“Learn to code” is the wrong frame.
Just build stuff, and learn exactly what is needed to accomplish your goal. AI makes this easier and more fun than ever before!
93
316
3,761
190,510
I like to think of @bryan_johnson as a civilizational milestone you can only get at $100T global GDP
19
7
741
116,034
I am extremely excited for the incoming explosion in the quality of software.
It’s so incredibly hard to make good software, and it’s so under-supplied as a result. Humanity deserves amazing UIs for even the most mundane use-cases.
12
14
343
25,067
🚨BREAKING: OpenAI announces new AI agent that can get HOURS of work done in SECONDS while also SACRIFICING the entire LIGHTCONE to an uncaring MACHINE GOD
10 tips to get the most out of this incredible development 🧵
10
19
348
37,321
Is anybody working on an AI agent that can follow high level text scripts to perform browser tests?
Ideally it could also self-generate scripts by playing with the UI, and recommend updates to scripts when UI changes.
I think we would pay a lot of money for this.
78
13
340
91,859
The vibe shift feels lacking in a dimension I think is extremely important: compassion for the experience of sentient beings
15
11
323
21,445
I’ve noticed a midwit phenomenon when it comes to predicting AI progress.
People really close to the technology (like lab employees) seem to have a good sense of what’s coming. They understand the details at a deep technical level and are seeing early previews of what is to come.
People who take an outside view also seem to do pretty well. Not over indexing on any particular approach, but looking at the overall trend.
People in the middle who understand some of the technological abstractions I very often see over-applying them and coming to the wrong conclusions.
11
8
265
37,958
The arrow of history is clearly going to point towards increasingly extreme body modifications, enabled by technology. Trans is just the beginning, get over it.
10
13
216
31,762
I’m out here building b2b saas
Meanwhile in synthetic biology: we’re trying to turn the ribosome into a more universal matter compiler using directed evolution on cells with a special side ribosome that uses a different codon mapping
9
6
180
17,125
I think it’s optimal to always be a bit on the edge, a bit overwhelmed, a bit unsure if what you’re trying to do is even possible.
If you’re feeling very comfortable, consider that maybe your task has become too easy.
3
8
165
27,006
We are now entering a confusing era, where two things will be true simultaneously:
The application of human agency towards a goal will be more powerful than ever, with the help of AI. The experience of doing stuff will be exhilarating.
At the same time, human competitiveness on any particular task will diminish rapidly, to the point where there is nothing that you could do better than an AI. It might feel hopeless to try to do anything at all.
Focusing on what you can do that AI won’t be good at will ultimately be a losing strategy.
I think the right strategy is to focus on what you want to see in the world, and figure out how you can maximally apply AI to help you achieve it.
If you’re not already doing something, go start today!
12
18
164
16,023
I wrote a short op-ed in support of SB 1047 — AI capabilities are advancing rapidly and I believe we need to start setting up sensible guardrails now:
latimes.com/opinion/story/20…
34
22
150
82,538
I'm super excited to have @thesephist joining us at @NotionHQ – we'll be exploring how AI can truly push us forward as humans, and moving quickly to ship useful products.
Come join us!
Life update🎉
I'm very excited to be joining @NotionHQ to continue prototyping and researching ways AI can help us be more creative, thoughtful, and productive!
Looking forward to learning from the team and bringing some of my ideas from the past year to a tool loved by many 👇
1
2
98
28,121
Join us next Thursday at @NotionHQ NYC for an AI show and tell – featuring demos from The Browser Company, Replit, Nomic, and more!
lu.ma/hxhxwmjg
1
9
92
53,544
It’s so sad that a generation has been brainwashed to believe humanity has no future.
We need to be dream maxxing. Building an insanely glorious future.
Oh god this just messed me up
ALT Tiktok screencap that says “day 1 of decreasing my 401k contribution every time I can comfortably wear just a t shirt in December”
3
3
86
14,498
If compassion isn’t a core part of your stance towards the world, you have lost your way and become morally bankrupt.
From here flows all of the most evil things humans have done to each other.
