Register and share your invite link to earn from video plays and referrals.

JackChen
@JackChen_x
Ex PM (¥100M+ revenue).
267 Following    1.1K Followers
Multi-agent's quiet problem: token cost scales with agents × turns × tool calls. It compounds fast , and that's the bill that kills production rollouts. @MiniMax_AI is now a native adapter in @JackChen_me's open-multi-agent, so goal-driven agent teams stay affordable in production, not just in demos. 👇
Show more
Shipped open-multi-agent 1.2. Biggest external-contributor release yet: @ibrahim_kzmv : MCP, context strategies, CLI, glob hkalex(github): DeepSeek + MiniMax providers mvanhorn(gihtub): Groq provider Plus tool output cost controls + error/abort fixes. 🙏
Show more
Thank you @guk2472 . This directly influenced v1.1.0's default-deny context isolation in open-multi-agent. Credit to @ibrahim_kzmv who wrote the PR that landed the change.
Show more
@JackChen_x 多agent编排看着很美,但跑过就知道最大的坑是上下文隔离。agent共享太多context会串味,一个agent的中间结果污染另一个的判断,debug极痛苦。Claude Code的做法是每个sub-agent只给最小文件集+工具权限,比花哨框架靠谱。框架解决调度,但生产里context泄漏才是真正杀人的
Show more
How I Found Real User Out of 5,387 Stars
分享一个我最近做的事:开源项目上线一周,5,387 个 Star 进来了,我没有继续做推广,而是去找谁真的在用。 Star 涨得快心里反而虚,点 Star 是一秒钟的事,但是把我的框架集成进自己的产品是另一回事。我想知道这一周里有没有人真的跨过了这条线。 方法很笨:用 GitHub 代码搜索,搜谁在自己代码里 import 了我的包名。搜出来一堆结果,大部分是噪音——fork 复制品、README 抓取器、自动文档站。逐个点进去排除,只留 package.json 里有依赖、代码里真的在调 API 的。 上线才一周,真找到了一个。一个土耳其安全研究员做的 WordPress 安全分析工具。传统做法是人工读几千行 PHP 确认漏洞,他用我的框架做 AI 执行引擎,让 Agent 替代人工读代码。我让 AI 帮我把他的 3 个源文件全读了,800 多行核心代码,框架的 4 种编排模式全用上了,还在上面自建了一套工具风险分级系统。不是跑个 demo,是产品级集成。 上线一周就有人做到这个程度,比 5,387 个 Star 让我兴奋多了。除了他,还有人私信确认在用、有人提了深度 bug report——但能在代码里确认的产品级集成,就这一个。 然后我想联系他。让 AI 帮助我起草了一条 GitHub Issue,发之前逐句核查,抓出 4 个事实错误——AI 自己查出来 3 个,第 4 个是我凭记忆追问才发现的。AI 也会漏,人必须复核。全部改完才发出去。 如果你也以开源为策略获客,建议早点去搜一下谁的代码里真的 import 了你。Star 数不等于用户数,早一天知道真实情况,早一天能做对决策。
Show more
Thanks @swyx, saw open-multi-agent got mentioned in your AI News from Reddit. Checked my traffic and there was a clear spike from Latent Space readers. Been a subscriber for a while. Good stuff.
Already works with open-multi-agent — pure local agent team with tool-calling, zero API cost.
Meet Gemma 4: our new family of open models you can run on your own hardware. Built for advanced reasoning and agentic workflows, we’re releasing them under an Apache 2.0 license. Here’s what’s new 🧵
Show more