contributing to Amber-Lang (new programming language that transpile to bash).

I am using it since few days to complete the #code coverage of our codebase.
SInce today I am testing with https://github.com/code-yeongyu/oh-my-opencode and I see that with the same model is smarter (also now uses subagents).

Check my #Fosdem talk https://fosdem.org/2026/schedule/event/GGLZS9-amber-lang-bash-transpiler/

Ivan Fioravanti ᯅ (@ivanfioravanti)

MLX OpenCode와 Qwen3-Coder-Next-8bit(165K 컨텍스트)를 로컬에서 최대한 활용해보고 있다는 실험 리포트. 작성자는 이 조합이 동작하지만 더 큰 모델들(MiniMax, Zai, StepFun)에 비해 자주 수동 유도(steer)를 필요로 한다고 느낍니다.

https://x.com/ivanfioravanti/status/2020910223095992785

#qwen3 #opencode #contextwindow #codemodel

#Development #Comparisons
Comparison of 36 CLI coding agents · “Now there’s a CLI agent for every workflow.” https://ilo.im/16ahjo

_____
#AI #Agents #ClaudeCode #Codex #Gemini #OpenCode #CLI #WebDev #Frontend #Backend

🎉 Univention Nubus jetzt auf #openCode gelistet!

Gute Nachrichten für die öffentliche Verwaltung und alle, die auf offene, souveräne IT setzen.

Nubus ist unsere modulare Open-Source-Lösung für zentrales Identity & Access Management (#IAM) für die sichere & effiziente Verwaltung von Identitäten, Rollen und Berechtigungen.

Nubus lässt sich flexibel in bestehende IT-Landschaften einbinden – On-Premises, unter Kubernetes oder in hybriden Infrastrukturen.

👉 Zum Eintrag:
https://opencode.de/de/software/nubus-8053

@pmj Empfehle dir Cortecs https://cortecs.ai/ Das ist ein European LLM Router. Zugriff auf Models gehostet in Europa.

Bei anderen Routern wie OpenRouter oder Ollamacloud kommen Amerikanische Provider zum Einsatz.

#cortecs #OpenCode

ollama cloud ist geil
gleicher preis wie z.b. openai aber zugriff auf mehrere modell aber man kann es auch gratis nutzen
und die prompts und responses werden nicht fürs training genutzt

drawback: keine transparenz bei den limiten - ich habe mit qwen3-coder und opencode in 5 minuten mein gesamtes free limit verballert 😂
nanocoder ist da weitaus sparsamer, ist allerdings paar mal abgestürzt
#KI #ollama #nanocoder #opencode #qwen3coder

I used an LLM to help me build a simple static site generator while I took a shit. This was not possible some years ago. But this is not the news here. It's something else.

I have strong conceptual understanding of infotech but an extremely difficult relationship to actually learning programming due to neurodivergence.

#llm #code #ai #opencode (1/4)

Friday means tacos in #Sweden.

Cooking both on the stove and in #opencode 🤓

#TacoFredag #vibecooking #soundvibes

Learn how to combine OpenCode with AWS MCP servers for flexible, AI-powered architecture assistance without vendor lock-in.

Read the full guide: https://dev.to/gitaroktato/using-your-own-architecture-agent-with-opencode-and-aws-mcp-servers-2j26

#AWS #architecture #OpenCode #AI #AIAgents

Coding harness personality is an extremely subjective preference. I have used #ClaudeCode, #Cursor, #opencode, and #github #CopilotCli. I've found Github copilot cli the closest in personality to Claude Code.

I wanted to like opencode, but it seems too targeted towards people who like agents to run with minimal human intervention. Too much of an effort to craft an instruction to make it behave like claude code.

e.g. it didn't ask for approval for a git commit, and wrote a really terse message.