4
7
81
3,247
I agree, these ads are a terrible vibe.
Two principles for use of AI:
1. It should be clear to all parties that AI is being used (no deception).
2. AI should do tasks it can do *better* than a person could have, because it is too tedious, time consuming, or expensive (no regressions).
The Apple Intelligence ads have been so bad:
▫️Mom forgets birthday gift for dad so AI makes a photo collage
▫️Lazy office worker can’t write an e-mail to his boss so AI does it
▫️Guy didn’t read meeting memo, so AI summarizes it and he can lead conversation
These are all use cases that people have been using Gen AI for but, man, you gotta let the shitposters joke about it (and not make it the official ad and basically call your users lazy idiots).
Apple products are supposed to be — as Steve Jobs said — a “bicycle for the mind”. Gen AI is real good a thinking and ideation aid. This is not that. Almost as bad as the iPad “Crush” ad from earlier this year.
And you know they dropped millions on this. So grim.
6
4
82
33,667
1e3 → 1e8 was quite a journey! Only 1e2 left now ⚡️
Last week, @NotionHQ passed 100M users!
I found the photo I took after we reached 1000 users.
And I still recall the day when we passed 1M in 2020 right before covid.
I feel grateful, introspective. Looking back, there's so much learning, yet much remains unchanged...
1/n
3
5
77
10,267
It was really fun to reminisce on the past decade+ of building @NotionHQ – including the time my dad flew out to make sure Ivan was a real person…
Ivan and Simon have been building Notion for 10 years. Today, they’re peeling back the curtain on the early days!
Their first meeting. The first version of the product. Our first conversations with users.
First Block, season 1 finale.
Watch now: ntn.so/firstblockfinale
2
6
68
29,405
I feel so much joy when I think about all the humans being born every minute, or experiencing a feeling for the first time, or thinking a unique thought.
There’s lots of suffering and we should absolutely work to reduce it, but even very poor people consistently rate their lives to be very worth living.
On net, this project feels like obviously the most wonderful thing in the known universe.
1
4
74
6,355
We just released our official MCP server – would love to hear what folks are using it for and how we could make it better!
We’ve open-sourced our MCP (Model Context Protocol) — now available on GitHub.
Plug it into your favorite client and start building richer AI integrations with Notion in minutes ✨
github.com/makenotion/notion…
1
8
70
24,916
I believe all 3 of these are true:
- AI could benefit us massively
- AI could cause short term harm
- AI could be an existential risk to humanity
It’s disappointing to see people push back on 3 without offering real object level disagreements. This is a really important debate!
9
2
67
9,375
I'm so excited to be showing Q&A to the world – the ability to instantly find any information in Notion has made this a daily essential for me.
More to come soon!
What if… you could talk to your workspace?
Meet Q&A, Notion AI’s newest superpower.
Join the beta: notion.ai
1
1
62
7,750
This was a really fun conversation!
I hope it has some useful information for folks building AI products out there.
Notion (@NotionHQ) cofounder @simonlast told me everything he’s learned from integrating AI into a platform that has over 100 million users.
Simon likes to keep a low profile, even though he’s the driving force behind Notion AI, one of the most widely scaled AI applications in the world.
In his first-ever podcast interview, we get into:
- What he would build if he started Notion from scratch today with AI
- How to get high quality and reliable results from AI at scale
- The future of human creativity in a world with machines that think
This is a must-watch for anyone interested in building reliable AI products at scale.
TIMESTAMPS:
00:01:57 - Introduction
00:02:28 - How AI changes the way we build the foundational elements of software
00:10:07 - Simon's take on the impact of AI on data structures
00:13:05 - The way Simon would rebuild Notion with AI
00:23:39 - How to design good interfaces for LLMs
00:28:22 - An inside look at how Notion ships reliable AI systems at scale
00:35:41 - The tools Simon uses to code
00:38:16 - Simon's thoughts on scaling inference compute as a new paradigm
00:49:10 - How the growing capabilities of AI will redefine human roles
00:50:28 - Simon's AGI timeline
2
8
64
14,872
AI Meeting Notes is insanely useful – I use it at least 3x per day. Check it out!
