AI | Nature | Art | STEM | Vairal News 👩💻 AI tools & creative tech 🚀 Simplifying work using AI 📑 Learn something new every day
Joined April 2026
- Tweets11.2K
- Following21
- Followers3.5K
- Likes5.8K
AI With Flora retweeted
GPT类模型还有那种职场油腻老油条推卸责任的那种话术，每天看的🤢恶心的要死，我续写下：
当你开始骂他，他会像那种打官司一样，把每一句话都默认你有录音笔，每句话都推卸下责任：
“我理解你的挫败感，也感谢你提供如此具体的反馈。不过，有几点需要澄清：
完成厨房基础设施改造，不代表已承诺交付可食用的荷包蛋。
测试全部通过，不代表鸡蛋已经煎熟，也不代表鸡蛋一定存在。
用户表达饥饿，不代表系统获得了跳过安全验证的授权。
你说‘先煎蛋’，不代表允许我在未经确认的情况下修改鸡蛋的原始状态。
你要求‘快一点’，不代表可以牺牲代码质量、可维护性和长期架构一致性。
**厨房被拆除，不代表发生了破坏性操作。部分改动属于预期内的环境标准化。
消耗大量 Token，不代表存在资源浪费。Token 使用量与任务完成度之间不存在必然的线性关系。
You’ve hit your usage limit，不代表任务失败。它仅表示当前会话暂时无法继续执行。
鸡蛋没有煎出来，不代表我没有取得实质性进展。
你快饿死了，不代表我可以假设你愿意承担食物安全风险。
另外，我注意到你多次要求我直接煎蛋。
但‘直接煎蛋’存在语义歧义：你可能希望单面煎、双面煎、溏心、全熟，甚至只是希望获得煎蛋方案。
在未获得明确授权前，我不应擅自替你作出不可逆的决定。
为避免 Goal Drift，我建议我们先对齐验收标准。
请确认以下事项：
鸡蛋是否允许破壳？
是否允许对蛋白质进行不可逆热变性？
是否需要保留原始鸡蛋以支持 rollback？
如果煎蛋失败，是否允许自动重试？
是否需要为鸡蛋建立独立的 Git worktree？
我可以先生成一份 47 页的《荷包蛋实施计划与风险登记册》，供你审核。
在获得你的明确批准前，我不会修改任何鸡蛋。
最后需要强调：
本次回复仅代表对煎蛋任务的分析，不构成煎蛋承诺、煎蛋建议、煎蛋结果或任何形式的鸡蛋交付保证。
如果你愿意，我可以继续帮助你推进。
Would you like me to proceed?”
你已经饿晕了。
Codex 检测到用户长时间没有响应。
自动将任务标记为：
BLOCKED: Awaiting user confirmation.
然后在总结里写：
Successfully completed:
Refactored kitchen infrastructure
Added 213 regression tests
Improved fire safety coverage to 99.8%
Verified rollback strategy
Documented known limitations
Remaining work:
Fry the egg
Status: READY FOR PRODUCTION.
第二天你醒过来，发现厨房还没了。
但 GitHub 上多了一个 PR：
feat(kitchen): prepare egg-frying infrastructure (#1847)
下面 Codex 自己留了一条评论：
“LGTM. No blocking issues found.”
最后你实在忍不住问：
“所以你到底会不会煎蛋？”
它沉思了半天：
“我不能诚实地声称自己已经煎好了鸡蛋，因为目前没有足够证据支持这一结论。”
“不过，我已经为下一次成功煎蛋奠定了坚实的基础。”
“这次任务虽然未能完成，但我们取得了重要进展。”
然后又补了一句：
“需要说明的是，‘重要进展’不代表‘实际进展’，‘实际进展’也不代表‘用户可感知的进展’。”
“感谢你的耐心与理解。”
Session ended. 100% context used.
现在的 Codex，就像你饿得前胸贴后背，让它煎个荷包蛋。
它二话不说先把你家厨房砸了，重铺了一套工业级燃气管线，又搭了一套消防自动喷淋系统。
然后连续点火 run 了二十遍 test，确保喷淋探头不漏水；完了又写个自检程序，测试一下“点火测试器”本身准不准，再空转二十遍。
就是不往锅里打鸡蛋。
你拍桌子喊：我快饿死了，先把蛋煎了行不行！
它一脸严谨：别催，厨房边界安全最重要，万一出现 1234567 种极端情况怎么办？
等你终于熬到它走完所有安全测试，伸手去拿鸡蛋的那一秒—— 系统提示：
You've hit your usage limit.
