Hurricane Dancing

Hi — I’m an indie builder practicing building in public .
I’m building AI products and sharing small demos + design notes as I go.

[Hurricane Dancing @HDancing49123]

@MoByhao 感觉就像这样 把一个事情 随着做的过程中 拆的越来越好细致 每一个点提升1%反馈在整体上就是巨大的

[Hurricane Dancing @HDancing49123]

随着通用人工智能(ai)离我们越来越近 成为生活不可缺少的工具 思考工具本身带来的双面性是一个不错的思考角度

[Hurricane Dancing @HDancing49123]

如果一直使用默认配置(官方的默认system prompt)的ai会容易出现这样的现象
由于通用智能助手的设计 你总是会站在离问题最近的那一侧(因为要展现它的通用和helpful)
让你一直处于碎片化的局部认知上 如果你没有外界组织作为依托时间久了很容易让你的认知处于一种悬空的状态

[Hurricane Dancing @HDancing49123]

我认为解决这个问题的方案 可以从这个角度去思考：你使用de是什么工具 工具将问题展开成了一个什么样的表征对象 长期处于这样的交互环境 又会怎么反过来塑造你自己的认知取向 你自己的注意力导向?

[Hurricane Dancing @HDancing49123]

我自己就是这种症状的受害者 它让我失去了组织知识 组织认知网格的能力 而不在像传统教学 课程传授的连续的可过渡的认知状态

[Hurricane Dancing @HDancing49123]

@timfrancisco @AriaWestcott You’ve already built the iCloudPC/external-drive setup I’m researching. If you also use Photos on a Mac with a 500GB+ library would you compare current options and one unbuilt concept for 10 min No install photo access purchase or compensation.

[Hurricane Dancing @HDancing49123]

@afrkml SSD 2 TB vs iCloud 2 TB ini persis studi saya. Apakah Anda juga memakai Photos di Mac dengan pustaka 500 GB+ Jika ya bersedia ngobrol 10 menit dalam bahasa Inggris tentang opsi saat ini dan konsep yang belum dibuat Tanpa akses foto instalasi pembelian atau bayaran.

[Hurricane Dancing @HDancing49123]

Use iCloud Photos on a Mac with 500GB+ Have or plan an external drive/NAS and expect to migrate or reduce iCloud use within 6 months Seeking up to 5 people for 10-min product research. No install login photo access purchase or compensation. Reply/DM “migration study.”

[Hurricane Dancing @HDancing49123]

@ClaudeDevs

[Hurricane Dancing @HDancing49123]

@neil_croy Exactly. I actually call this the 'fingerprint of thought.' Extracting and utilizing this fingerprint is the only way to pull the model's output away from the 'mean' and align it with our true intent.🤛

[Hurricane Dancing @HDancing49123]

Most AI products fail because founders fundamentally misunderstand what an LLM actually does.

[Hurricane Dancing @HDancing49123]

If you miss this, you are missing the whole point.
Everything else is just in service of this core principle. How to get more precise context? How to maintain continuous memory? It all revolves around this single idea.

[Hurricane Dancing @HDancing49123]

AI is all about CONTINUING something, not GENERATING a whole new thing.

[Hurricane Dancing @HDancing49123]

When building AI-native products, the most critical step is understanding the true nature of AI.
At a fundamental level, AI is simply predicting the probability of the next token. But if we abstract the action of "predicting," we realize something profound:

[Hurricane Dancing @HDancing49123]

@p_d_d_t Cool icon! Just out of curiosity, what kind of app is this for?"

[Hurricane Dancing @HDancing49123]

@tf_PrabaL hi

[Hurricane Dancing @HDancing49123]

I’m actively trying to break this cycle. No more letting AI bloat my ideas. Time to touch grass and ship real things.
If you’re stuck in this same loop, let’s talk. Let's hold each other accountable. Drop a 🤝 or DM me.

[Hurricane Dancing @HDancing49123]

When you treat AI as an external organ and the friction of building drops to zero—that is the exact moment everything starts falling apart.
A harsh truth I recently realized about the "vibe coding" trap. 🧵👇

[Hurricane Dancing @HDancing49123]

You get trapped in an endless loop of "paradigm shifts" and AI-fueled scope bloat. The project gets infinitely iterated, but never sees the light of day. Never gets used by real humans.
It's just phantom progress.

[Hurricane Dancing @HDancing49123]

The worst part? A total dopamine hijack.
All the pleasure comes from the moment a new idea pops up in your chat window. The thrill of thinking completely overpowers the grit and joy of actually shipping.
Everything distorts.