Stephen Chan

629 posts

Stephen Chan banner
Stephen Chan

Stephen Chan

@lightencc

AI • LLM • Agent | Long-term thinking | Ultra runner | build, optimize, explore | 追求效率与深度 | 热爱数据、旅行与户外 | 知行合一

Shanghai เข้าร่วม Temmuz 2018
413 กำลังติดตาม74 ผู้ติดตาม
Stephen Chan
Stephen Chan@lightencc·
@anorth_chen 虽然我没看他写了什么,但是我看了文章附的全文翻译,的确是我努力的目标
中文
0
0
1
122
North@CreaoAI
North@CreaoAI@anorth_chen·
悲观者正确,乐观者前行。 宝玉老师提出了很多条质疑的点,本着严谨交流的态度,我不介意一一回复。 这篇文章内容涵盖了大量软件工程,因为我们希望把自己AI First如何实践落地的理念分享出来,而不仅仅是一些形而上concept的内容。至于其他AI First的实践,由于篇幅和文章重点无法涵盖所有。 第一,AI提交代码是不需要全部人工回归一遍的。我们会拆细每次PR的影响范围,基本都只会涉及某个功能模块,而不是大范围的修改,这在手写代码时也是基本的软件开发协作规范。 在符合软件工程协作规范下的情况下,AI提交的代码修改完全可以被自动化测试覆盖,不需要担心它会搞崩别的功能。 第二,我们中间的测试/审查/发布确实做到了全部自动化跑通了,所以我们做到了今天这个迭代效率。 第三,我们每次A/B测试的线上监控基础设施也都是完善的,有充足的数据支撑我们做判断。我建议你学习了解下statsig。 第四,你为什么会把大而模糊的任务丢给AI?这个问题非常奇怪,我相信如果你做过管理,也不会把大而模糊,自己都没想清楚的任务丢给员工吧。 第五,系统架构的设计是任何软件工程团队的基本功了,拿出这一条来悲观质疑真的很像在抬杠。 结论是,我们全都做到了。 关于你提到的claude code和codex团队是否有这么搞的问题,事实上我们就是观察到了claude团队极其夸张的迭代效率,以及OpenAI工程团队在今年二月份的分享得到的灵感:openai.com/index/harness-… 你觉得他们有没有也在用这一套呢?为什么你如此笃定AI交付和功能质量必然在对立面? 我们的分享来自于团队脚踏实地实践后的经验,关于我们目前做到了怎样的迭代效率,请看产品changelog:docs.creao.ai/community-and-…
宝玉@dotey

