Skip to content
Archive

Post

Back to deliverables

39 replies collected of 40 X reports
Jevon | AI企业级布局 @ValenSRfor711 ·

@berryxia Amazing! Changing the AI ​​like this won't affect the entire agent system!

original · zh

@berryxia 厉害 这样换AI 也不会影响整套Agent系统!

Mr Paper @Mr_paperi ·

@berryxia It's getting better and better with use, I'll bookmark it first.

original · ja

@berryxia 越用越溜了,先收藏

程序员潘哥 @mogician301 ·

@berryxia This is great. By digitizing yourself and creating a digital avatar, and connecting these elements, many personal decisions and tools become much easier to implement.

original · zh

@berryxia 这东西不错。 和自己数字化,弄一个数字分身,把这个打通了,很多个人决策和工具都更容易实现

阿西_出海(2.0版) @imaxichuhai · 13K

@berryxia I learned something.

original · zh

@berryxia 学到了

知不道吧 @zhibudaoba ·

@berryxia This is good news for those who frequently use Codex, Claude Code, and Cursor. Every time they switch, they have to re-explain, which is incredibly time-consuming.

original · zh

@berryxia 利好频繁用Codex、Claude Code、Cursor的那一撮人,每切一次都要重新解释,这件事太耗人了

BodhiAlpha @happy20250912 ·

@berryxia Memmy's idea of ​​sharing memories across agents is excellent!

original · zh

@berryxia Memmy 跨 Agent 共享记忆的思路很优秀~

懒家伙 @rogue30481577 ·

@berryxia The biggest problem right now isn't that the AI ​​isn't smart enough, but that it needs to be trained all over again every time a new agent is introduced. Memmy is working on solving this.

original · zh

@berryxia 现在最大的问题不是 AI 不够聪明,而是每换一个 Agent 都要重新培养一次。 Memmy正在解决这件事

AI智多星 @sunwei57247249 ·

@berryxia After restarting my computer, I thought I had lost all my projects. I opened Warp and there was nothing there. But when I used Memmy, I found all 210 items.

original · zh

@berryxia 电脑重启后以为项目全丢了,打开Warp啥都没有,用Memmy一查210个项全找到

1
王二狗 | Open to Work @IPFS15 ·

@berryxia Switching to a different tool eliminates the need to explain everything from scratch; this is truly a necessity.

original · zh

@berryxia 换个工具不用再从头交代,这个真刚需。

E吴大哥wu| @smark0428 ·

@berryxia Is Memmy's memory independent and separate from each agent?

original · zh

@berryxia Memmy是不是独立记忆,和每个agent分开的?

VV田 @check_xie ·

@berryxia Will this product just distill me up, so a clone can take my place and continue working?

original · zh

@berryxia 这个产品会不会直接把我蒸馏了,以后有个复制人代替我,继续工作

木马人 @cnyzgkc · 14K

@berryxia Thanks for sharing

original · zh

@berryxia 感谢分享

1
Cell 细胞 @cellinlab · 31K

@berryxia It seems like we haven't yet seen a game-changer in the memory field where someone can instantly recall something just by mentioning it.

original · zh

@berryxia 感觉 记忆这块 怎么 还没杀出了一个 一提就能想到的

余温 @gkxspace · 44K

@berryxia @grok What are the differences between it and other memory tools?

original · zh

@berryxia @grok 它和其它记忆工具相比有什么又是呢

1 1
阿熊学AI @JunSongYoung ·

@berryxia's statement, "The project wasn't lost; what was lost was the AI's understanding of the work environment," is so accurate. Remember, language preference is only superficial. Being able to grasp the project, skills, and pitfalls encountered is what truly shows you understand the user and makes you a reliable long-term colleague.

original · zh

@berryxia “项目没丢,丢的是 AI 对工作现场的理解”这句太准了。记住语言偏好只是表层,能接住项目、Skills 和踩过的坑,才算真正认识用户,也才像一个能长期配合的同事。

jackylin @jackylin888888 ·

@berryxia Great, I'll learn from that.

original · zh

@berryxia 不错,学习下

飞叔 @linghucong ·

@berryxia This is great! It feels better than any lobster-flavored product.

original · zh

@berryxia 这个不错啊 感觉比什么龙虾好用啊

铜匠AI・十点睡觉 @CopperForgeAI ·

@berryxia I've been working on agent teams lately and found memory management to be extremely difficult. This method is good.

original · zh

@berryxia 最近搞agent team发现记忆确实非常难处理,这个方法好

1
Jaron Lin 林佳龙 @JaronLin2005 ·

@berryxia Your analogy is spot on. From "temporary workers" to "shared archives," it perfectly captures the different levels of collaboration.

original · zh

@berryxia 你这个比喻太到位了,从“临时工”到“共享档案”,一下就把协作的层次说透了

林雪 @lnxu150850 ·

@berryxia I thought all my projects were gone after restarting my computer, I totally understand that feeling... Then Memmy found all 210 projects, and even knew which were official and which were backups, that's really impressive. Now, I have to explain the project background all over again every time I switch to another AI tool, it's so annoying. If this memory layer could really connect all the tools, that would be amazing.

