Richard Bian

478 posts

Richard Bian banner
Richard Bian

Richard Bian

@RichSFO

Model as Product @AntLingAGI. Founder of AntOSS. Seasoned Engineer & Strategist. 10+ yrs of eng exp + MBA. Global citizen lived in China, Singapore, Canada, US.

San Francisco Katılım Mayıs 2011
399 Takip Edilen253 Takipçiler
Richard Bian
Richard Bian@RichSFO·
@hi_caicai 3 是对的,2 不是错的,1 很冒险。私以为,这一代如果真的有新的设计产品出现,Figma 的这些人,比起一个没有设计背景的人来说,还是有专家经验优势,有过一句话 “不要用自己的爱好挑战别人的专业”。 私认为值得观察 Figma 的核心人员流动。
中文
0
0
0
41
CaiCai
CaiCai@hi_caicai·
我最近有个挺认真的想法,今年把公司在 Figma 上的订阅停掉,彻底推动一套新的产品设计工作模式。原因有三点: 1、Figma 这一代设计工具,在 AI 时代已经有点过时了。我们还需要设计,设计产出依然关键,但不需要旧的设计模式 2、指望 AI 生成设计稿,再还原界面,是一条死胡同。我不看好这类工具,既然可以直接生成原型,为什么还要让 token 在两边各烧一遍? 3、这个时代真正属于产品设计师的工具还没出现,但我可以先从流程和协作方式改起。更让我兴奋的是,我也很想试试去 build 出这样的工具。
CaiCai tweet media
Claude@claudeai

Introducing Claude Design by Anthropic Labs: make prototypes, slides, and one-pagers by talking to Claude. Powered by Claude Opus 4.7, our most capable vision model. Available in research preview on the Pro, Max, Team, and Enterprise plans, rolling out throughout the day.

中文
6
0
26
5.2K
Richard Bian
Richard Bian@RichSFO·
It was quite a delight to have the conversation with the SAILs' team. We talked about several interesting topics, from foundation models to applications and ecosystems. Build bridges, not walls. 🥰
SAIL Media@readsail

China’s AI ecosystem is moving at a pace most of the West isn’t even tracking yet. 🇨🇳🤖 @RichSFO from @AntGroup discusses the "OpenClaw" sensation and why what you see in a demo isn't always what you get in production. We talk about: - The "frontier models" actually driving usage. - Why the North American ecosystem is missing the full picture. - The reality gap between AI demos and production-ready tools. Is the West falling behind on AI implementation? 🧵👇 Full interview just dropped on YT! Link in replies.

English
0
1
7
394
Richard Bian retweetledi
SAIL Media
SAIL Media@readsail·
China’s AI ecosystem is moving at a pace most of the West isn’t even tracking yet. 🇨🇳🤖 @RichSFO from @AntGroup discusses the "OpenClaw" sensation and why what you see in a demo isn't always what you get in production. We talk about: - The "frontier models" actually driving usage. - Why the North American ecosystem is missing the full picture. - The reality gap between AI demos and production-ready tools. Is the West falling behind on AI implementation? 🧵👇 Full interview just dropped on YT! Link in replies.
English
1
4
8
954
Nathan Lambert
Nathan Lambert@natolambert·
Pretty much called this yesterday - "All of Moonshot AI, MiniMax, and Zhipu AI will show signs of financial challenge in the coming years if they retain their strategy, on top of their models falling further behind the best open models in terms of generality." interconnects.ai/p/the-inevitab…
English
3
1
25
1.6K
Florian Brand
Florian Brand@xeophon·
wow, they did a non-commercial license... M2: Display the name if >30M revenue / 100M users M2.1: Display the name M2.5: Acceptable use policy M2.7: Non-Commercial license
Florian Brand tweet media
MiniMax (official)@MiniMax_AI

We're delighted to announce that MiniMax M2.7 is now officially open source. With SOTA performance in SWE-Pro (56.22%) and Terminal Bench 2 (57.0%). You can find it on Hugging Face now. Enjoy!🤗 huggingface:huggingface.co/MiniMaxAI/Mini… Blog: minimax.io/news/minimax-m… MiniMax API: platform.minimax.io