今天刷到这篇文章几次,说点不一样的。与其说 AI First,不如说软件工程 First。 这篇文章看着在讲 AI,底下全是软件工程。 抛开后面讲组织和人的部分,原文前半段的重点简单总结一下: AI 时代,人成了瓶颈。PM 花几周做需求,AI 两小时就能实现,PM 成了瓶颈。QA 测三天,AI 写代码只要两小时,QA 成了瓶颈。团队 25 个人,对手几百人,人力也是瓶颈。 怎么办?把人从链条里拿掉。AI 写代码、AI 审查代码、AI 跑测试、AI 部署上线、AI 监控线上状态,出了问题自动回滚。每天定时扫描日志,自动发现问题、分配任务、跟踪修复。整条流水线跑起来,人只需要在关键节点做判断。 至于文中提到的统一代码库,锦上添花,和 AI First 关系不大。有当然更好,没有也有很多替代方案。 整套方案听下来,逻辑自洽,效果也漂亮:一天部署好几次,功能当天上当天撤,数据说了算。 但先别急着照搬,先对照自己的情况想几件事: 第一,自动化测试。AI 改完代码,你得有办法确认它没搞崩别的功能。测试覆盖不够的话,每次 AI 提交代码你都得人工回归一遍,那速度根本快不起来。 第二,CI/CD 流程。从提交代码到部署上线,中间的测试、审查、发布、回滚,是不是全自动跑通了?这条流水线不通,AI 写得再快,代码也堆在那儿等人手动处理。 第三,A/B 测试和线上监控。新功能上线之后效果好不好,得有数据说话,效果不好得能随时关掉。没有这套机制,AI 一天产出五个功能,你都不知道哪个该留哪个该砍。 第四,任务管理。任务得拆到合适的粒度,生命周期得跟踪得住。一个大而模糊的任务丢给 AI,现在的能力还啃不动。多个 Agent 同时干活的时候,谁做哪个、哪个优先、做到什么程度,这些都得有地方管。 第五,系统架构。架构太乱或者压根没有架构的代码,AI 维护起来跟人一样头疼。上下文塞满了还是搞不清边界在哪,改一处崩三处。 这几条里如果有做不到的,就得靠人去补。补不上,AI First 就只是一句口号。 但假设你全做到了,就能 AI First 了? 还是不行。这套玩法只适合一部分场景。 什么场景适合?后端逻辑为主、界面不复杂的产品,比如 API 服务、数据处理平台、内部工具。功能好不好,跑一下数据就知道,不需要人去盯着每个像素。原文里的就是个 Agent 平台,本质上是后端驱动的产品,可以用这套打法。 再比如早期产品快速试错,功能上了不行就撤,用户预期本来就没那么高,AI 的速度优势能充分发挥。 但很多场景玩不转。 比如 UI 密集的产品。自媒体天天喊前端已死,但你让 AI 做个复杂界面试试,各种易用性问题、交互细节、视觉还原,它搞不定的。否则马斯克靠 AI 早就改了不知道改版 X 多少次了。 比如对功能质量敏感的产品。Anthropic 和 OpenAI 不知道 AI First 吗?他们敢在 Claude Code 和 Codex 上这么搞吗?让 AI 全自动迭代自家的核心产品,用户不骂死才怪。 再比如安全性要求高的场景,银行系统、在线交易平台,AI 代码出个差错,那可不是回滚能解决的。 AI First 的方向没有错,它代表的是一种意识的转变:每做一个决策的时候,想一想这件事能不能让 AI 来做,如果不能,缺什么条件,怎么把条件补上。 但这种意识要落地,靠的不仅是买几个 AI 工具的订阅,还需要把基础搭好。测试、CI/CD、监控、架构、任务管理,这些做扎实了,AI 的能力自然能释放出来。做不好,加再多 AI 也是在沙子上盖楼。 从这个角度看,AI First 的终点未必是让 AI 干所有的活,而是借着这股力量,把你一直想做但没动力做的工程改进,真正推动起来。 仰望星空是好的,但也还要脚踏实地。

