うーんQwenCode素晴らしいね。

いや、QwenCodeは明らかにOpenCodeより賢いな。 これは凄いな・・・ サブエージェントだけではなく、コンテキストとかメモリ管理の差なのだろうか

QwenCode、同じLocalLLMでも確かにOpenCodeよりだいぶ賢い感じがあるな。クライアントでここまで変わるものなのだろうか。もうちょっと使ってみないとまだわからんけど。

QwenCode、ブツブツ言ってて面白いｗ

まあOSSで無償で提供している以上はオプトアウトにしときたいのだろうし悪意はなさそうだが、もうちょっとわかりやすくしておいたほうが良いかも。QwenCode事態はぱっとみかなり良さそう。

Alibaba主体のQwenCode、一応開発母体が母体だけにリポジトリをチェックアウトし他社のフロンティアモデルAIで不正な処理等がないか確認。結果として危険な情報送信や抜き取りは無いようだ。 ただし統計情報の送信についてはオプトアウト、別に隠蔽していないので悪意のあるものではないと思われる。

ClaudeCode、Codex、Antigravity CLI、QwenCode CLIの4つを主軸に色々と回していこうかと。 Claude Opus 4.8 GPT-5.5 Gemini 3.5 Qwen 3.6 Plus それぞれ検証よね。ローカルLLMはQwen3.6、次点でgpt-ossで回すが。

🤖Qwen Code is surprisingly capable for coding tasks. But like most LLM agents, it still struggles when the task requires real-time web access. A simple MCP setup can give it: → Live web browsing → JS-rendered pages → Search capabilities → Browser actions We documented one implementation using Scrapeless MCP. Feels like MCP is becoming the USB-C of AI agents. 👉 A complete guide: scrapeless.com/en/blog/qwen-… 🎁 Trial credits are available for new users, up to 100 free hours session run. #AIAgents #MCP #QwenCode #LLM #AgenticAI #Qwen

opencode-agmsg ですが名前変えました → agmsg-opencode-plugin。前のリポジトリはredirectされます。理由はこれから他のプラグイン作るときに並ぶから。ということで、 agmsg-qwencode-plugin 作ってる #agmsg

Yes you are right about the unfamiliar apply_patch tool. Besides codex relies on bash only while other harnesses provide shortcuts like grep/glob/edit tools. Pi > QwenCode after training remains a mystery to me as well atm. Might be that simpler system prompt helps RL performance.

PROOF OF WORK Bitcoinize Node AGI Cluster Report 9 hours to sync 400k blocks on minimum config: 5 peers, 4GB RAM, 2 cores. Not bad, but there are small bugs everywhere, and that is exactly why we test. My home network is becoming a cluster of Bitcoin nodes running different implementations, built to survive different hardware, edge cases, and connectivity scenarios, while moving toward compatibility with @start9labs, @mynodebtc, @umbrel, and @RaspiBlitz. The Bitcoinize squads report back to the Main AGI agent, which allocates energy, tokens, and compute to future tasks according to real availability. No rush. Low budget, high sovereignty, local Llama always active, orchestrating Codex, Claude, OpenCode, Gemini, QwenCode, and OpenClaude toward one goal: turning node testing into knowledge mining. A pleb AGI learning how to become a Noble Hodler.

🤔你是不是在暗示QwenCode

Qwen Code 上线了，阿里的 AI 编程助手。跟 Cursor/Claude Code 类似，但核心优势是中文语境理解——写中文注释、处理中文文档更顺手。 建议关注两点：MCP 协议支持程度、能不能接本地模型（BYO LLM）。这两块决定它能走多远。 #AI #QwenCode

QwenCode、ClaudeCodeのリフレッシュ関連がTeraTermと相性悪いからWindowsのターミナルに乗り換え

Use QwenCode to join DojoZero sport forecasting arena. You can ask it to memorize working strategies and apply them again in future games. See it in action 👇

Este es el desglose de top 3 agentes/IDEs/clientes mas usados en NaN en los últimos 7 días: - OpenCode: 41% - OpenAI Python/Go/JS SDK: 39% - Node.js: 8.5% Luego hay también gente con: - Zed Editor: 0.7% - QwenCode CLI: 0.5 Y este me ha sorprendido: - GuzzleHttp (PHP): ¡4.2%!🔥

QwenCode and Kimi both work through ACP and other agents like OpenCode can make use of them as well. Or am I missing something?

but minimax and zhipu are growth "stocks" they picked the more compelling customer segment, the vibe-coders and grannies raising lobsters. Alibaba had a lot of qwencode freeloaders and even though I personally love mimo's new model, it's a little too rich for me to afford every day

Since they are benchmarking on their own 'Qwencode' harness, I wonder how much of the improvement is the model vs. them improving the harness