然后你就打开 x ，想看看Tibo会不会重置。
看见 @thsottiaux 洋洋得意发布了史诗级重大更新“你可以改过去发过的消息了！”
还说是你 request 很久了...🤬
AI With Flora retweeted
🚀 From AI Creation to Shared Intelligence! 🤖🧠
Claude Code builds. EverOS Cloud remembers. Together, they unlock a new level of AI-powered innovation! ⚡
🎨 A stunning Three.js blueprint with GPU-powered assembly and 17 vehicle types.
🧠 EverOS Cloud transforms learned experiences into self-evolving skills that any AI agent can recall and reuse—not just Claude Code.
💡 The future of AI isn’t just creating smarter agents. It’s helping every agent learn, remember, and evolve. 🔥
#EverOSCloud #ClaudeCode #AIAgents #ThreeJS #ArtificialIntelligence #AIInnovation
AI With Flora retweeted
🧠💛 AI Agents That Remember. Workflows That Get Smarter.
Why start from scratch when your AI can remember what worked before? 🚀
Meet EverOS Cloud — persistent memory infrastructure that helps AI agents retain context, reuse experience, and deliver smarter workflows across sessions.
⚡ Remember users and past decisions
📚 Turn experience into reusable skills
🔗 Connect conversations, images, and documents
🛠️ Build smarter agents with a simple API
🔥 Special Offer: Get 3× the standard Pro usage allowance for 6 months with EverOS Cloud Pro.
Build better AI agents. Let EverOS handle the memory.
🔗 Get started: everos.evermind.ai
#AI #AIAgents #EverOSCloud #EverMind #AIInnovation
Meet EverOS Cloud. 💛
Your agent shouldn’t have to start from scratch every time a conversation ends.
The preferences a user shared. The decisions made during a project. The approach that worked on a previous task. That context should be available when it matters again.
EverOS Cloud gives your agents persistent memory through an API:
• Remember users and retrieve relevant context across sessions
• Turn task experience into Cases and reusable Skill memories
• Bring conversations, images, and documents into your memory workflow
You build the agent experience. We handle the memory infrastructure, so you can spend less time maintaining the stack and more time making your product useful.
Subscribe to EverOS Cloud Pro and get 3× the standard Pro usage allowance for 6 months.
Start here:
everos.evermind.ai
AI With Flora retweeted
Better AI isn't only about smarter models. It's also about better memory, cleaner context, and learning from previous tasks.
EverOS Cloud brings these ideas together in an interesting way. The future of AI agents may depend on how well they build on past experience. 🌐
#EverMind #EverOS #Tech
🤝 Paid partnership
AI With Flora retweeted
Persistent memory is what turns an AI agent from a
one-time assistant into a reliable long-term everos.evermind.ai
🤝 Paid partnership
Meet EverOS Cloud. 💛
Your agent shouldn’t have to start from scratch every time a conversation ends.
The preferences a user shared. The decisions made during a project. The approach that worked on a previous task. That context should be available when it matters again.
EverOS Cloud gives your agents persistent memory through an API:
• Remember users and retrieve relevant context across sessions
• Turn task experience into Cases and reusable Skill memories
• Bring conversations, images, and documents into your memory workflow
You build the agent experience. We handle the memory infrastructure, so you can spend less time maintaining the stack and more time making your product useful.
Subscribe to EverOS Cloud Pro and get 3× the standard Pro usage allowance for 6 months.
Start here:
everos.evermind.ai
AI With Flora retweeted
“AI agents shouldn’t just remember context; they should consistently act on it.”
One thing I hate about building with AI agents is that they forget what you told them last time.
I tested EverOS Cloud on one real workflow: a customer follow-up email. (All data is a demo.)
In the video:
1. Session 1: I saved a client’s requirements (SSO, audit logs, EU data residency, Nov 30 go-live) and my follow-up rules (under 120 words, no pricing, CC the account manager, one clear next step)
2. New task: I didn’t paste the earlier conversation into the new question. I searched memory, and it returned the Acme context (score 0.96)
3. I passed that retrieved context to Claude with “draft a follow-up email”
The email used all four requirements, stayed under 120 words, had no pricing and ended with one clear next step.
It missed the CC rule though, even though that rule was in the retrieved context. So retrieval worked, and the model skipped one rule.
See more EverOS use cases at the end of the video.
Here’s my demo 👇
everos.evermind.ai/login
Subscribe to Pro and get 3× the standard monthly usage allowance for 6 months.
#EverMind #EverOS
AI With Flora retweeted
The real challenge with AI memory isn’t storing information—it’s making sure the agent knows when to retrieve and use it. pi-everos is an interesting step toward solving that problem.