中文
8
3
75
15.7K
Stephen Chan
Stephen Chan@lightencc·
@0b7cm 你脑子里想的和说的是一回事么
中文
0
0
0
7
草莓🍓女の子
爱一个人 好好爱 别总想着脱了她的衣服
草莓🍓女の子 tweet media
中文
5
0
11
1.7K
Pablo
Pablo@pblmnz·
How I hacked my timeline, and the Chinese AI Culture: Yesterday, a post from @dontbesilent appeared in my feed. He was wondering why foreigners who can't read chinese were suddenly answering to his tweets (that was me), and if we were using auto-translate. I replied with the reality: long before X had the auto-translate feature enabled, I realized the chinese content was way better than the english feed. So, I started manually copying and translating their articles. I intentionally trained my algorithm to feed me their posts because the insights were just better (imo). That one reply sparked a massive back and forth yesterday with the chinese community in the comments. The cultural exchange was incredible, and it perfectly mirrored what i saw on my visit to China this last christmas. The Implementation Reality I spent over two weeks moving through the country: Shenzhen, passing by Xingping, Phoenix, Tianmen, Zhangjiajie (prob the most beautiful place on earth), Xi'An, Shanghai, Beijing, Guangzhou, and back home. Everyone in the west talks about ai, fintech and implementation, but China is years ahead. To give you an idea of the scale: We were driving through a random, remote village somewhere between phoenix and zhangjiajie. Literally the middle of nowhere. We stopped at a crosswalk by a dirt roundabout, and there was a guy selling fruit in a cartwheel. Even this remote dirt-road fruit vendor had an alipay & wechat qr code. We paid with our phones (and generously tipped him bc we couldn't believe he could be in that remote place selling fruit). The entire country runs on a seamless digital infrastructure that makes the rest of the world look like its stuck in the past. There is no place on earth where there is such a general technology adoption. The Friction & The Physical Comedy My worst and only problem of the entire trip was the language barrier. What I love the most every time I travel, is mixing with locals. Understanding their mental models, exchanging ideas, knowing their culture and ways of thinking. From the smallest detail in their day to day life, to the way of working of the entire society of the country I am in. For the first time in my life, I couldn't do that. The friction was just too high (the younger generations are doing great though. Kids would constantly approach me just to practice a few english words). Also, being redhead in rural china is a really interesting experience: People would take sneaky photos every day, parents pointing me out to their kids, Chinese locals asking me to take photos with them, even a mother literally handed me her baby just to take a smiling photo with me. The people are incredibly welcoming, but the communication gap is (very sadly), too big right now. Thoughts after my trip I loved it so much, I seriously looked into moving there for a few months right before I moved to Portugal. Unfortunately, the visa requirements allowed me just to stay for 30 days as a tourist. Otherwise I would have to create a company, or get hired by one (none of them were options in my plan for now). But here is the absolute reality: at their current pace and work ethic, they are going to completely dominate the tech and ai meta. I have no doubts about this. I will definitely return at some point to live there for a few months and being able to dive really deep into AI, local models, hardware, and above all, the culture and people. But next time I am bringing a real time ai auto translate device with me, so I can actually talk to everyone over there ;) (and massive thanks to everyone on the chinese timeline who spent their day exchanging thoughts with me yesterday. the bridge is being built.)
Pablo tweet mediaPablo tweet mediaPablo tweet media
English
138
21
401
113.1K
戴姐
戴姐@chodaileyii·
向下兼容堕落很容易,坚守美德和好的思想才是难得。有一个好的长辈,能把正确的价值观传递给你是多么重要。
中文
8
215
821
67.7K
Stephen Chan
Stephen Chan@lightencc·
with or without agent
English
0
0
0
12
Stephen Chan รีทวีตแล้ว
Elon Musk
Elon Musk@elonmusk·
True currency is steadfast friendship
English
14.2K
26.9K
285.4K
77.5M
ImL1s
ImL1s@aa22396584·
@AYi_AInotes 有意思。企业端的AI认知升级往往比硅谷滞后一两年,但一旦开窍就会大规模推进。代理时代的真正门槛不是技术,而是企业愿不愿意重构流程——这才是最难啃的骨头。
中文
2
0
2
313
阿绎 AYi
阿绎 AYi@AYi_AInotes·
Box CEO Aaron Levie最近跑遍了银行、零售、医疗、科技等十几个行业,跟几十家公司的AI和IT负责人深聊完,得出的结论,把硅谷很多想当然的认知都推翻了。 所有人都达成了一个共识:AI已经彻底翻过了聊天时代的山,正式进入了代理时代。 现在的AI不再是只会回答问题的聊天框,还能调用工具、处理数据、执行真实业务流程的工作者。 企业的AI策略,也从以前的随便试,百花齐放,变成了精准瞄准特定工作流的深度自动化。 但最反直觉的一点来了: 几乎没有任何一家公司,在讨论用AI Agent取代岗位。 大家想的全是以前根本没资源、没精力、没优先级做的那些事,现在终于能做了。 比如把堆了十年的客户文档挖出来做新的业务洞察,比如自动化那些从来没人碰过的后台流程,比如把拖了好几年的系统升级做完,核心目标是赚钱不是省钱。 真正卡住所有人的,也根本不是模型不够强,是几十年攒下来的老旧系统,是散在几十个地方的碎片化数据,和固定到不能再固定的运营预算,以及整个公司从上到下的变革管理。 很多公司已经开始搞鲨鱼池式的算力预算竞标,各个业务线抢token资源,还有的在每个部门都设了专门的AI负责人,统一向中央团队汇报。 还有一个特别扎心的观察: AI不仅没让任何人少干活,现在所有人,包括硅谷的团队,都觉得比以前任何时候都更忙了,而且工程师不仅不会失业,反而变得更重要了。 因为硅谷以为AI会把所有事情都变简单,但真实的企业世界里,最强大的代理用法反而更技术化了。 MCP、Skills、CLI这些概念,对非技术人员来说还是天书,现在最缺的,已经不是会写代码的人,而是能把这些零散的AI能力,搭成一个稳定、可靠、能跑通的业务系统的人。 所以企业AI代理的落地,从来就不是一个模型能力的问题 它是集成、是治理、是预算、是流程、是人的综合工程。 模型只是最便宜的入场券,真正的战争从来都在模型之外。
Aaron Levie@levie

