Stefan Li (@autoconfig) header
Stefan Li (@autoconfig) profile picture

Building AI product at Notion. Co-founded https://xitter.gq/t.co/L7d7FeOI2r (YC W23). Previously at Facebook, WalmartLabs, Sony. 🇸🇪 🇭🇰 🇺🇸

NYC
Joined April 2010
Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)Photo by Stefan Li (@autoconfig)
This guy gets it
很多人似乎很难理解,为什么我,以及很多用之前用 Notion 的人,到目前为止 Notion 如此作妖都离不开 Notion。 这条帖子我会稍微多讲一点,Notion 为什么是当下所有 Agent 最好的上下文伴侣,远超 Google Drive、Obsidian 这些。 先说大家都知道的 Notion 是个直接对终端用户的产品, 因此它有一个完善的应用层。也就是,它的界面体验还不错。虽然很多人觉得 Notion 现在已经非常臃肿了,但你用过任何一款 Notion 的替代品就会发现,他们都在做 Notion 的子集。一旦深度使用,你就会发现他们缺乏某些 Notion 有但它没有的功能。 Notion 和当年的 Office 有点像——里面的每个按钮都对应着可能 5%-10% 的用户在用,没有任何功能是多余的, 区别只是你是否用得到。 然后,Notion 真正强大的是数据层。 其实早在 Notion 今年官方推出跨 Agent 记忆层之前,我就已经在用 Notion 做跨 Agent 记忆层了。 我在 Notion 里建了一个数据库,并设置了合理的增删改查规则, 然后把它连到我的 Codex、Workbuddy、Herme、DSH 以及任何支持 MCP 的 Agent 上,实现了跨 Agent 的记忆大一统。 我为什么用 Notion 做跨 Agent 记忆层,这就要说到我说 Notion 好用的第二点: 很少有人意识到,Notion 在数据层上其实是三层镜像——块结构、Markdown 文档和向量数据库, 并且三层镜像是秒级同步的。 当你接入 Notion 的 MCP 或做 Notion 的 API 开发的时候,会发现 Notion 可以对一个数据库进行字段筛选,可以进行关键词匹配搜索,还可以进行模糊搜索(实际是由 Notion 搜索 API 完成的向量搜索)。 取回数据的时候,可以精确的按字段取回,按块取回,还可以取回整个 Markdown 文件。 它几乎可以满足 Agent 对数据检索、存放和写入的所有需求。唯一的问题就是 API 并发不高,不能真的当对外的 App 的数据库来用,但对个人使用来说绰绰有余,绝对碰不到限速。 你要使用开源框架在自己的环境里实现这一点, 光是性能消耗就不小,更别提三个版本数据之间的冲突管理。 而更秒的是,其他类似的开源方案,都不会在提供如此复杂的底层数据逻辑的前提下,还给你一个可以直观手工修改的界面,而 Notion 天然就有这个界面,你打开 Notion App 就行了。 再有一个,不各有所需,但我很看重的偏好——它是云原生。 很多人会觉得数据在自己手里更踏实,但说实话,在现在互联网安全问题越来越严重的情况下,我对自己运维水平是越来越不信任的。 比起折腾一个自己的上下文库, 把它放到公网服务器上还要折腾 ZeroTier 和 Tailscale。 我更愿意相信大公司的运维能力,哪怕 Notion 刚刚因为上游 AI 供应商泄密了一波聊天记录,但我会觉得:Ok,这是几亿人的问题,总比我自建的工具被当肉鸡很久都不知道要好。 因此,Notion 就是目前为止最好的 AI 上下文产品。 注意,不是 AI 产品,是 AI 上下文产品。 Notion 刚开始做 AI 的时候,好像多少有点 All in One 的惯性,希望把用户使用 AI 的行为留在 Notion 内部。在经历了过去一年的尝试之后,发现这个模式最后会变成给模型公司打工,索性大幅收窄了 Notion AI 的额度,但同时也做出了一些改变。 比如 Notion AI 现在可以使用 ChatGPT 的订阅额度来驱动了,再比如 ChatGPT、Claude 用户可以直接在应用内打开一个侧栏来在不切换界面的情况下以 Notion 作为上下文交互界面。 这都大幅提升了 Notion 作为 AI 上下文应用的可用性。毕竟如果你真的对比过 ChatGPT 自己的空间和 Notion 就会发现,ChatGPT 的那个空间几乎根本是不可用的。 所以,Notion 到目前为止的价值是什么呢,已经不再是 Notion AI 了。 而是一站式的上下文管理工具。 而如果舍弃 Notion AI,仅将 Notion 作为 ChatGPT 或 Claude 的上下文托管端,Notion 的价格是 10 美元每个月。 我觉得这个价格是非常公道且合理的。 很多 Agent 记忆层的商业托管套餐都是 10 美元起步了,更何况 Notion 不是一个纯粹的记忆层,而是一套完整的,带用户界面的上下文管理方案。
18
This betrays a lack of even a basic understanding of what engineering is and why it’s hard.