Introducing — AI Meeting Notes to @NotionHQ & @NotionCalendar!
May your memory be perfect, fingers be free.
4
3
65
16,965
The success of firms is going to depend increasingly on how much they leverage AI to drive productivity.
Firms that do this better while outcompete firms that do this worse, with the effects intensifying over time.
The nuance here is that AI has to be actually driving productivity and not just having the appearance of doing so. I think there will be a lot of savvy involved in figuring that out for different use-cases.
3
4
59
11,193
It’s a fact of our universe that, with the right application of technology, there is enough energy and matter for everyone alive today to live extremely prosperous lives.
Our duties (in order of importance) are to (1) preserve the light of consciousness, (2) press forward towards technological maturity, and (3) do the best we can to make current conditions as comfortable as possible.
3
4
60
5,076
Some AI predictions for 2025:
Coding agents that can do end-to-end software tasks become actually useful, reliably doing tasks that would take an engineer at least one hour. Adoption is slower than many expect because it requires a significant change in behavior (much more so than autocomplete).
Models will become strongly superhuman at math, but it won’t have much of a direct effect.
We’ll see extremely cheap and reliable small reasoning models. They’ll be very good at coding. This will unlock use-cases that involve a lot of fan-out.
The next large scale-up in model size will succeed in all the labs, but it’s unclear to me whether they will actually be released. I think probably yes just because it would make the lab look like they are ahead, even if the cost tradeoff doesn’t make sense and the main trend is inference compute.
The gap between top labs narrows further, with OpenAI, Anthropic, and Google very competitive. Google in particular seems to be catching up and I expect more convergence.
Meta successfully replicates the reasoning paradigm, but remains behind the top labs and isn’t the best cost/performance tradeoff.
Always-on ambient AI hardware devices hit the market, but are not widely adopted yet.
1
4
57
6,081
Building Notion AI brought @ivanhzhao and I back to the early days. Since then, we've built a solid team and I'm super excited for the future!
Notion 1.0 was built by our founders in Kyoto, coding in their underwear.
Notion AI was built in their hotel room at a team offsite.
You know the product well. Now here’s the story behind it: notion.so/blog/behind-the-sc…
2
1
49
23,190
Unlike semiconductors or medicine, where progress gets exponentially harder over time, AI progress seems to be getting easier?
With reasoning models, we’ve now shifted from supervising outputs to supervising the reasoning process. Result is faster iteration speed, less data needed, and more powerful generalization.
What seems to be happening is: as AI gets smarter, we can leverage its intelligence more for development.
As we get closer to true AGI, AI will increasingly drive the entire iteration process, eventually improving fully autonomously.
1
1
49
7,260
How we use Braintrust: braintrust.dev/blog/notion
1
7
50
9,439
I just finished reading Ray Kurzweil's new book. "The Singularity is Near" had a big effect on me as a teenager so I was curious to see how his views have updated. Some thoughts:
I really like and vibe with the vision of human brain augmentation as the pathway to transformative AI. However, we seem to be firmly on the path of synthetic superintelligence well before brain augmentation being meaningful, and he doesn’t provide any clear arguments why this is what will actually happen.
I thought his treatment of AI risks was pretty good. He covered the main areas, argued for a large effort to address them, and came out thinking the risk is justified by the significant benefits (a view I share).
Overall, may of his predictions have stood the test of time and I think it demonstrates the power of looking at large scale tech trends and taking the conclusions seriously.
1
3
46
3,800
Our approach is to ship new features as soon as they are ready, and create a marketing release when we feel we have a cohesive bundle. Principle is to never hold up our product team shipping value to users :)
1
5
41
3,939
o1-pro prompt that was insanely useful today:
Please exhaustively analyze and produce a report of:
- Everything in both "EPD plans" and "GTM requests"
- Everything in "GTM requests" but NOT in "EPD plans"
- Everything in "EPD plans" but NOT in "GTM requests"
Please think REALLY hard and be exhaustive.
