
Dim is now open source. An agent paradigm built on an event-driven timeline and a fully plugin-based runtime — it grows and iterates within the environment it runs in. Explore Dim and the source : github.com/arcships/dim

Your AI agents working as one. opensource in github : github.com/arcships/dim discord : discord.gg/cZx3MycCm5

Dim is now open source. An agent paradigm built on an event-driven timeline and a fully plugin-based runtime — it grows and iterates within the environment it runs in. Explore Dim and the source : github.com/arcships/dim

Pi和Dim是我最近一个月用的最多的Harness! 我已经很久没有打开过Claude Code和Codex了。
We heard there was a harness race. Dim-alpha ran its own lap. Our FrontierHarness self-run 68.5% pass rate $0.133 average cost per task The cost column is our favorite part.

欢迎newapi的领导莅临 dim 进行工作指导😎 @Ion_Mio_ @yyhhyyyyyy @t0ng7u @Card198454

dim alpha在frontier harness 1.1 上取得了68.5%的通过率,以及平均任务成本$0.133的成绩。 dim alpha还在迭代中,我们会不断发布新的成绩。
We heard there was a harness race. Dim-alpha ran its own lap. Our FrontierHarness self-run 68.5% pass rate $0.133 average cost per task The cost column is our favorite part.

Here are the charts for both Dim-alpha self-runs. FrontierHarness v1.1 68.5% pass rate · $0.133 avg. cost per task FrontierHarness v1.0 · DeepSeek V4.1 Flash 82.2% pass rate · $0.07625 median cost per task (The original v1.0 leaderboard used Kimi K3.)

One more lap, this time on FrontierHarness v1.0. Dim-alpha + DeepSeek V4.1 Flash 82.2% pass rate $0.07625 median cost per task Results from our own run. The original leaderboard used Kimi K3.

The video overlays our self-run result on the public FrontierHarness chart for illustration. If you’ve run the eval too, we’d love to compare setups. Explore Dim on GitHub ↓ github.com/arcships/dim

We heard there was a harness race. Dim-alpha ran its own lap. Our FrontierHarness self-run 68.5% pass rate $0.133 average cost per task The cost column is our favorite part.

Rutis 0.22 今天发布了。这次主要为了保证安全和可靠,补了大量的形式化证明和测试:会话协议有了正式规范,规则逐条对到实现,模型检查、竞态测试、模糊测试都加了一批,原本排在 0.23 的形式化工作也一起做完了。插件使用者平稳升级即可。 github.com/arcships/rutis/re… 一些变化/bugfix: - emit 已派发出去后收到 cancel:照常派发,回复丢掉 - 宿主调用运行时刚撤掉的监听器:返回空值 - 插件开始卸载后调 ctx.provide:拒绝,Python 和 Bun 同样 - 心跳启动时机:改为 hello 发成功之后,清理旧租约慢也不误关新连接 - Kotlin / Swift 收到未知字段:拒绝该次调用

dim-alpha 批量蹬issue,内存也就几百兆 还是很爽的😎 昨天分享的时候忘说了,求求大家走过路过点个star github.com/arcships/dim

dim alpha在巨量agent数量的情况下(50-100),内存只有400MB占用。cpu差不多只有5%-10%。 做企业级场景真心可以考虑dim做基础框架,这是ts类agent无法支持的能力。

忘记说了, Dim 也是 MIT 开源的国产 Harness , 而且是 Rust,相对性能占用还是很小的. 如果你需要搓个东西不想自己折腾 Agent 的时候也可以考虑考虑,直接把仓库丢给你的 Agent 他就懂了 github.com/arcships/dim

To run 1 million tokens, we split our inference service twice.


不要在启用Dark Reader时打开dimcode的官网,不然你没办法欣赏到官网做的有多好看。 @DimAgentai

分享一下使用deepseek v4.1 flash做TLA模型实践过程 可能在架构与需求的设计阶段就先做证明会更好


We're going to COSCon'26 this Sunday! @imwritingbugs will join the Open Agent Ecosystem BoF to discuss DimAgent and OpenSource. Hope to chat with you there! Can't make it? Leave a question below, and we'll share our thoughts during the BoF livestream.See you in Hangzhou! bagevent.com/event/9280412