Searching...
18 results for “scaling mindset”
scaling and the obvious opportunity to improve from here? There's still meaningful dimensions, of scaling that, like, haven't been, I think, fully explored in terms of, you know, IQ. I think there's a lot of cool efforts out there. There's a lot of c
15:03 / 16:10
from transcript15:03 – 16:10
There may just not be buyers even if it's a great company, and so you have to reflect that in your thinking. Yep. Alright. Let me squeeze one last one in, and then we'll break. Especially because you've done a lot of developer focused stuff that's be
54:42 / 55:53
from transcript54:42 – 55:53
...part of scaling laws and that's part of reasoning and that's part of all these things that have, you know, all the post training stuff that's been rolled out. So, there's just been so much innovation in such a compressed period of time that that's th
38:13 / 39:16
from transcript38:13 – 39:16
and we have a lot more to do to go and scale those. You know, it's it's RL, for example, is, I would say, the the biggest breakthrough of the last year and a half, call it. It's it's crazy to think that you you roughly can solve any benchmark. It's i
14:58 / 17:29
from transcript14:58 – 17:29
I then was obsessed with figuring out what's the chain of thought that would cause someone competent to choose that. That brought me to discovering Paul Graham's essays and Y Combinator and learning about the entire Silicon Valley ecosystem. I rememb
6:44 / 7:45
from transcript6:44 – 7:45
...like, we're scaling up on existing dataset in pre training. Is that a fair contrast? Spikier, like, what's the plot here? What's the, like, the x axis and the y? Domain is the x axis and y is capability? Yes. Because you you're, like, choosing what d
29:20 / 30:27
from transcript29:20 – 30:27
needs to be that it should be trying to do that. Like, the model needs to actually be, like, trying to learn all of these, like, trying to learn about you, trying to learn about, you know, the things that you care about. Like, the the actual objectiv
33:38 / 34:59
from transcript33:38 – 34:59
It doesn't need to be like but if you kind of train the model to, like, respect this, you would hope that if you'd, like, execute this correctly, the model learns to, like, roughly think the right amount. Okay. This actually changed my opinion of thi
11:16 / 12:16
from transcript11:16 – 12:16
Like, we completely changed the core interaction of the product. If we had launched that publicly, we never would have been able to switch it. There's no way because users would have learned a new behavior. User would have said, oh, this is cool. The
39:31 / 40:49
from transcript39:31 – 40:49
But as an investor, when you're a series a investor, when you see someone growing at the rate of Codian, Windsurf, or do you just not care about motes because you you you, like, you gotta pick your jaw off the floor for growth? Or how do you rethink
55:28 / 56:31
from transcript55:28 – 56:31
“AI gives you 100x leverage to generate personalized experiences at scale”
You might have tested if you ran AB test or experimented with some images as you were trying to promote the product. Now you can generate 100 x, 1,000 x, of a similar experience. It's very personalize
6:29 / 7:08
tactical advice6:29 – 7:08
“Why early stage VCs should hunt for 100x wins, not safe 2x returns”
As an as an as a master, you wanna learn lean into risk, and you really wanna, in a sense, look at things from the perspective of, like, what could go right, which is a very different approach. Right?
4:41 / 5:04
framework4:41 – 5:04
...cross some scaling threshold or new algorithmic development. For us, the solution has been to first try to create a system that is pretty agnostic, that can use any LLM that's out there and use different systems to accelerate their performance or bui
18:43 / 19:53
from transcript18:43 – 19:53
So micro in these games, something like StarCraft is like, alright. Can you click really fast and, like, move your people and build things really quickly? And, it's like the mechanics, like, every second. Like, how many how many hundreds of things ar
1:19:27 / 1:20:32
from transcript1:19:27 – 1:20:32
Like, we instead of hiring PR firms and buying AdWords and stuff, we were creating these, like, viral videos, and we were using, you know, using math from epidemiology to create these viral loops. In a lot of ways, Dropbox took that, transplanted tha
1:23:55 / 1:25:00
from transcript1:23:55 – 1:25:00
I'm not sure that they're, like, that diff I think, like, we don't know how they do it under the hood, but my guess is that the whole reasoning effort thing is essentially a token budget that the model has been, like, RL ed to like, you would hope th
11:54 / 13:03
from transcript11:54 – 13:03
Or, you know, can we really get the model to learn, like, longer time horizon kind of meaningful extended behaviors? Like, it it it feels like kind of a natural place to start in some ways and that it, you know, also is fairly likely to be, like, use
12:45 / 13:53
from transcript12:45 – 13:53
play? And the work isn't done, but but came back with with kind of three key ideas. What can the models not do? Aspire. Set the right aspiration. Do you go to low earth orbit? Do you go to the moon? Do you go to Mars? That's a uniquely human capabili
23:14 / 24:23
from transcript23:14 – 24:23
Have a podcast?
Run your podcast through Clypt
Get ranked clips, hooks, and ready-to-post copy from your own episodes. Free to try.