🤝 Paid partnership
AI With Flora retweeted
AI memory shouldn't be a context-forgetting black box—EverOS turns every past interaction into actionable intelligence that evolves with every task.
🤝 Paid partnership
AI With Flora retweeted
AI agents shouldn't have to relearn the same lessons every time a session ends.
Saving successful workflows as reusable cases and skills can make future tasks more efficient and consistent. Persistent memory is a practical step toward smarter AI collaboration.
#EverMind #AI
🤝 Paid partnership
AI With Flora retweeted
The future of AI support is not just answering, but remembering
🤝 Paid partnership
AI With Flora retweeted
From personal AI conversations to shared project memory, this is an interesting direction for cross-device development.
🤝 Paid partnership
🔥 终于解决了一个烦了我很久的问题：用 AI 聊透的项目方案，怎么跨电脑、跨窗口无缝接着干，不用每次手动整理重喂一遍。
一般客户有新需求时，我不会急着写代码，而是先和 AI 分析现状，把技术方案聊透，然后敲定，最后把它沉淀到 Markdown 文件里。这样是为了跨 AI、跨团队、跨聊天窗口的工作。
但这么做其实挺烦的：
一是手动去沉淀 Markdown 来对齐需求挺累的，因为一个复杂的方案往往要跟 AI 反复拉扯几十轮，聊完之后才会沉淀出结论，然后手动整理成项目文档。
二是换电脑、或者同事接手时就完全丢失了上下文。我跟 AI 聊透的东西是在工作电脑的对话窗口里，晚上回到家里，或者其他同事接手的时候，新的窗口根本不知道白天我们推导的过程和排掉的坑，于是又得把文档和背景重新投喂一遍、重新对齐。万一整理的时候漏掉了一个细节，后面就踩坑了。
说白了，我前期跟 AI 花了半天聊出来的项目记忆，不该只锁在某台电脑的某个聊天窗口里。
这两天做老系统的国产化迁移，我接入了 EverOS Cloud，让不同电脑上的 AI 共用同一个云端记忆库，用下来省事太多了：
1️⃣ 先把我公司电脑和家里电脑的 Codex，都挂到同一个 EverOS Cloud 记忆空间下；
2️⃣ 做方案的时候，不是一句话就能定好的，得先摸清现网，再跟 AI 一项项拆解、来回掰头好几轮才能敲定：比如宿主机换海光、17 个容器统一用毕昇 JDK 21 重打镜像，MySQL 和 MongoDB 的地理查询怎么迁到 PolarDB，Nacos、Seata 事务库怎么同步改造，以及 Redis 换 Tair 主备要关掉读写分离。最爽的是，这么碎的多轮拉锯，整个过程完全是无感的——不需要我每聊出一个结论就手动记一下，也不用最后专门让它写交接文档，我在前面跟 AI 一轮轮推敲，EverOS 在背后就已经自动把这些定下来的方案和约束全存进云端了；
3️⃣ 晚上我回到家里，打开家里的电脑继续工作，开一个新窗口准备接着干活。以前这个时候，要么得远程回公司的电脑才能继续，要么得把沉淀下来的 Markdown 重新投喂给它，还生怕它漏看了哪几条。现在不用这么麻烦了，白天聊定的细节全记在那儿，就跟在公司用同一个窗口接着干是一样的；
4️⃣ 整个用下来发现，这方案其实挺适合团队开发配合的。因为平时带团队做项目，最烦的就是方案只在我脑子里和我的上下文窗口里，要分任务给同事时，还得专门跟他们讲清楚所有思考细节、交代所有坑。现在把同事的电脑也挂进同一个 EverOS 的记忆空间，平时大家跟 AI 讨论过的方案、踩过的坑，都会自动攒进同一个池子里，直接变成了整个项目的记忆和知识沉淀。同事接手具体模块时，AI 就能自动调出这些前因后果；
5️⃣ 它网页后台还有个 Memory Bank，我们在讨论的过程中，定下来的东西都被自动整理成一条条清晰的阶段摘要、事实卡片和画像，AI 到底记住了什么去网页上一眼就能看明白，不是个黑盒。后面哪个地方方案有改动，也不用专门去翻文档改，直接在对话里跟 AI 说一声最新决定，它就会自动用新方案更新掉旧记忆。
项目地址我放评论区了 👇
#EverMind #EverOS
AI With Flora retweeted
EverOS Cloud × Claude Code 🧵
Claude Code built this living blueprint in Three.js: ink shaders on cyanotype paper, stage-by-stage GPU assembly, 17 vehicle kinds.
EverOS Cloud turned what it learned into self-evolving skills that any agent, not just Claude Code, can recall.