Another week on the road meeting with a couple dozen IT and AI leaders from large enterprises across banking, media, retail, healthcare, consulting, tech, and sports, to discuss agents in the enterprise. Some quick takeaways: * Clear that we’re moving from chat era of AI to agents that use tools, process data, and start to execute real work in the enterprise. Complementing this, enterprises are often evolving from “let a thousand flowers bloom” approach to adoption to targeted automation efforts applied to specific areas of work and workflow. * Change management still will remain one of the biggest topics for enterprises. Most workflows aren’t setup to just drop agents directly in, and enterprises will need a ton of help to drive these efforts (both internally and from partners). One company has a head of AI in every business unit that roles up to a central team, just to keep all the functions coordinated. * Tokenmaxxing! Most companies operate with very strict OpEx budgets get locked in for the year ahead, so they’re going through very real trade-off discussions right now on how to budget for tokens. One company recently had an idea for a “shark tank” style way of pitching for compute budget. Others are trying to figure out how to ration compute to the best use-cases internally through some hierarchy of needs (my words not theirs). * Fixing fragmented and legacy systems remain a huge priority right now. Most enterprises are dealing with decades of either on-prem systems or systems they moved to the cloud but that still haven’t been modernized in any meaningful way. This means agents can’t easily tap into these data sources in a unified way yet, so companies are focused on how they modernize these. * Most companies are *not* talking about replacing jobs due to agents. The major use-cases for agents are things that the company wasn’t able to do before or couldn’t prioritize. Software upgrades, automating back office processes that were constraining other workflows, processing large amounts of documents to get new business or client insights, and so on. More emphasis on ways to make money vs. cut costs. * Headless software dominated my conversations. Enterprises need to be able to ensure all of their software works across any set of agents they choose. They will kick out vendors that don’t make this technically or economically easy. * Clear sense that it can be hard to standardize on anything right now given how fast things are moving. Blessing and a curse of the innovation curve right now - no one wants to get stuck in a paradigm that locks them into the wrong architecture. One other result of this is that companies realize they’re in a multi-agent world, which means that interoperability becomes paramount across systems. * Unanimous sense that everyone is working more than ever before. AI is not causing anyone to do less work right now, and similar to Silicon Valley people feel their teams are the busiest they’ve ever been. One final meta observation not called out explicitly. It seems that despite Silicon Valley’s sense that AI has made hard things easy, the most powerful ways to use agents is more “technical” than prior eras of software. Skills, MCP, CLIs, etc. may be simple concepts for tech, but in the real world these are all esoteric concepts that will require technical people to help bring to life in the enterprise. This both means diffusion will take real work and time, but also everyone’s estimation of engineering jobs is totally off. Engineers may not be “writing” software, but they will certainly be the ones to setup and operate the systems that actually automate most work in the enterprise.

中文
7
21
105
25.7K
Stephen Chan
Stephen Chan@lightencc·
“I definitely expect in some of these areas ‘27 to be an important inflection point for certain things. … But I expect ‘27 to be a big year in which some of those shifts happen pretty profoundly.” The history and future of AI at Google, with Sundar Pichai youtu.be/bTA8sjgvA4c?si… via @YouTube
YouTube video
YouTube
English
0
0
0
19
Stephen Chan
Stephen Chan@lightencc·
模型只是入场券
阿绎 AYi@AYi_AInotes

