dmux just shipped support for 11 coding agents including A/B testing for any/all of them. Now supports: Claude, Cline, Qwen, OpenCode, OpenAI, Amp, Charm, Cursor, GitHub copilot, Gemini CLI, pi Also new ui & merge diffs ➡️

@cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli dmux has a discord server now too!

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli the a/b testing across agents is smart. i've been manually switching between claude code and cursor for different tasks but having actual data on which one performs better for what would save so much guesswork

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli A/B testing across 11 agents is huge. The merge diffs feature alone makes this worth trying - comparing agent outputs side by side is the killer workflow everyone needs.

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli A/B testing across coding agents is such a smart move. Been wanting something like this — knowing which agent actually ships better code for your stack is worth more than any benchmark.

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli A/B testing across agents is such a clean way to keep preferences honest, especially once you’ve got a few you bounce between.

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli I tried dmux, it is nice, but I really don't see big advantage over plain tmux. agents can be instructed to use worksheets. A/B is nice, but only for specific cases. Not convinced.

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli sounds right until you think twice

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli What's the most effective strategy for balancing A/B testing across multiple agents with varying strengths and weaknesses?

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli @Zai_org ?

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli Can you use it via web? Tried in console mode and it does not work well, because mostly my dev machine is remote: my agents are running in a sandboxed env on a remote server in vms for isolation.

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli Haha. We used to have that feature actually and removed it. Maybe someday in the future. Does ssh not work?

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli dmux supporting 11 coding agents with A/B testing out of the box is genuinely unhinged in the best way 🤯 now you can literally race Claude vs Cursor vs Copilot on the same task and see who wins. competitive coding agents is the energy we needed 🔥

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli Why stop here? Next-gen devs pick multi-agent

@claudeai @cline @Alibaba_Qwen @opencode @OpenAI @AmpCode @charmcli @cursor_ai @github @geminicli [Keanu Reeves Woah]