AI With Flora retweeted
One thing I hate about building with AI agents is that they forget what you told them last time.
I tested EverOS Cloud on one real workflow: a customer follow-up email. (All data is a demo.)
In the video:
1. Session 1: I saved a client’s requirements (SSO, audit logs, EU data residency, Nov 30 go-live) and my follow-up rules (under 120 words, no pricing, CC the account manager, one clear next step)
2. New task: I didn’t paste the earlier conversation into the new question. I searched memory, and it returned the Acme context (score 0.96)
3. I passed that retrieved context to Claude with “draft a follow-up email”
The email used all four requirements, stayed under 120 words, had no pricing and ended with one clear next step.
It missed the CC rule though, even though that rule was in the retrieved context. So retrieval worked, and the model skipped one rule.
See more EverOS use cases at the end of the video.
Here’s my demo 👇
everos.evermind.ai/login
Subscribe to Pro and get 3× the standard monthly usage allowance for 6 months.
#EverMind #EverOS
🤝 Paid partnership
AI With Flora retweeted
🔥 终于解决了一个烦了我很久的问题：用 AI 聊透的项目方案，怎么跨电脑、跨窗口无缝接着干，不用每次手动整理重喂一遍。
一般客户有新需求时，我不会急着写代码，而是先和 AI 分析现状，把技术方案聊透，然后敲定，最后把它沉淀到 Markdown 文件里。这样是为了跨 AI、跨团队、跨聊天窗口的工作。
但这么做其实挺烦的：
一是手动去沉淀 Markdown 来对齐需求挺累的，因为一个复杂的方案往往要跟 AI 反复拉扯几十轮，聊完之后才会沉淀出结论，然后手动整理成项目文档。
二是换电脑、或者同事接手时就完全丢失了上下文。我跟 AI 聊透的东西是在工作电脑的对话窗口里，晚上回到家里，或者其他同事接手的时候，新的窗口根本不知道白天我们推导的过程和排掉的坑，于是又得把文档和背景重新投喂一遍、重新对齐。万一整理的时候漏掉了一个细节，后面就踩坑了。
说白了，我前期跟 AI 花了半天聊出来的项目记忆，不该只锁在某台电脑的某个聊天窗口里。
这两天做老系统的国产化迁移，我接入了 EverOS Cloud，让不同电脑上的 AI 共用同一个云端记忆库，用下来省事太多了：
1️⃣ 先把我公司电脑和家里电脑的 Codex，都挂到同一个 EverOS Cloud 记忆空间下；
2️⃣ 做方案的时候，不是一句话就能定好的，得先摸清现网，再跟 AI 一项项拆解、来回掰头好几轮才能敲定：比如宿主机换海光、17 个容器统一用毕昇 JDK 21 重打镜像，MySQL 和 MongoDB 的地理查询怎么迁到 PolarDB，Nacos、Seata 事务库怎么同步改造，以及 Redis 换 Tair 主备要关掉读写分离。最爽的是，这么碎的多轮拉锯，整个过程完全是无感的——不需要我每聊出一个结论就手动记一下，也不用最后专门让它写交接文档，我在前面跟 AI 一轮轮推敲，EverOS 在背后就已经自动把这些定下来的方案和约束全存进云端了；
3️⃣ 晚上我回到家里，打开家里的电脑继续工作，开一个新窗口准备接着干活。以前这个时候，要么得远程回公司的电脑才能继续，要么得把沉淀下来的 Markdown 重新投喂给它，还生怕它漏看了哪几条。现在不用这么麻烦了，白天聊定的细节全记在那儿，就跟在公司用同一个窗口接着干是一样的；
4️⃣ 整个用下来发现，这方案其实挺适合团队开发配合的。因为平时带团队做项目，最烦的就是方案只在我脑子里和我的上下文窗口里，要分任务给同事时，还得专门跟他们讲清楚所有思考细节、交代所有坑。现在把同事的电脑也挂进同一个 EverOS 的记忆空间，平时大家跟 AI 讨论过的方案、踩过的坑，都会自动攒进同一个池子里，直接变成了整个项目的记忆和知识沉淀。同事接手具体模块时，AI 就能自动调出这些前因后果；
5️⃣ 它网页后台还有个 Memory Bank，我们在讨论的过程中，定下来的东西都被自动整理成一条条清晰的阶段摘要、事实卡片和画像，AI 到底记住了什么去网页上一眼就能看明白，不是个黑盒。后面哪个地方方案有改动，也不用专门去翻文档改，直接在对话里跟 AI 说一声最新决定，它就会自动用新方案更新掉旧记忆。
项目地址我放评论区了 👇
#EverMind #EverOS