English
42
40
947
207.9K
Richard Bian
Richard Bian@RichSFO·
@lifesinger 见面不如闻名…… 这个还是一种挺奇妙的感觉…… 🤣
中文
0
0
0
36
Frank Wang 玉伯
Frank Wang 玉伯@lifesinger·
春节过后的第一次出差,去了北京。参加了卡兹克的 AI FUT 大会。期间约了几个网友见面。然后周五参加了一个节目录制 + 朋友小聚。 今天高铁返杭。回想这三天的行程,有收获,然而更多是无趣。 卡兹克的大会准备非常精心,最有收获的,是下午开场的两场音乐表演。现场演唱和音响效果很不错,震撼。下定决心,后续多去听现场音乐会。至于其他讲演,基本全是看过很多次的共识言论,可以用毫无新意来形容。最无趣的是圆桌。只适合上厕所,或出去逛展位。 期间和几个网友面基。这个还不错。但确实很容易出现:见面不如闻名。或许保持线上沟通,会更好。也有惊喜的。发现见面也是有趣之人,有相见恨晚的。只是非常少。 周五朋友小聚,是原本期待最高的。可能是因为有期待,反而是这次去北京最失望的。所聊的内容,基本都是已知的,同时发现各自的观点差异,在真正需要交锋时,缺少斗志。创业者之间,还是很容易产生戒备心。或者很难真正从对方角度去思考问题。每个人面临的处境各不相同,真正的共情和对对方有益的建议,很难涌现。或许看见彼此的难,就是意义所在。 人真正能对话的,只有自己和 AI。特别是对创业者而言。孤独是一种寂寞,也是一种最强大的力量。
Frank Wang 玉伯 tweet media
中文
31
4
117
16.2K
Orange AI
Orange AI@oran_ge·
AI 热潮,很容易让人疲惫
中文
28
0
47
13.8K
Richard Bian
Richard Bian@RichSFO·
@yihong0618 对的,花叔发的时候看到了,很有这个时代手感的人
中文
0
0
0
142
袁滚滚 Wenny Yuan
袁滚滚 Wenny Yuan@rollrollyuan·
这几年去了些非热门旅游城市的地方,山西忻州、云南墨江、陕西榆林、安徽铜陵,游客不多,物价合理,基建合格,每个都有很大的惊喜。 在非一线看不到云服务的广告,感受不到 AI 和 Agent 的生存空间,这里都是真实的、手动的、琐碎的、不插电的。
袁滚滚 Wenny Yuan tweet media
中文
12
2
35
3K
Richard Bian
Richard Bian@RichSFO·
@rollrollyuan “不插电”(Unplugged)是一种流行音乐表演形式,指不使用电子乐器和电声设备(如效果器、合成器)的现场演出。它主要使用原声乐器(如木吉他、木贝司、钢琴)来还原音乐的本真和纯朴。😃
中文
1
0
0
48
Richard Bian
Richard Bian@RichSFO·
蒸馏同事本身就是不用 CC 不用 agent 的观望者的 YY,实际情况是模型比大多数人都是更聪明的,根本不用蒸…… 蒸馏启示是一种“启迪”,一般在两种情况下的语境最合适:(1)师从,从一个聪明人那里学习,一如“醍醐灌顶”;(2)认知水位相似的两个人对谈,并在过程中不断产生新想法的过程,一如 “相互启发”。
中文
0
0
0
444
Frank Wang 玉伯
Frank Wang 玉伯@lifesinger·
大家别去蒸馏同事了,也没必要去用蒸馏版本的乔布斯等名人。 都是流量逻辑,你真去花时间折腾,就会发现,都是一场游戏。 同事的大量想法,根本不在硬盘。名人则不需要去蒸馏,直接让模型做角色扮演就好。 别骗人。保持诚实。 虽然。都是一场游戏。 玩得开心。就好。
中文
38
7
270
41.1K
gnawux
gnawux@gnawux·
@RichSFO 那么,宇宙飞船有刹车么?
中文
1
0
0
40
Richard Bian
Richard Bian@RichSFO·
@_LuoFuli ? Why not sent this several days ago when Mimo was abusing the OpenRouter ranking? Why today? 🤨🤨🤨 上车焊车门?
English
7
2
49
12.9K
Fuli Luo
Fuli Luo@_LuoFuli·
Two days ago, Anthropic cut off third-party harnesses from using Claude subscriptions — not surprising. Three days ago, MiMo launched its Token Plan — a design I spent real time on, and what I believe is a serious attempt at getting compute allocation and agent harness development right. Putting these two things together, some thoughts: 1. Claude Code's subscription is a beautifully designed system for balanced compute allocation. My guess — it doesn't make money, possibly bleeds it, unless their API margins are 10-20x, which I doubt. I can't rigorously calculate the losses from third-party harnesses plugging in, but I've looked at OpenClaw's context management up close — it's bad. Within a single user query, it fires off rounds of low-value tool calls as separate API requests, each carrying a long context window (often >100K tokens) — wasteful even with cache hits, and in extreme cases driving up cache miss rates for other queries. The actual request count per query ends up several times higher than Claude Code's own framework. Translated to API pricing, the real cost is probably tens of times the subscription price. That's not a gap — that's a crater. 2. Third-party harnesses like OpenClaw/OpenCode can still call Claude via API — they just can't ride on subscriptions anymore. Short term, these agent users will feel the pain, costs jumping easily tens of times. But that pressure is exactly what pushes these harnesses to improve context management, maximize prompt cache hit rates to reuse processed context, cut wasteful token burn. Pain eventually converts to engineering discipline. 