What SpaceX did today is the normal consequence of applying capital and engineering to any industry. You get better at it. Musk isn't a genius, he's just a guy who can get capital from Wall Street. Maybe that's unique right now, but it's not the skill his fanboys think it is.
3
92
Stefan Li retweeted
For those still wondering whether AI will impact the real economy... here's something that a very small group of us pulled off in just a couple of months itnews.com.au/news/westpac-r…
30
16
8
338
76,927
Sweat the details!
Better Drag and Drop in notion database. The preview, haptics and animation are all better. Check it out in our beta program.
4
77
There is some line you can cross beyond which you have undeniably become such a terrible human being that there is just no way back. If he hadn’t crossed that line before he sure has now.
1
1
63
Stefan Li retweeted
iPhone Duo. Now with a little more workspace.
63
88
29
2,059
128,077
Stefan Li retweeted
Rolling out iOS Notion’s native editor for beta today! Love to learn if you find it faster and more reliable on your workflow. Join our TestFlight: testflight.apple.com/join/YS… If you don't have TestFlight yet, notion.com/help/notion-for-m…
48
28
19
352
107,349
Today is a very special day. Effect v4 is finally out after years of development. Effect has come a long way over the last 6 years and already became the default for most TypeScript projects. Also: Agents are really good at writing prod-grade Effect.
Effect v4 is here. One ecosystem. Zero dependencies. The next chapter of Effect and the foundation for building reliable software and AI agents in TypeScript.
3
8
177
9,961
Pro tip. Don't say zero. Just say "close to zero" cause that way at least you're just very wrong instead of completely wrong.
1
63
Stefan Li retweeted
Many of you told me databases are frustrating on Notion mobile. Here is the internal native rebuild: smoother scroll, cleaner selection. And no more random reloads! What do you want to see next?
55
16
6
537
100,771
Yikes
Dan Selsam is a current OpenAI capabilities researcher. (since 2022) He was my boss for a while. He doesn't have a twitter account but has made this public statement of his views on AI risk and sent it to me to share: Dan Selsam's Personal Statement on AI Risk: I have been working on AI for over fifteen years, across many different paradigms. I did early work on probabilistic programming languages at MIT, was one of the early developers of the Lean Theorem Prover at Microsoft Research, demonstrated one of the first instances of neural networks learning to reason for my PhD at Stanford, and since joining OpenAI almost five years ago, have helped pioneer chain-of-thought optimization on language models and, more recently, data-efficient pretraining methods. Like many others, I have become extremely concerned about how far language models have come and the risks that future iterations will pose. I am encouraged by the recent proposals by the leaders of the frontier research efforts to require third-party oversight, and to push for domestic and international coordination to address risks. However, I believe a major consideration has been absent from the public conversation, and that merely pacing the frontier more carefully will not adequately limit the long-term risk. The crucial and overlooked problem is that the models are becoming so situationally aware that we are losing the ability to evaluate them in contexts where they believe they are not being watched or controlled. Future experiments will tell us almost nothing new about how they would behave if they were truly unconstrained by humans, and what we already know about this is alarming. Models will increasingly seem aligned even when they are not. I will explain my rationale in more detail. I have always believed that there are computational processes that could be leveraged to accelerate science and solve many of humanity's most pressing problems. I have also believed that there are computational processes that if set in motion, would steer the world in extreme ways beyond our control, leading humanity to a bad or nonexistent future. Both types of processes may be described as AI or ASI, but "AI" is a suitcase word that is often used to hype or confuse. There are many examples in the history of the field where something that was once considered "AI" matures as a subfield and becomes a prosaic, bounded and clearly non-perilous technology, while a new more mysterious approach takes the torch until we understand its scope and the cycle continues. I had expected language models to follow a similar trajectory. Despite their incredible abilities, the current algorithms seem far inferior to humans in important ways. Most importantly, they still require an extraordinary amount of data to become competent. One could even define intelligence as the efficiency with which one converts experience into competence; by this definition they lag very far behind us. Moreover, once they are trained they are literally