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.

Fascinating work! Beyond the impressive visual rendering in Three.js, turning session insights into universal agent skills solves one of the biggest friction points in modern AI development.

Setup was one prompt. I pasted the EverOS console's install prompt into Claude Code, and it wired up the API itself. One build later: 22 episodes, 7 cases, 7 skills, all distilled by EverOS. Then a brand-new session recalled those skills and added a city bus, limousine, off-roader, ambulance, fire engine, and race car. All six matched the blueprint very well.

Cars are just my excuse. Anything an agent learns the hard way can stick around. Your support bot solved a gnarly ticket last month? Next time it already knows the fix. Forgot why the team picked this approach back in spring? Ask, and the reasoning is still there. One agent finally cracks a tricky bug? Every other agent on the team gets that skill too. People are already building wild stuff with it: a memory aid for people with Alzheimer's, coding agents that share one brain, a mystery game where the NPCs remember, smart glasses that recap your runs. See more:

Want your agents to keep what they learn between sessions? Try EverOS Cloud: Subscribe to Pro now and get 3x your monthly usage allowance for 6 months.

I'd love to see how this holds up on the next build. If the shader and assembly lessons from this one get picked up by a different agent on a different project, that's a real test of the idea. #EverMind #EverOS

The cyanotype blueprint look is gorgeous, and the stage-by-stage assembly is a great touch. Building something this specific with shaders is hard enough once, so the idea of keeping what the agent figured out is the part that gets me. #EverMind #EverOS

The part about other agents being able to reuse skills learned by Claude Code is interesting. Can an agent using a different model adapt those skills to its own workflow, or does it need to follow the original implementation? #EverMind #EverOS

这个太酷了！EverOS Cloud × Claude Code 🧵 Claude Code搭了个Three.js的living blueprint：cyanotype纸上的ink shader，分阶段GPU组装，17种车型 最关键的是EverOS Cloud把学到的东西转成self-evolving skills，任何agent不只是Claude Code都能recall。经验变成可迁移的技能，这个太重要了

The drafting-table UI with the stage slider is really well done. Curious which parts of the process ended up as skills: the shader setup, the assembly sequence, or the vehicle data? #EverMind #EverOS

"Any agent, not just Claude Code, can recall" is the interesting bit. Skills that aren't locked to one tool would mean the next project doesn't start from zero, even if you switch agents halfway through.#EverMind #EverOS

Claude Code built living blueprint in Three.js + EverOS Cloud turned learning into recallable skills 这个组合是未来：Agent做项目，然后EverOS把轨迹沉淀成Skill，别的Agent也能直接调用。下次做类似3D组装、shader效果就不用从零开始了 ink shaders on cyanotype paper这个审美也太绝了

Anything an agent learns the hard way can stick around - that's exactly the gap. Support bot fix once and remember forever is huge. #EverMind #EverOS

Love the use cases. One agent cracks a bug, every other agent gets that skill too. That's real team memory. #EverMind #EverOS

This is the perfect demo of EverOS Cloud × Claude Code Claude Code builds the complex thing (Three.js blueprint, ink shaders, stage-by-stage GPU assembly), EverOS Cloud turns what it learned into self-evolving skills that any agent can recall. Not just code saved, but the HOW saved as Skill. That's how you stop rebuilding the same thing

着色器那部分最难沉淀成经验，参数一换风格就跑偏

The blueprint looks incredible! The real game changer is turning what Claude Code learns into reusable skills across different AI agents. That’s far more valuable than a one-off project.

Impressive on its own. But then: EverOS Cloud turned what it learned into self-evolving skills.

This is a perfect demo of why persistent agent memory matters. Claude Code learns the blueprint style once, then any agent can pick it up later.

Amazing how persistent memory can turn one project’s experience into reusable skills for every AI agent!

Curious: 请问 EverOS 和直接用 Markdown 文件管理相比,主要优势在哪?跨设备同步是怎么做的? (How does EverOS compare to managing plain Markdown files? How does cross-device sync work

17 vehicle types with a consistent blueprint style? That's impressive. Definitely a skill worth reusing.

With 17 vehicle kinds, I’m curious what became a reusable skill—the GPU assembly process, the ink shader setup, or both? Seeing another agent apply it to an 18th vehicle would be a compelling follow-up.

The fresh-session test is the most convincing part. It’s one thing to store skills, but seeing another session apply them to new vehicle types shows why cross-agent memory is worth exploring.

这个 demo 很能说明问题： • Claude Code 一次提示完成 EverOS 接入 • 构建出带自定义着色器的 Three.js 交互蓝图（17 种车型 + 分阶段组装） • EverOS 自动把过程蒸馏成 Cases + Skills • 全新会话直接调用这些技能，继续扩展车型且风格一致 Agent 终于不用每次从零开始了。

Have you tried handing those seven skills to Codex for vehicle #18? I’d be curious whether it keeps the ink shader and assembly sequence without you having to explain the blueprint style again.

Yep, tried it. Codex recalled the same skills and got going without me re-explaining anything. Pretty good.

This is a great example of turning one-off AI work into reusable intelligence. Self-evolving skills could make cross-agent workflows much more powerful! 🚀

This is a sharp demo of what persistent agent memory actually looks like in practice.

This is wild, Claude Code’s cyanotype blueprint in Three.js with those ink shaders looks stunning. Even better EverOS turns the session into self-evolving skills any agent can reuse. One prompt → persistent knowledge. Seriously impressive work.

🔥 This is the future of AI! 🤖 Claude Code brings creativity to life, while EverOS Cloud transforms experience into reusable, evolving skills. 🧠🚀 The ability for any AI agent to learn, remember, and improve is a massive step toward truly intelligent automation!

The useful part is that the experience doesn't stay locked in one session. The build was saved as a Case, distilled into reusable Skills, and the next agent can retrieve them before starting a similar task instead of figuring it out again.

The real power of AI isn't just what it can build—it's what it can remember and reuse. 🧠

The portable part is what stands out: skills an agent can recall instead of re-deriving them every session. Are those versioned? Curious how you stop a stale skill from poisoning later runs.

Claude Code builds it, EverOS Cloud makes the learning reusable. 🔥

A healthy indicator of the AI future

The ability to turn past experience into reusable skills is what makes AI agents truly useful long-term. Great concept!

yes

The real power is turning one agent’s experience into skills for every agent. 🚀