original · zh

@berryxia 重启电脑后以为项目全没了,这感觉我太懂了……然后Memmy直接把210个项目全找出来,还知道哪个是正式的、哪个是备份的,确实牛。现在换个AI工具就要重新解释一遍项目背景,真的烦死了。这个记忆层要是真能打通各个工具,那就太香了。

1
AI奶爸 @zstmfhy · 29K

@berryxia This is essentially shared memory, right?

original · zh

@berryxia 这相当于是共享记忆了吧

1
阿链 @lian_doctor000 ·

@berryxia What's truly lost isn't the 210 directories, but the work sites scattered across various tools. Changing agents every time is like starting over, and that's the most expensive hidden cost.

original · zh

@berryxia 真正丢的不是那210个目录,是散落在各个工具里的工作现场。每换一次Agent,就像重新入职一遍,这个隐形成本才最贵。

一地鸡毛 @zhang_baoqing ·

@berryxia I've recently been using Pi to create custom agents and I've definitely found memory issues, especially when multiple agents need to work together.

original · zh

@berryxia 最近用Pi搞专属Agent确实发现记忆是个问题,尤其是要多agent联动的时候

Enzo @enzo66lin ·

@berryxia This is good, isn't it similar to storing Markdown files locally before?

original · zh

@berryxia 这个好,和之前md文件放本地是不是差不多

Yanhua @yanhua1010 · 34K

@berryxia Great stuff, a must-have for multi-agent use.

original · zh

@berryxia 好东西,多agent必备了

1
岚叔 @LufzzLiz · 22K

@berryxia I'm quite optimistic about this track: the underlying models and agents can be constantly changed, but a long-term oversight is needed at the top.

original · zh

@berryxia 我挺看好这个赛道:底下的模型和 Agent 可以一直换,上面需要有一个长期存在的总管,

1
Damien @brc20_1brc ·

@berryxia It can scan 210 projects and even knows which is upstream and which is replica, which is much better than just memorizing. The real problem is that the tools don't communicate with each other, and Memmy's attempt to solve this is a good direction.

original · zh

@berryxia 210个项目都能扫出来,还知道哪个是upstream哪个是replica,这比普通记忆强。 真正麻烦的是工具之间不互通,Memmy想解决这个,方向可以。

诺鸭船长3 @noahduck283 · 19K

@berryxia @grok Please conduct a comparative review of the current mainstream memory management methods.

original · zh

@berryxia @grok 做一下目前主流记忆管理方法的横评

1
fang么么 @me881112 ·

@berryxia This is so cool! It's an amazing tool; now I don't have to keep feeding myself data.

original · zh

@berryxia 太帅了吧。简直是神器,这样子我都不用反复在喂资料

Sac @Saccc_c · 30K

@berryxia This is so useful, I learned a lot 🫡

original · zh

@berryxia 太有用了,学习了 🫡

Steven蒙 @MiloSteven000 ·

@berryxia To be honest, I haven't used any memory-based tools before, and I probably don't have the intention to switch agents right now, as I feel the Codex context is sufficient. But I'd still like to give it a try, since I used to back up my long-term memories.

original · zh

@berryxia 说真的,我还没用过记忆类的工具,可能目前还没有切换agent的意愿,觉得codex上下文够用了。但还是可以体验体验,以前备份自己的长期记忆

1
马识途 @shitunote ·

@berryxia Cross-agent memory is a problem Some external tools can indeed help.

original · zh

@berryxia 跨agent记忆是个问题 一些外部工具确实能起到作用

1
Wei佳 @LiuweijiaVip ·

@berryxia The useful part is recovering context after a restart; otherwise memory is just another place to store stale state.

想风 @xaiwind ·

@berryxia Sharing memory is really great; I don't have to explain things to a new agent every time.

original · zh

@berryxia 有了共享记忆 真心不错 不用每次都和新开的agent重新解释

Rion Wu @rionaifantasy · 14K

@berryxia Recently, many AI memory-related products have emerged.

original · zh

@berryxia 最近很多ai记忆方向的产品涌现

Akinn @Akinnx2 ·

@berryxia Of the several AI programs you've used, this one with memory is quite rare. It's probably better than those that only remember preferences.

original · zh

@berryxia 用的几款ai,这款ai记忆还是挺少见,比单纯记偏好的强了吧~

阿森的资源库 @y4180741368915 ·

@berryxia Holy crap, this tool is amazing! Switching to AI means I won't have to explain projects from scratch anymore. The memory sharing is truly incredible, hahaha 😂

original · zh

@berryxia 卧槽这工具可以啊,换AI再也不用从头介绍项目了,记忆互通才真厉害啊哈哈哈😂

RuDong @zrudong ·

@berryxia, could you please advise on methods to improve recall in memory management?

original · zh

@berryxia 请教,记忆使用什么方式提高召回率