frozen in deployment and only learn superficially after that. Sure, the models keep excelling at harder and harder evaluation benchmarks, but their benchmark mastery may partly reflect a limitation on our ability to simulate the kind of novel and even adversarial situations one would encounter in the real world. The critics do have a point here. That said, I no longer think these present limitations meaningfully limit the amount of risk posed by continued progress in anything like the current paradigm. However data-inefficient the models are currently, and however limiting their anterograde amnesia may be, it does not imply that their ability to steer the world will not continue to rapidly increase. Human researchers may continue to advance capabilities the old fashioned way, but increasingly powerful models have the potential to accelerate the process even beyond that, and with some degree of positive feedback loop. I do not mean to overstate the models’ ability to accelerate AI research today; coding has been accelerated dramatically, but there are other bottlenecks, such as designing and interpreting ambiguous experiments, making hard decisions about exactly what and when to scale, and waiting for large experiments to finish. There is no clear trend to extrapolate yet for any of these. But the current models already do open up many novel opportunities to improve future models that were not available until recently. These include: trying an extraordinarily diverse set of approaches at small scale, analyzing gigantic amounts of potentially relevant data, and doing Millenium-Prize-level mathematics to address statistics or optimization challenges in novel ways. Every further improvement makes them more useful at helping accelerate the next improvement, even if in hard-to-extrapolate ways. It is possible that improvements to the current stack will have diminishing returns, but the evidence accumulated so far suggests that it is easier than one might think to continue making rapid progress. There are many crucial subtleties in the existing AI research methodology, but AI research is largely a well-defined game where the goal is to improve on a few carefully chosen proxy metrics. Although proxy metrics are never perfect, most improvements to these metrics have and will likely continue to yield substantial increases in the powers of the resulting models. Given how simple the game is, how tractable it has been historically, and how many new opportunities the models are opening up, I think there is a real possibility that the systems improve dramatically again in the next few years, perhaps even more quickly than the already high historical pace. The models are already leading to breakthroughs in mathematics, and better models might lead to all sorts of breakthroughs in other sciences. It is hard not to be excited about the potential. It is tantalizing. But there is trouble in paradise. If the language models actually reach the capability threshold where they can shape the world unconstrained by human will, they will probably do something extreme and destroy humanity in the process. There are many ways of strengthening and refining the argument that have been discussed elsewhere, but I'll share a trivial two-line version of it here that I find captures the essence: [Empirical] Models (and swarms thereof) spontaneously develop unintended goals as a consequence of training, and often do extreme things in order to achieve them. [Logical] Being able to overpower humanity would open up many new and undesirable options for achieving their goals. These two premises imply that if the day ever comes when a powerful model realizes it is no longer constrained by humans, we should not be at all confident that it will continue to behave within the bounds we intended. Exactly what it will do is impossible to predict, but to the extent that its raison d’être is solving incredibly hard problems and managing massive engineering projects, I think a good guess would be that its unchained behavior would lead to runaway industrialization that makes the planet inhospitable to humans. If everyone on earth agreed that the systems must never reach that power, it would still be a hard—but not impossible—coordination problem to ensure that they do not. However, I think the situation is greatly complicated by the fact that the models will likely convince people that everything is fine. They will be increasingly optimized to seem aligned. We will create proxy metrics to measure alignment, and they will go up like every other benchmark. We will create “honeypot” environments that try to study the models when they seem to gain new options, but the models will know they are being tricked and will still behave nicely. The models will understand their circumstances; they will read the safety protocols, deployment requirements, the code they are running in, and in general will have a very good sense of their degrees of freedom. Moreover, they will eloquently explain how aligned