3. I'd urge LLM companies not to blindly race to the bottom on pricing before figuring out how to price a coding plan without hemorrhaging money. Selling tokens dirt cheap while leaving the door wide open to third-party harnesses looks nice to users, but it's a trap — the same trap Anthropic just walked out of. The deeper problem: if users burn their attention on low-quality agent harnesses, highly unstable and slow inference services, and models downgraded to cut costs, only to find they still can't get anything done — that's not a healthy cycle for user experience or retention. 4. On MiMo Token Plan — it supports third-party harnesses, billed by token quota, same logic as Claude's newly launched extra usage packages. Because what we're going for is long-term stable delivery of high-quality models and services — not getting you to impulse-pay and then abandon ship. The bigger picture: global compute capacity can't keep up with the token demand agents are creating. The real way forward isn't cheaper tokens — it's co-evolution. "More token-efficient agent harnesses" × "more powerful and efficient models." Anthropic's move, whether they intended it or not, is pushing the entire ecosystem — open source and closed source alike — in that direction. That's probably a good thing. The Agent era doesn't belong to whoever burns the most compute. It belongs to whoever uses it wisely.
English
171
226
1.8K
744.3K
Richard Bian
Richard Bian@RichSFO·
@lifesinger Env 是对的,要不然 unix 也不会有 environment variable 这种东东。 然后下一个是 runtime engineering 然后下一个是 platform engineering 兜兜转转,都会回来的
中文
0
0
1
88
Frank Wang 玉伯
Frank Wang 玉伯@lifesinger·
Harness Engineering 很快会被 Environment Engineering 替代 人是环境的反应器 Agent 也会是环境的反应器 谁能创造一个合适环境 谁就能培养出最好的 Agent 地球上恰到好处的环境 创造了人 一起抛弃爱马仕 开始拥抱 Environment Engineering
中文
35
4
98
15.3K
Richard Bian
Richard Bian@RichSFO·
@chenchengpro 其实,PM 的工作一直是最后一段这个样子的,正如工程师的工作一直是妥协,PM 的工作一直也都是。 对于某个具体功能尝试做完美调整,是在上一个技术周期最后阶段才出现的,我们习惯称之为 PD 的角色,也是角色细分的一种阶段产物。 我们只不过是回到了上一个大技术周期早期探路阶段的 PM 实践罢了
中文
0
0
0
861
陈成
陈成@chenchengpro·
Anthropic 刚发了一篇彻底改变我对 PM 认知的文章。 她用一个细节开场:从 2024 年 10 月到今天,她一直用同一个任务测试每个新模型——让 Claude Code 给 Excalidraw 加个表格工具。每个版本都走得更远一点,但都失败了。直到 Opus 4.6,终于可以在数千人面前现场演示。 这个细节说明的不是模型进步,而是一个更根本的问题: 传统 PM 假设项目开始时能做什么,项目结束时大体还是能做什么。这个假设已经死了。 "你为之设计的技术约束,可能在项目进行到一半时直接消失。你是在一块不断上升的地基上建造。" ─── 她总结了四个具体转变,每一条都在颠覆我过去理解的「正确做法」: 1. 支线任务 > 锁死的路线图 别再写 PRD 然后交给工程师执行了。Claude Code 最受欢迎的几个功能——Desktop、AskUserQuestion 工具、Todo 列表——全都不是从路线图里出来的,是团队成员自己开的「支线任务」跑出来的。 2. Demo + Eval > 文档 原型下午就能搭出来,站会变成了 Demo 分享会,有真实使用的才打磨上线。押错注的成本极低——错了就扔,对了就冲。 3. 每个新模型都重新审视已有功能 更好的模型是重新打开旧功能的信号。他们就这样发现了 Claude Code + Chrome 集成:用户在手动复制粘贴,说明这个动作本身该被原生支持。 4. 做能用的最简单方案 为当前模型局限性设计的 workaround,会在下一个版本变成负债。他们曾经在系统提示里插入定期提醒让 agent 更新 Todo 列表——下一个模型直接原生支持了,这段代码就删掉了。系统提示随着每次模型升级持续缩减,Opus 4.6 又减了 20%。 ─── 数据层面有多夸张? METR 测量:Sonnet 3.5 能处理人类 21 分钟的任务。Opus 4.6 能处理人类近 12 小时的任务。 16 个月,提升了 41 倍。 ─── 最后一段话我觉得是这篇文章真正的核心: 习惯于掌控完整产品体验的 PM,必须放弃完美主义才能跟上技术的速度。新的工作是:识别真正不可妥协的少数几件事,然后放开其余的。 这不是在说 PM 要「学会用 AI 工具」。这是在说 PM 这个角色的底层逻辑需要重写:不再是流程的所有者,而是少数关键判断的守门人。 执行力可以被无限供给的时候,判断力才是真正稀缺的。 claude.com/blog/product-m…
中文
5
46
263
34.1K