Benjamin Pan
425 posts

Benjamin Pan
@panmingming
engineer, businessman 時間是幻覺,創造乃對抗熵增的唯一手段。 time is delusion, building things is the sole means to defy the entropy increase.

Unpopular opinion: "AI makes everyone a developer" is true the same way "cameras makes everyone a photographer"


A conversation I've had several times lately: Them: "LLMs are so good for learning whatever you want — there's never been a better time to learn!" Me: "When was the last time you learned a new domain with LLMs?" Them: "Oh, there's no point. LLMs do a better job than me already"

现在有一股 AI 攀比风:你要尽可能多的往高里说 AI 提升了你多少倍工作效率,这样你会显得很时髦,比其他人反应快,跟得上潮流,更能应对变化;否则你会显得古板,同社会和趋势脱节,不善于面对变化;这股攀比风,从公司到个人都存在,越是本职工作和主业做不好的,越期望自己显得更加适应潮流。

最近在弄公司做系統的事,我覺得資訊工程師是一個很有趣的領域,就是做IT很需要溝通跟理解,不然很容易做出不食人間煙火的東西,因爲我們用系統其實就是建構在資訊工程師想像出的世界當中,可是寫coding的能力,發展的方向又跟這個靈活的能力完全相反,就是你在IT這領域如果要做到卓越的話,你是需要具備兩個自相矛盾的能力,個人覺得還蠻有趣的。說到這個有一次臉書建議我加一個超級久沒聯絡的朋友為好友,我想說好吧,那我就按下去,結果他跳出一個訊息寫說請確認你有認識對方再加,然後就不給我加,我???


Here’s the result from LinkedIn translator: I quit xAI after three months. It was a total nightmare. I’ve worked in some high-pressure spots, but this place was just dysfunctional and exhausting. I like to think I’m a "problem solver," but the truth is I was just grinding in the trenches for a mission that didn't actually matter to me. I realized pretty quickly that I hated the culture and it was a terrible fit. It’s embarrassing to admit because I bought into the Elon hype, and I’m sure they’ll do fine, but I couldn't stand being there another day. If I learned one thing: don't stay at a company just because the name looks good on your resume. If you're miserable and not getting anywhere, just quit. Your sanity is more important than a brand. What’s next? I’m going to sleep for a month because I’m burnt out. Then I’ll start looking for another job that I can hopefully tolerate. I’m still trying to stay relevant and network, mostly because I need to find my next paycheck. I'll probably start some random projects to keep myself busy.

This is the productivity trap nobody talks about. AI removes friction from execution, so you say yes to everything because now "it's easy." But the bottleneck was never execution speed. It was knowing which 20% of tasks actually move the needle. Without that clarity, 10x speed just means 10x busywork.

最近一个月,不是我销声匿迹了,是AI编程的时代已经结束了。 工程师的架构经验、巧思、前瞻、代码坏味道直觉,以及最重要的——在软件大方向上的定调和指挥的价值,正在呈断崖式下跌。 最近一个月,最常用的提示词是:顺着你的新思路走。因为在接收到大量启发式反问之后,我越来越感到,"爹味先验"和"自信的人择",正在限制AI的能力。 最近一个月,越来越多地看到人们轻而易举地制作出自己想要的工具,一夜之间,一小时,不再是一个可以被嘲讽的自媒体噱头。人人都自恋,使用自己的工具,打磨自己的工具,这件事情开始像短视频一样杀时间。 最近一个月,我开始意识到,软件诞生所需的时间,和它的生命周期一样,越来越短暂。接下来的日子,软件不需要被迭代,只需要被不断生成。所有的版本都是临时版本,就连最终版本也是"重新生成",而非迭代而来。 当"重构成本",这个过往使软件必须始终如一地发展下去的引力基石,在成本上被瓦解。意味着AI开始像霍金辐射一样蒸发着人类在软件工程上的一切经验所凝固而成的黑洞。一切都变得柔软而不再重要了。 人们没有那么多时间再陪着AI闹了,他已经开始满场跑,不停地扣篮。在短短一年时间里,人先是在场上,让AI给他传球。后来,把球传给AI。 再后来,人心想,做教练。 教练做了两个月,人回到了观众席。坐在观众席,背后凉凉的,那是因为背后就是球馆通道,吹来了阵阵的冷风。 走在江边,江水,草地,人群,变得比往日更加鲜活,动人。他们是唯一不可以生成的东西了。


在日本、台湾、欧洲旅行,从来没感觉受到本地人歧视。在美国生活,也没感觉受到本地人歧视。在美国、欧洲遇到的不愉快,几乎都是发生在说普通话的中国人开的店。可能算是一种中国特色的歧视:粗鲁、没教养、没礼貌,而且专门针对中国顾客。

Marc Andreessen believes that true innovators share 5 specific personality traits that allow them to do groundbreaking work. First, they have high openness, so they are always ready to hear new ideas from any field. But being open is not enough if you never build anything, so they also need high conscientiousness to work hard for many years. It is actually quite rare to find both traits in 1 person because they usually contradict each other. These people are also disagreeable so they do not quit when others call their ideas stupid. To tie it all together, they need a high Intelligence Quotient to handle information and low neuroticism so they do not break under the pressure. --- Video from 'Andrew Huberman' (@hubermanlab ) YT channel (link in comment)
