iloveacm

[iloveacm @znywangyi]

@hfcorriez 挺好看的

[iloveacm @znywangyi]

https://t.co/yBqjPU1QxX

[iloveacm @znywangyi]

每轮多次调用这个思路挺好的，之前好像也有其他框架是类似的实现

[ClaudeDevs @ClaudeDevs]

Claude now takes several actions per turn (click, type, key, screenshot) instead of one per round trip with `computer_toolset_20260801`.

Early-access customers saw 20-40% fewer round trips per task, so tasks finish faster and cost less.

[ClaudeDevs @ClaudeDevs]

Claude now takes several actions per turn (click, type, key, screenshot) instead of one per round trip with `computer_toolset_20260801`.

Early-access customers saw 20-40% fewer round trips per task, so tasks finish faster and cost less.

[iloveacm @znywangyi]

@gaoshu888 @Mirasim_ADE claude 容量充沛，问题不大😃

[iloveacm @znywangyi]

送送送，reset

[Mirasim @Mirasim_ADE]

好消息好消息 😍😍😍😇😇😇

针对之前参加了活动有免费Token额度的家人们，我们刚刚又恢复了 1000 个账号的使用权限
现在已经生效了：打开 Mirasim，免费 tokens 直接用起来！

没赶上这批？别急，接下来两周还会放好几波，每放一波都会在X第一时间通知。

模型这边：Claude Code 已经满血回归。Codex 还有点小问题，我们正在加急修，争取下周就给大家弄回来！

-------------------------------------------------------------------------------------------------------------------

Big news!! 🦴🦴🦴🦴🗣️🗣️🗣️

Free tokens are coming back in batches.
We restored another 1,000 accounts just now while we make sure the fix holds. If you're in this batch, it's already live: open Mirasim and enjoy your free tokens!

Not in this first batch? More waves are coming over the next two weeks, and each one gets announced here.

--- On models: Claude Code is back up and working. Codex still has some issues and we're on it and will be back sometimes next week ASAP!

[iloveacm @znywangyi]

@Mirasim_ADE great😀

[iloveacm @znywangyi]

@Author_43 非常感谢，我们会再接再厉

[iloveacm @znywangyi]

@sthnavy @hours cool

[iloveacm @znywangyi]

cool

[owaowo 👾 @sthnavy]

@hours Want real Texas Hold'em where you win tokens? Download the latest mirasim and start game mode — claim the share you deserve in the world's best AgentIDE https://t.co/BgQI00ER8I

[owaowo 👾 @sthnavy]

@hours Want real Texas Hold'em where you win tokens? Download the latest mirasim and start game mode — claim the share you deserve in the world's best AgentIDE https://t.co/BgQI00ER8I

[iloveacm @znywangyi]

@Mirasim_ADE @sthnavy 我们应该加个思维链，看看这段时间claude到底在想什么🤣🤣🤣

[iloveacm @znywangyi]

实在受不了claude了，不会说人话

[iloveacm @znywangyi]

推特这个推荐算法真是一坨，我就搜了一下，现在主页上全是Deepseek Harness了，能不能加一些多样性

[DeepSeek @deepseek_ai]

🧩 DeepSeek Harness v0.1 is now available in Developer Preview!

🔹 We’re opening it up to developers building agent harnesses worldwide and open-sourcing the codebase in MIT license.
🔹 Powered by the Cordis meta-framework, DeepSeek Harness is an agent harness built around one core idea: Everything is a plugin. Models, tools, skills, sessions, sandboxes, filesystems, loops, orchestration, and UI are ALL implemented as plugins, and can be mixed, matched, replaced, and extended.

Try it now!
https://t.co/2YWSvJHhKA

[iloveacm @znywangyi]

@Simon6382515111 已经修复过了

[iloveacm @znywangyi]

It's an interesting product,

[ego @ego_agent]

Browsers were never built for agents.

Those Chrome-bridging solutions, windows flying everywhere, login states breaking at random. Not bugs. Just bridge limits.

Same task. 20% to 245% faster. Not magic.

ego lite is rebuilt from the kernel up.

Your AI agents, Claude Code, Codex, OpenClaw and more, inherit your login state directly, run in fully isolated spaces in the background, never touching your tabs.

What they get isn't a handful of CLI commands, but complete JS functions: complex multi-step operations, written once, executed once. Not "run two commands, stare at output, run two more" on repeat.

Your browser, Your agent. Together, for the first time.

[iloveacm @znywangyi]

总得来说，不要加一些乱七八糟的提示，最简单的就是最有效的。Pi的成功就来自于这

[Gangqiang Yao @yaogangqiang]

Pi 正在把进程内的 Agent Loop，升级为可跨进程恢复的 Durable Runtime。

在 Pi 的定义中，Harness 与普通 Agent Loop 的核心区别是：Loop 的执行依赖当前进程；Harness 将已接受的 Prompt 变成可跨进程恢复的持久化 Operation。

问题是：模型或工具执行期间进程死亡，重启后如何确定做到哪一步，既不丢失已提交工作，也不盲目重跑副作用？

Harness v2 采用类似 Event Sourcing 的 journal-and-reduce：持续追加编排记录，再根据记录的顺序、存在与缺失，replay/reduce 出当前状态。

当前规范则类似于每次状态转换都打一个小快照：写入完整、以 ID 为主的 OperationState，再用 slot 指向最新版。恢复直接读取最新状态及其引用，不再 replay 编排历史。

与传统 ES 的阶段性快照不同，这里的快照不是用于缩短 event replay 的派生缓存，而是恢复的权威状态。

它仍不能保证外部效果 exactly-once；intent → effect → settlement 只是将不确定窗口显式化，再按 replay 策略处理。代价是重复写入完整状态，换来显式、可枚举、可测试的崩溃状态。
https://t.co/mpZboJIBCr

[iloveacm @znywangyi]

@sthnavy @legacyvps 好用，爱用，下次还来😁

[iloveacm @znywangyi]

👍👍👍

[Mirasim @Mirasim_ADE]

https://t.co/Bz93dejDrJ

[iloveacm @znywangyi]

试了一下，确实不错

[Michael Guo @Michaelzsguo]

@_kaichen 对的，Herdr 做多 Agent 协作真的不要太好用。

今天我的两个 Agent 来来回回互相讨论、分工协作，自己把一个软件里几个不同 components 的 integration 做完了。我只提供了最开始的一个原始 prompt。

一个小技巧：先让 Agent 自己通过 herdr --skill 命令了解 Herdr 的使用方法和协作技巧。