4
3
35
5,145
A key principle I try to hold: that in the fullness of time, I will always be honest with myself.
Upon receiving critical feedback, our brains will spontaneously produce a plausible story for why it is invalid.
The critical loop is to be able to unwind this, and upon reflection, tend closer towards the truth.
1
33
2,698
I've noticed that criticisms of controls on AI from open source tech folks share a common sense that freedom is the overriding priority.
This works fine most of the time but critically breaks down for dangerous technology.
4
3
32
5,945
I never really used a calendar until 2019 – before that, I just showed up every day and we built stuff.
I still remember the exact moment I realized I needed it (it was 1:1 meetings and increased external interviews).
2
32
2,596
Things get proven correct or incorrect over time based on progress.
Very recent example is a lot of people made arguments about how synthetic data wouldn’t be useful, but this is clearly been proven wrong by the new reasoning models.
6
33
2,377
Braintrust has become an invaluable tool for building quality AI products. Huge congrats on the raise!
We’re thrilled to announce that we've raised a $36M Series A led by @martin_casado at @a16z to advance the future of AI software engineering, bringing our total funding to $45 million.
We’re also introducing functions — a flexible primitive for building with foundation models.
1
27
4,014
My personal use of AI chatbots really shifted this year.
Early 2024, my main use-cases were code writing and search.
Now, I find myself asking about a really wide range of topics and it feels like a genuinely useful cognitive aid. Wild times!
4
28
1,357
I’ve found that “how would I actually do this task?” is a really valuable and underrated question to ask when designing an AI workflow.
It’s easy to get locked into unnecessary technical constraints.
27
2,395
I’m not that deep in the technical details so I try to be really cautious about making inferences based on my knowledge of them.
2
26
4,596
What if we gave the worlds top 1% smartest people a device that transforms their internal train of thought into tokens, then did unsupervised training on all of it?
4
2
24
4,767
You would think signs and notices in Japan would have perfect English now because of ChatGPT. Not yet! 😆
1
2
24
1,924
It’s so peculiar that we seem to be at a knife’s edge between
“the end” of an explosion of growth and progress that cannot possibly continue much longer and
“the beginning” of a long, universe-spanning era of extreme abundance.
Why are we here now?
1) You might think we are actually headed towards stagnation, so the situation is not peculiar.
2) You might try to explain the selection effect using conventional science, like in “grabby aliens” discounting the early and later universe.
3) You might say we’re actually in a simulation, and we are in one of many simulations of this period.
What’s your favorite explanation?
6
22
4,414
Wow Claude is insanely good at medical triage / diagnosis
General flow to try:
Use dictation to ramble about the symptoms and all the context you can think of. At the end, ask it to start thinking of what the diagnosis is, and to ask you more questions to narrow it down.
Keep iterating and having it ask you questions until (hopefully) there is a single clear diagnosis.
4
24
4,939
Brainstore is crazy fast! Our team can now run exploratory searches over logs really easily.
Braintrust is now 80x faster than any other LLM observability platform on the market.
To achieve this benchmark, we built Brainstore, the first database designed to handle the high scale and complexity of AI data.
3
22
5,085
Notion’s GitHub connector is out!
It’s been so cool to see it weave together knowledge across GitHub, Notion, and Slack into a comprehensive answer — give it a try!
As promised, Notion AI now connects to your @Github!
Now you can ask questions about anything stored in your repositories—documentation in code files, recent pull requests, and more—and get answers directly within Notion.
21
1,582
I see a lot of discussion about if or how certain jobs will be “safe” from ASI. Arguments in favor of humans usually break down pretty easily if you think it through.
Consider childcare with ASI + humanoid robots:
- Infinitely patient & kind
- Available 24/7, never tired
- Expert teacher in every subject
- Perfect memory & adaptability
- Emergency response expert
- Fraction of the cost
Want some human imperfections? It could do that too.
To be clear, I think is a very desirable outcome! We should seek to fully automate all economically useful tasks, doing them at much higher quality than was possible before.
We just need to ensure everyone gets fair access to the benefits.