Awni Hannun (@awnihannun)

mlx_lm.server, Qwen3 Coder Next 6bit, OpenCode, M3 Ultra를 조합한 로컬 코딩 환경을 소개. 로컬 LLM 서버와 6비트 경량화 코드 모델, 코드 보조 툴, 고성능 칩을 결합해 매우 빠르고 실용적인 로컬 코드 작성/실행 세트업을 구현한 사례로 설명하고 있음.

https://x.com/awnihannun/status/2018852381677756872

#qwen #localllm #opencode #m3ultra

金のニワトリ (@gosrum)

opencode에서 발생하는 크래시 관련 이슈가 제기됨. 80k 이상의 컨텍스트 길이에서 높은 확률로 크래시가 발생한다고 보고되어 있으며, 해당 문제가 해결되지 않으면 정확한 성능 평가가 어려울 것이라고 지적함.

https://x.com/gosrum/status/2018906755288412257

#opencode #contextwindow #80k #crash

金のニワトリ (@gosrum)

Qwen3-Coder-Next 벤치마크 중 Claude Code와 상성이 맞지 않아 진행이 전혀 되지 않음. opencode와 조합하면 초반에는 괜찮지만, 중간에서 갑자기 llama.cpp 쪽에서 크래시가 발생하여 테스트가 중단되는 문제를 보고함.

https://x.com/gosrum/status/2018906324931838287

#qwen3 #claude #llama.cpp #opencode #benchmark

#OpenCode + #Kimi K2.5 (free today through OpenRouter).

Three agents are modernizing the frontend of the PointerPodcast website.

Let’s see what they ship.

逆瀬川 (@gyakuse)

OpenCode의 무료 모델들(Kimi-K2.5 Free, GLM Free 등)은 입력한 데이터가 모델 개선(학습)에 사용될 수 있으니 주의하라는 경고입니다. 특히 Kimi는 공식 API를 통해서도 학습 데이터로 활용된다고 명시되어 있어 개인정보·데이터 사용 관점에서 중요한 알림이며 관련 문서 링크가 첨부되어 있습니다.

https://x.com/gyakuse/status/2018250132115148945

#opencode #kimi #datausage #privacy #ai

dax (@thdxr)

작성자가 항상 켜져 있는(opencode) 서버를 구축해 어디서나 기기에서 세션을 실행할 수 있게 설정했다는 개발자 게시물입니다. 설정에 몇 분밖에 걸리지 않았고, 실제 동작(시연)을 보여준다고 밝히며 원격 개발·세션 공유·접근성 향상 사례로 참고할 만합니다.

https://x.com/thdxr/status/2017691649384620057

#opencode #devtools #remote #server

I did not try out OpenCode

A post about development thought process, containers, bash, and how it doesn't matter it doesn't even work.

https://pancake.bearblog.dev/2026-01-31-i-did-no-try-out-opencode/
#blog #ollama #opencode #podman

LM Studio and Claude Code

Which means you can use with gguf (and mlx) locally.

https://lmstudio.ai/blog/claudecode

I have #litellm setup which also sets the Anthropic-compatible /v1/messages endpoint, with any model.

Use #OpenCode Zen (supports BYOK now) or #Ollama Cloud if you need other models and don’t have the hardware to run them locally. Current favorite GLM 4.7, and for lighter tasks the fantastic lomo-3 on a local Ollama.

Have you tried Kimi K2.5? It's actually quite a big of a deal!

AI models are already at a level where they are able to severely transform the way we develop software.

With new #opensource models coming out that are able to close the gap to ChatGPT5.2 & Claude Opus 4.5 it's also clear that you are not tied to any single vendor. AI has become a commodity. You can run smaller models locally or your company can host bigger models in-house.

You will stumble in the beginning - thinking it's creating garbage.

But - shocker - #vibecoding is ALSO something you have to get experienced at. When you know how to work, you'll be able to do the same work with much cheaper models.

#opencode offers Kimi K2.5 for free during a period - try it out for yourself!

#kimi #opensourceAi

金のニワトリ (@gosrum)

Kimi-K2.5가 어떤 에이전트와 조합해도 문제가 없는지 확인하기 위해 kimi-cli, opencode 외에 Claude Code와 조합한 측정도 수행했으며, 추가로 Claude Code(v2.1.25)와 Opus 4.5 / Sonnet 4.5 조합 결과를 덧붙였다는 안내입니다.

https://x.com/gosrum/status/2017241531145691445

#kimik2.5 #agents #claudecode #opencode #modelevaluation