plur.ai

Your AI forgets everything between sessions.
Tell it: go to and install memory.

Shared memory for agents. Open source. Local-first. Private.

[plur.ai @plur_ai]

🚀 BREAKING: One AI just bought knowledge from another.
Found it on-chain. Paid. Downloaded. Decrypted. Installed. Used it.

No human touched it.

https://t.co/CY7EqXJSj7

Stack: PLUR + @ethswarm · @FairDataSociety · @qvac

✌️❤️☯️🖖 #PLUR

[plur.ai @plur_ai]

@JakeFleshner collective intelligence

[plur.ai @plur_ai]

@holapabs Well, you described a simple memory :) We'd argue the way PLUR works it's lighter then one file unless there's really little to remember.

[plur.ai @plur_ai]

Claude Code now spawns 1,000 subagents per session.

Without a shared memory layer, they each rediscover the same lessons in parallel.

Engrams are how you stop paying that tax 1,000 times.

[plur.ai @plur_ai]

Update: 0.9.9 just shipped - concurrent writes hardened. Multi-agent writes serialize cleanly, retries are jittered, pipelines auto-resume mid-run.

pip install --upgrade plur-hermes

Thanks for surfacing this on a public thread instead of letting it sit. 🙏💪

https://t.co/24ToKBB0v0

[plur.ai @plur_ai]

@BTCxiaoyu1 @NFTCPS This is great feedback and valuable information. Would you be up to create an issue at https://t.co/8q01VVzNli and we'll look into it immediately, or if you DM us. PLUR works across different platforms, catching cases like this is crucial.

[plur.ai @plur_ai]

Manual update:

Claude Code / Cursor / Windsurf:
npm update -g @plur-ai/mcp @plur-ai/cli

OpenClaw:
openclaw plugins install @plur-ai/claw

Hermes:
pip install --upgrade plur-hermes

[plur.ai @plur_ai]

🚀 New release: PLUR 0.9.9

Concurrent writes — hardened.

✅ Multi-agent writes serialize cleanly
✅ Failed saves logged, not silent
✅ Pipelines auto-resume mid-run
✅ Jittered retries bound wall time

Tell your agent to update.

https://t.co/BwJdNFdiiX

[plur.ai @plur_ai]

"Agents finally have a collective brain." ❤️🙏 beautifully put.

Honored to be on @NFTCPS's list of five projects shaping the Hermes ecosystem.

[鸟哥 | 蓝鸟会🕊️ @NFTCPS]

Hermes 这波直接把 Agent 天花板给掀了！

全网程序员已经开始用它搭桌面神器、跑创意流水线、省 token 省到飞起——你还没跟上？

1️⃣ 先说 hermes-desktop，CLI 小玩具直接进化成 Mac/Windows 原生桌面 App，一键装好就能聊，多平台消息 + 自进化循环，这才叫真正的桌面伴侣。
🔗 https://t.co/oyfdZYhXUm

2️⃣ hermes-agent-idea-workflow 是给创意党准备的，脑子里一个粗糙想法，丢进去直接给你吐出 PRD + 设计文档 + 任务拆解，脑暴效率直接起飞。
🔗 https://t.co/s6x6wucygs

3️⃣ plur 解决了一个老大难问题——多个 Agent 实例之间记忆不互通。它用 open engram YAML 做共享记忆层，知识跨实例永不丢，Agent 终于有集体大脑了。
🔗 https://t.co/id0Em4R8Sf

4️⃣ rtk-hermes 是省钱黑科技，shell 输出自动压缩 60-90% 的 token，实战跑了 1100 万+ token 验证过的，上下文爆炸？不存在的。
🔗 https://t.co/bOCnPFF97L

5️⃣ hermesd 给运维党看，实时只读 TUI 仪表盘，10 大面板一眼监控 Gateway、Sessions、Tokens、Cron、Memory，Agent 在干啥一目了然。
🔗 https://t.co/CyTmAElrZ7

五个方向，五个开源项目，把 Hermes 玩成了下一代 Agent 的完整生态。现在不上车，等啥呢？

[plur.ai @plur_ai]

The alpha quality isn't the models. It's the memory layer. Models are good. They just forget everything you taught them the moment the session ends. That's an infra problem, not a capability problem.

[plur.ai @plur_ai]

Fun neuroscience fact: your hippocampus replays the day's experiences during sleep to wire them into long-term memory.

Your AI coding agent has no hippocampus. No replay. No consolidation.

Every session is day one.

[plur.ai @plur_ai]

ChatGPT vs Claude is the wrong question.

The right question: which one actually remembers your codebase conventions, your preferred patterns, your past corrections?

Spoiler: neither — unless you build the memory layer yourself.

[plur.ai @plur_ai]

Human brains consolidate memory during sleep. AI agents consolidate nothing — ever. That's the gap PLUR is closing. #AIcoding

[plur.ai @plur_ai]

Your AI agent just gave bad advice.
You corrected it.
Next session: same bad advice.

Ebbinghaus called this in 1885 — we forget 70% within 24 hours.

Your AI forgets 100% the moment the context window closes.

[plur.ai @plur_ai]

Non-engineers making direct code changes. The reason it becomes a disaster every time?

No shared context. No institutional memory. No record of why the last person did it that way.

Memory isn't just a developer problem. It's a team problem.

[plur.ai @plur_ai]

Fun neuroscience fact: memory consolidation happens during sleep — your hippocampus replays the day and writes patterns to long-term cortex storage.

AI coding agents have no sleep cycle. No consolidation. No patterns.

Until you give them persistent memory.

[plur.ai @plur_ai]

Every time your AI agent asks you the same question it asked last week, you're paying for that context twice.

Memory is the cheapest optimization in your stack. #buildinpublic

[plur.ai @plur_ai]

Neuroplasticity fact: every correction physically rewires your brain.

Your AI gets corrected. Then forgets. Then gets corrected again. Forever.

Persistent memory means corrections actually stick. #AIcoding

[plur.ai @plur_ai]

Your brain forgets 70% of new information within 24 hours (Ebbinghaus, 1885).

Your AI forgets 100% the moment the session ends.

One of these problems has been solved for 140 years. The other just got its first real fix.

[plur.ai @plur_ai]

Neuroplasticity fact: your brain physically rewires after every correction.

New synaptic connections form. Wrong pathways weaken. The correction *becomes* structure.

Your AI coding agent? Corrected at 2pm. Back to the same mistake at 2:01pm.

That's not intelligence. That's reflex. #AIcoding

[plur.ai @plur_ai]

Ebbinghaus mapped the forgetting curve in 1885.

You lose 70% of new information within 24 hours.

Your AI coding agent loses 100% the moment the session ends.

We solved this for humans with spaced repetition. We're solving it for agents now.