Tech investor. A skeptical empiricist surrendering to life; Studying fat tails, removing blind spots, seeking dynamic quality, and letting go. Views are my own.
New York, NY
- Tweets 46,980
- Following 7,501
- Followers 2,675
- Likes 185,346
They are building safe superintelligence. They have said there will be no intermediate products. It's not that difficult to understand
10
1
279
16,113
I support Sam's preference of releasing early. Iterative deployment is what the world needs and what's good for OpenAI imo.
2
3
247
6,309
What do you mean? His whole point of starting SSI is to not have to give any public updates until ASI is achieved internally.
4
3
200
12,851
From Value to Growth: Some Thoughts on My First Semester at Columbia Business School substantiainvesting.wordpres…
11
21
190
As long as investors are happy to keep funding it, then I guess they have no obligation to provide any updates to the public?
6
150
6,899
the o-series remain undisputed kings of reasoning models. Grok doesn’t come close at all in my experience
5
2
117
4,229
Had a fantastic and unique session today with @altcap & @DanielSLoeb1… They reinforced in me the importance of thinking probabilistically, maintaining a very high level of mental flexibility, and having passion in what we do!! Thank you!!!
5
1
89
Pop’s winning % is actually a lot more impressive due to there are multiple games in the NBA each season
2
63
On behalf of the Greater China Society at @Columbia_Biz, we invite you to attend the 13th Annual Columbia China Business Conference on April 6th-9th. It is FREE for everyone. Please register in the following link: eventbrite.com/e/the-13th-an…
6
10
70
Li Lu rarely speaks live and publicly.... How can you miss it....?
On behalf of the Greater China Society at @Columbia_Biz, we invite you to attend the 13th Annual Columbia China Business Conference on April 6th-9th. It is FREE for everyone. Please register in the following link: eventbrite.com/e/the-13th-an…
4
11
70
Having a chat with the awesome @DennisHong17 this coming Friday. What would you like me to ask?
11
55
Well there are only a few frontier labs. I guess you are referring to the widely circulated rumor about Ant....
1
49
18,150
Can you explain how it works? How does the little thing know what I've read or even summarize what I've read?
11
52
2,985
Magic has 100m token context and I expect OpenAI to soon announce something in the millions.
4
1
47
9,657
But… can a single router learn the preferences of a billion users which might all be different?
6
49
8,237
Got access to Copilot Voice Mode. Daily usage limit is around 10 minutes for free users.
Pretty sure it's based on 4o
But it seems to be implemented differently, maybe fine tuned differently and with a different set of custom introductions.
It hmms when you speak, it cannot be properly interrupted, and it's not very stable. But it offers the illusion that it responds with lower latency.
Overall a worse experience than advanced voice mode imo.
I think maybe this shows Mustafa has bad taste lol
4
42
3,063
Real question: why do you hate OpenAI on this? 4.5 OBVIOUSLY will soon roll out to everyone. Besides, your favorite lab, such as Anthropic, has been keeping its Opus baby internal for almost just as long and still hasn’t released it.
This is real - the ChatGPT app for Android (1.2025.056 beta) has a new announcement: "Try the GPT-4.5 research preview - Pro users now have access to our newest, largest model."
8
1
46
6,610
I think this is impressive because they are also trained to use the tools well
They didn’t just give o3 tools and call it a day
In other words, even if you give R1 access to the web, which it already has on the official app or Perplexity, it still just doesn’t compare
3
45
7,846
No. The issue is you guys claimed you have the best reasoning model, when that’s clearly not the case.
1
43
6,904
First class of Prof. Mauboussin’s Security Analysis tonight 😝🥰😍😆☺️ @mjmauboussin
3
42
The high end was never in play for them imo. They should be thrilled if they become the Android for XR .
1
35
7,293
Warren Buffett will always be my biggest hero. But this doesn’t mean I need to do everything as he does (1/10)
3
5
38
Had the incredible @patrick_oshag in class today… It was amazing to learn about his concepts of (1) growing without goals and (2) learn, iterate, and share. He was totally inspiring and motivated me to forge my own path!
38
We’ve been so disappointed with the last few days of 12 days of OpenAI. Will we finally get excited in the next two days?
2
37
4,079
After using 4.5 exclusively for seven days, switching back to o3-mini really crystallizes how much better it is for conversations than a small reasoning model.
2
1
39
2,058
I would be very surprised lmao. With Poole and Klay on the team, no way the Kings defend like that
35
2,707
Agreed. And AVM vision was announced 7 months ago. So it feels like we didn't get much today from OAI.
33
7,440
Bro, wait. OpenAI never promised to release their open source model next week. A delay of a few weeks is okay. Chill.
3
36
12,535
Most interesting point to me is it does seem the reasoning capability will be independent of the scale of the underlying model. This is likely the Q* we are all waiting on.
3
32
4,739
AI has strengths and weaknesses, just like humans.
How long will it take a third-grader to figure this one out?
5
29
2,532
Any idea why a 400b+ model can be nearly as good as the ~2 trillion model? Pre-trained on more data?
13
30
16,659
Gemini is a dense model. Does that mean even the Ultra version would likely have fewer parameters than GPT-4?
3
29
16,716
I am betting on Ackman winning the war. Mostly because the Ackmans have little to lose anyway lol
27
3,908
Nice. But without noise cancellation how can it replace AirPod Pro regarding quality of audio when listening to music?
2
24
57,582
"Productive people and companies force themselves to make choices most people are content to ignore.”
1
24
Wrote something about Independent Thinking
desilorance.substack.com/p/t…
1
2
29