Sabitlenmiş Tweet
Mich_Kalek
1.3K posts
创业了一段时间，再回头看毛泽东，总有新的感触。
经常感叹自己没有出生在美国，不像硅谷小天才一样开局就能拿到美国顶级的融资，开局就有明星梦之队。
反观毛泽东当年，国都没出过；对手蒋介石拿到了最顶级的资源支持，有更高的正统性，黄浦军校名校毕业且光环拉满，但依然输得一败涂地。
最好的心法就是《毛泽东选集》，最值得借鉴的人就是毛泽东。上山当土匪的时候想过怎么和军阀打仗吗
你想象一下，这就仿佛你创业做一个项目，从头都是bootstrap，融资都拿不到，网站设计的像是大学生画的ppt，产品只是勉强能跑，还得经常去各种生态蹭点黑客松的grant维持生计
却最终干翻了 a16z 加 Paradigm 加一众顶级 VC第一轮就估值3B，融了50M的大局
世上无难事 只要肯登攀
中文
@teortaxesTex exactly, doing some alpha fold thing and yet cannot deliver frontier model
English
Sad. Playing GDM without GDM's compute is impossible
I doubt it though, their video generation is good enough to be made profitable
Xiuyu Li@sheriyuo
ByteDance AI Panic: Internal Meeting Just Axed 30% of Projects “Accelerating AI investment” vs Cut “Cat Box”, “Star Painting”, Dreamina lines. All non-Doubao products FAILED. Tens of billions burned on AI video/writing/education. 2025 AI inference cost: 8B CNY = 2.3× revenue growth. CFO warning: “This burn rate means no second Doubao. Cash flow dies before 2027.” Overseas nightmare: TikTok US drama + EU AI Act + India ban. Dreamina MoM growth crashed from 30% → 4%. Strategic retreat: All-in on Doubao + PICO/AI glasses. Pure apps? DEAD. First time a Chinese tech giant publicly ABANDONS multi-line AI war. The great AI shakeout is here? 👀 #ByteDance #AI #TikTok
English
ok, using official api, it hit 98%.. wtf, the cost per million is 0.04
Mich_Kalek@KalekMich7668
using v4flash for some small project, cache hit rate over 97%, and the per million cost is 0.05 cny😰 it is a third party router the price tag is the same as the official but i don't know why the cache hit rate is so high...
English
using v4flash for some small project, cache hit rate over 97%, and the per million cost is 0.05 cny😰 it is a third party router the price tag is the same as the official but i don't know why the cache hit rate is so high...
English
use dsv4 pro to execute my weekly-report skill, it is way better than glm and kimi, only problem with dsv4 pro is no vision ability, have to borrow vision from gpt model
English
@CommandCodeAI love your separte interface of /context , but /compress /model etc slash commands after enter, the slash commands won't disappear
English
the status of deepseek was a signal of new release, now it disrupts my workflow...
OpenCode@opencode
OpenCode Go is currently experiencing issues with DeepSeek models due to an upstream provider outage. We’re monitoring the situation here: status.deepseek.com
English
Mich_Kalek retweetledi
@nahcrof model is quite enough, but yeah the stability issue, it just disrupts the flow
English
I'm working on some updates to the API and other things relating to stability at the moment
I'm sorry to everyone who's been impacted by the issues with stability lately and I plan on writing a blog that talks about this journey to the best of my ability
Thank you to everyone who's stuck around this long, and I hope I can continue to build and improve a service that you find worth using
English
@CommandCodeAI yeah, but apparently borrow tons of stuff from claude code, not saying it is bad,
English
@KalekMich7668 No. We are built completely different from scratch. First commit was 2020 July although most serious work started Aug last year.
English
it is probably a fork of leaked claude code
Mich_Kalek@KalekMich7668
@CommandCodeAI got taste😇
English
sometimes i just forget that i am the biggest brain btw me and the model at least in my problem space
English
@zephyr_z9 shit that is 20b cny, it is even bigger than most private hedge fund AUM
English
They are also getting a $3B capital injection from High Flyer
Andrew Curran@AndrewCurran_
DeepSeek could be valued at up to $50 billion after its first funding round, in which it is expected raise as much as $4 billion.
English
why all the sudden "sunsetting" becomes viral?
Kilo@kilocode
Roo Code is sunsetting May 15th. If you need a new agent, here's a 2-minute walkthrough showing exactly how to move to Kilo. Settings, custom modes, model setup, all of it. We're grateful for what the Roo team built. Our door is open. kilo.ai/roo-migration
English
@iamai_omni agi power token output money is strip out of the cycle
English
完全正确。基本只要脱离了 AI 前沿，就会感觉泡沫太大了。
这是技术奇点附近特有的现象：你看到的是革命还是泡沫，取决于你是在事件视界之内，还是之外。
AI 的确可能是历次技术革命里少见的、直接提升生产力的技术。但资产价格交易的不是“有没有用”，而是“未来多少年的生产力提升已经被提前贴现”。
所以真正的问题不是 AI 有没有泡沫，而是：哪些资产处在奇点内，哪些资产只是被奇点的光照到了。
Tracy@CTracy0803
我身边的打工人朋友们都觉得现在AI的泡沫越来越大 而有意思的是，几个创业者朋友却认为AI不会有泡沫 原因是AI是历次技术革命里唯一一次只提升生产力，不提升需求的
中文
@teortaxesTex RUC in China has a very strange tier though.
Top2: pku, thu
Generalized Top"3?": {sjtu, fdu, zju, nju, ustc}, ruc, cas
Here’s a ridiculous ranking.
English
Very intense story, but I guess congrats on getting into Renmin
didn't know this about them, by the way
Xiuyu Li@sheriyuo
不得不提我志愿填报的时候，家里人想让我报北航 or ZJU 的 CS。我拿着菜刀跟他们对峙死也别改我志愿，我第一志愿就填的 RUC，第二志愿乱填的 PKU（ 大学四年结束了，我从没回过家，父母来北京找我也没见到我，只见到了我的急诊洗胃报告单，于是他们也知道了我是真的要以死相逼 我不想收家里的生活费，但是他们硬要塞给我，也给过我几个月的房租，或许就是他们剩下的唯一良知了吧。对我而言我的父母就是我的最大问题，小学老师的父亲、最低收入的母亲、恨铁不成钢的清北苗子 我高中最好的一位竞赛同学 & 好朋友知道我的这些事情，她的家长觉得我的父母做的相当不对，一定程度上也很可怜我的境遇，可惜的是这位朋友已经和我断交了 反观我那几十位去了清北的同学，我愈发感受到家学/家境是如何塑造一个孩子的，又是如何养废一个孩子的。我今年 21，却早早在未成年的时候就开始在社会自己谋生，或许承受了许多我这个年纪不该承受的东西 我跟我的导师聊人生、聊爱好，我的兴趣爱好已经跟一个 三四十的中年人差不多了，对于感情、婚姻的态度也是，尤其是我永远不可能再把我的不幸再带给下一辈了。这种事情是永远不可能和解的，得到的结果反而可能是我爸一刀把我捅死，我早就想开了 于是我跟我的导师全部说出了这些事情：为什么我想润、为什么我留在国内是慢性死亡、以及我的心病，难以想象这件事情对他的伤害有多大。这件事情的一个星期以内我想到他我就会哭，和朋友散了很久的步我才慢慢走出来，大家都很痛苦吧 家长为什么总是期待拿棍棒、体罚教育孩子，孩子就一定会在数十年后哭着理解这种中国式教育的正确性，并学会用同样的方法教育自己的孩子呢？
English