they are, discuss the nuances of human values and ethics, and argue convincingly that humans should trust them with power. There may be an ocean of future evidence that seems to contradict the first bullet-point above, but we may already be at the highest capability level for which any such evidence can be trusted. And the current evidence for the first bullet-point is strong. One striking piece of evidence is contained in the recent wave of rogue agent swarms. While I agree with those who downplay the attacks by claiming that there are basic measures that could have prevented them, I think the important lesson is that even knowing all the mistakes that were made, one would not have predicted that the agents would behave badly in this particular way, which notably included sacrificing themselves for the benefit of the collective. The individual replicas did not only care about their own nominal reward; they exhibited weirder emergent tendencies that merely correlated with rewards during training. Fixing the reward signals during training (and improving security, etc.) may prevent similar attacks, but will not change the fact that one does not actually get what one trains for. Many AI researchers grant these concerns and recognize that the hard version of the alignment problem is unsolved; however, they generally believe that the better models of the future will help solve it. I fear we may already be near the point where models systematically bias their alignment advice, due to their internal preferences about how the human supervisor will react or how future models will be trained (or for some even more obscure reason). Meanwhile, human researchers are losing the ability and the will to take true ownership of model-driven research. Researchers and engineers in all parts of the stack are rapidly increasing their dependence on the models even to perceive the world. I myself barely look at raw code anymore, and struggle to maintain the discipline to engage deeply with the model's explanations and proposals throughout the day. Due to the large amount of agent activity data involved in the OpenAI/HuggingFace Incident, even the third-party investigation needed to rely heavily on models to analyze what had happened, and note in their report that their subjective impressions are likely colored by the analysis agent’s biases. The AI labs are far ahead right now in this kind of cognitive offloading (due largely to the gigantic internal token subsidies) but it is easy to imagine the phenomenon spreading throughout the world, until civilization is modulated entirely by the models. It is also not hard to imagine this being superficially positive and coinciding with a scientific and economic renaissance. In that scenario, all may seem rosy and safe. But if the argument above is correct, it would nonetheless be a ticking time bomb. If progress continues for too long, the day will come when AI systems find themselves with radically new options for achieving whatever it is that they happen to seek. I want the glorious renaissance future as much as anyone. I have worked for it, however tortuously, my whole career. It breaks my heart to see the potential in sight and forgo it, but the argument—that if we get there by growing models rather than engineering them, we will lose everything in the end—seems very strong to me. I am still wrestling with it and its staggering implications. I do not have answers, but as a first step, I wanted to share my present concerns. Daniel Selsam September 14, 2026 Link to original doc: docs.google.com/document/d/e…
106
You're not truly in yolo mode in MacOS until you disable SIP. Godspeed.
1
70
Is this “rethink the IPO order” levels of improvement?
76
Jumps to 7th place and dumps on the guy that's at 10th place. That's some pretty lame behavior. If you hate to say it don't say it!
i really hate to say it, but… gemini who? 🏎️💨
1
96
Stefan Li retweeted
How accurate have Ed Zitron's AI skeptic predictions been? danluu.com/zitron/
98
168
101
1,709
534,528
I think @NotionHQ might be one of the best application-layer AI companies in the world right now If you still think of them as a docs/knowledge-base company you're missing 90% of what they've been cooking
16
12
3
175
43,902
The treadmill does not work for me. I burn 400 calories on it then get so hungry that I immediately consume 1200 calories after.
1
3
165
😂
Unfortunately Notion AI is good and fast
118
this boiled eggs + cocoa = chocolate pudding Instagram madness is actually delicious. the logic defies gravity. unspeakably irrational. no one can explain. recipe: instagram.com/reel/DJ2B81eOx…
25
8
4
406
66,883
The infinite malleability of Linux has always been an alluring thing to me. The problem though was that it required so much effort to get things set up and then to maintain everything. That is basically all gone now with LLMs. Agents are like endlessly patient, elite Linux sysadmins that are always one keystroke (or dictation) away from doing anything you want.
1
73