2
21
2,650
Reasoners let us shift from describing step-by-step instructions, to describing the goals of a task
1
22
3,226
I mean, it clearly does protect single family zoned neighborhoods. You’re just saying that you don’t like single family zoned neighborhoods.
I tend to agree with you but I think you could make your argument more clear.
2
19
14,228
The community has shifted largely to thinking that the best path to impact for most people is via their career, not “earning to give” 80000hours.org/make-a-differ…
1
2
19
“We have to be careful, now, that we rule over the digital world, and are not ruled by it.” – Dieter Rams
1
18
1,669
That’s not good! In my experience EAs are exceptionally kind and very open to criticism about the ideas
1
20
I'll be in Bengaluru Dec 18th to share the story of Notion AI - from early prototypes to what's next. Super excited to meet folks building in India – come by!
1
2
18
3,553
Sonnet 3.7 is so extra – I asked for a recap of a paper and at the end it added an SVG diagram of all the concepts (which looked pretty good!)
1
17
1,500
AI models can’t think!!
Here’s one (1) example where I tricked one into doing something dumb: 🧵
1
18
1,706
This is really interesting – has anyone read his book?
WHY CITIES KEEP GROWING, CORPORATIONS AND PEOPLE ALWAYS DIE, AND LIFE GETS FASTER
edge.org/conversation/geoffr…
4
1
16
“In the beginning was the Word. Then came the fucking word processor. Then came the thought processor. Then came the death of literature. And so it goes.”
— Hyperion Cantos
2
15
1,815
I don’t see how this follows.
The algorithmic efficiency trend has been extremely strong over the past few years, and o3-mini seems to hint at further low hanging fruit to continue the trend.
1
16
1,513
I feel like the escalation of people messing with Waymo cars can only end in Waymo self-defense – water guns might be effective.
2
14
1,670
So hyped for this worldofreel.com/blog/2024/2/…
15
2,209
Bio / nano tech will make it increasingly easy and reversible to modify your body.
Many (possibly most) will choose to stick to normie aesthetic standards and we may even see a convergence of what people look like.
But in parallel I think there will also be an explosion of more niche and weird modifications.
2
13
1,174
The only natural institution is family – anything beyond is invented and doesn't exist by default
4
1
15
2,400
Very interesting read.
I think a lot depends on how “cheap” human needs and desires become at terminal technological development.
It seems likely to me that they become extremely cheap and easy to satisfy, so in spite of a centralizing trend on capital, life for ~everyone alive becomes very very good.
this is one of the best essays I’ve read all year and really cleanly articulates all of the thoughts I’ve been yelling to ppl about for a while
2
1
14
3,008
I’ve been reading history of great powers and this came to me: while obviously a German victory in wwii would be terrible, a German victory in wwi seems… kind of good?
- France loses out but probably wouldn’t have lead to action like Germany in wwii due to weaker industrial power
- Constitutional monarchy in Germany gradually evolves to democracy
- Much less developed and isolated USSR, not in control of Eastern Europe
- More gradual decline of British power
- Potentially slower and less heated development of nuclear weapons
- More decentralized world order between USA, Britain, and Germany
- No wwii or Holocaust
- No rise of fascism in Europe
- Potentially no Stalin purges or Mao’s famine
~100M lives saved + more stable development path.
1
13
2,608
AI is close to unlocking a new way to explore ideas: you describe the process at a high level, AI executes the details, then you can check in to nudge and iterate.
The closest existing analogy is something like a research manager, with humans carrying out the details. But with AI extremely willing to carry out tedious tasks, follow precise direction, and iterate at very high speeds, it will feel more like a “research architect”.
I think we pretty much have the core reasoning box now, someone just needs to build the harness around it.
1
13
2,016
Someone needs to make GMO versions of foods that are significantly tastier or healthier than the originals — I bet anti GMO sentiment would go away quickly
5
12
2,227
The ultimate tragedy of the commons: if only we can agree to go slow, we could have use *up to 10^30x more total compute. Will we?
arxiv.org/abs/2104.01191
12
1,419