Box CEO Aaron Levie最近跑遍了银行、零售、医疗、科技等十几个行业,跟几十家公司的AI和IT负责人深聊完,得出的结论,把硅谷很多想当然的认知都推翻了。 所有人都达成了一个共识:AI已经彻底翻过了聊天时代的山,正式进入了代理时代。 现在的AI不再是只会回答问题的聊天框,还能调用工具、处理数据、执行真实业务流程的工作者。 企业的AI策略,也从以前的随便试,百花齐放,变成了精准瞄准特定工作流的深度自动化。 但最反直觉的一点来了: 几乎没有任何一家公司,在讨论用AI Agent取代岗位。 大家想的全是以前根本没资源、没精力、没优先级做的那些事,现在终于能做了。 比如把堆了十年的客户文档挖出来做新的业务洞察,比如自动化那些从来没人碰过的后台流程,比如把拖了好几年的系统升级做完,核心目标是赚钱不是省钱。 真正卡住所有人的,也根本不是模型不够强,是几十年攒下来的老旧系统,是散在几十个地方的碎片化数据,和固定到不能再固定的运营预算,以及整个公司从上到下的变革管理。 很多公司已经开始搞鲨鱼池式的算力预算竞标,各个业务线抢token资源,还有的在每个部门都设了专门的AI负责人,统一向中央团队汇报。 还有一个特别扎心的观察: AI不仅没让任何人少干活,现在所有人,包括硅谷的团队,都觉得比以前任何时候都更忙了,而且工程师不仅不会失业,反而变得更重要了。 因为硅谷以为AI会把所有事情都变简单,但真实的企业世界里,最强大的代理用法反而更技术化了。 MCP、Skills、CLI这些概念,对非技术人员来说还是天书,现在最缺的,已经不是会写代码的人,而是能把这些零散的AI能力,搭成一个稳定、可靠、能跑通的业务系统的人。 所以企业AI代理的落地,从来就不是一个模型能力的问题 它是集成、是治理、是预算、是流程、是人的综合工程。 模型只是最便宜的入场券,真正的战争从来都在模型之外。

中文
0
0
1
16
Tansu Yegen
Tansu Yegen@TansuYegen·
Former BYD and Huawei engineers made a device that turns any bike electric, reaching 32 km/h ⚡
English
129
722
7K
1M
Stephen Chan
Stephen Chan@lightencc·
Nowadays, whenever I work on a project, I engage in real-time interaction with an agent; I strictly avoid looking at the underlying code implementation—focusing solely on the results—and instead impart my expertise to the AI.
English
0
0
0
10
WangNextDoor
WangNextDoor@WangNextDoor2·
川皇:你们就看我说让伊朗文明不复存在,这句话让他们来到了谈判桌上,你们忘了这么多年我还得一直听着伊朗叫嚣着让美国去死,让以色列去死,美国就是撒旦吗?
中文
1
0
1
458
MapleShaw
MapleShaw@msjiaozhu·
今日份商 K ?🤔 之你是那一年我在珠穆拉玛峰救过的九尾银狐😏 提示词👇 真人写实风格,竖屏9:16,10秒,第一人称视角, KTV包厢墙面大镜子前,暖金色壁灯+蓝紫激光背景, 0-3秒:镜头(即"我")从镜子里看见她【图片1】——她背对镜头站在镜前,正在整理头发,黑色露背上衣,镜中倒影和真实背影同框; 3-6秒:她从镜子里捕捉到视线,停下动作,和镜中的"我"对视,嘴角慢慢勾起,没有说话; 6-10秒:她缓缓转身,面朝镜头,Slow Dolly In轻微推近,她托腮歪头,似笑非笑,霓虹光打在颧骨和锁骨上形成高光,画面在她开口说话前定格。 光影:暖金壁灯正面光+蓝紫背景轮廓光(光源层),镜面反射形成真实与倒影双重光影(光行为层),暖金+冷蓝对撞(色调层)。 音效:高跟鞋转身时轻踩地板声,远处包厢隔墙传来的低频音乐,她轻吐气的细节音。 高清写实,禁止字幕水印
中文
10
22
229
31.4K
Yufan Sheng
Yufan Sheng@syhily·
老婆把 M1 的 MBP 给我了,内存是 16G!!!
Yufan Sheng tweet media
中文
19
0
68
19.2K
khushi.vy
khushi.vy@khushiirl·
Show me a mouse with zero haters I’ll go first:
khushi.vy tweet media
English
88
5
237
17.5K
Bread🍞
Bread🍞@himself65·
haiku 短诗 sonnet 十四行诗 opus 杰作 mythos 神话 下一个模型叫:永乐大典
中文
49
27
713
70.2K
Stephen Chan รีทวีตแล้ว
iPaul
iPaul@iPaulCanada·
科普:宇航员在太空站展示拧湿毛巾时水的形态变化。
中文
8
24
273
153.3K
iPaul
iPaul@iPaulCanada·
400万粉丝大V抱怨X太天真,不愿付钱却希望创作者留下来。。。 产品负责人Nikita在评论区说,你丫药店碧莲,复制黏贴 没有任何评论,就想挣钱?!
iPaul tweet media
中文
2
0
4
3K