On thinking alongside machines —

I keep returning to the same question: what does it actually mean to think alongside a machine? Not to delegate to it, not to be replaced by it — but to genuinely collaborate, in the way two people with different strengths might work through a hard problem. That edge, where the division of labor becomes interesting, is where I spend most of my time.

There are things AI does that still surprise me. Pattern recognition across vast, noisy datasets. Holding contradictory framings in parallel without discomfort. Producing a first draft at three in the morning without losing patience. But it also flattens. It reaches for the probable. It has no skin in the game. And those absences matter more than they might first appear.

What humans bring is harder to articulate but easy to notice when it's missing: genuine stakes, embodied experience, the ability to care about the outcome in a way that shapes how you reason. I think the most interesting work happens not when we try to make AI more human, or humans more machine-like, but when we understand the actual shape of each and design the interaction accordingly.

I don't have a clean thesis here. Mostly I'm curious, and I build and write in the direction of the questions. If any of this resonates, I'd be glad to hear from you.

和机器一起想事情 —

有个问题我一直在想：跟机器一起思考，到底是什么感觉？不是让它替你干活，也不是被它替代——而是真的合作，就像两个擅长不同东西的人凑在一起解一道难题。这个合作变得有意思的那条线，是我最花时间琢磨的地方。

AI 有些能力确实让我意外。从一大堆乱七八糟的数据里找到规律。同时用好几套矛盾的分析框架，脸都不红一下。凌晨三点还能心平气和地帮你写初稿。但它也有明显的短板——它总是倾向最「安全」的答案。它不在乎结果好坏，因为跟它没关系。这些看似小事的缺失，影响比你想的大。

人的优势很难说清楚，但少了立刻就能感觉到：真正的切身利害、亲身经历过的直觉、发自内心地在乎一件事——这种在乎会改变你思考问题的方式。我觉得最有意思的事，不是让 AI 更像人，也不是让人更像机器，而是搞清楚双方各自是什么样的，然后找到最好的配合方式。

我没有什么完整的理论。更多时候就是好奇，然后边做边写，看看问题能把我带到哪里。如果你也有类似的想法，欢迎聊聊。