Sarah Catanzaro
@sarahcat21
Followers
13K
Following
18K
Statuses
6K
“All methods are sacred if they are internally necessary” (GP @amplifypartners, prev @canvasvc; Head of Data @Mattermark; @palantirtech; @c4ads)
Joined May 2014
I’ve used Bambino; good for finding people - but honestly, I’d pay for AI screening.
0
0
0
0
0
1
I strongly believe that technical founders need investment partners who not only understand how to scale but also deeply understand their product and technology. @narayanarjun is the investor *I* would want on my board. So stoked to welcome him to our team!
Please join me in welcoming @narayanarjun to @AmplifyPartners, and read more about our winding path to today's announcement:
0
1
11
@immad The Hatch Grow changing pad helped allay my concerns that the baby wasn’t eating enough:
0
0
1
You should just read everything Ari has to say on DeepSeek because every single comment he has made is 🎯.
This was all very explicitly stated in the paper -- the $5.5M number covered the cost of compute for the training run alone. But that's still a huge deal! Many have been arguing that frontier models will soon cost 100s of millions to billions in training costs alone. DeepSeek's ability to do it far more efficiently demonstrates that this is patently false. All of that R&D will be commoditized into easy-to-use solutions for training models (this is our explicit goal at @datologyai -- make it so that you don't need to be an expert in order to train a model on your own data with the best possible data curation). This means that in a few years, an enterprise that wants to develop their own incredibly powerful and specialized small model for whatever use case their business requires, will be able to do so end-to-end for a few million dollars at most in marginal cost. Jeven's Paradox has become surprisingly popular over the last week and it's because it perfectly applies here. If training costs $100s of millions to billions, very few entrants can work on it. But in a world where training costs a few hundred thousand to a few million, it will massively change the landscape. This will be especially important as inference costs become the main driving factor of cost in model development and deployment. In the enterprise, small specialized models that don't have the general ability of frontier models, but which can perform the single task that they need to five 9s of reliability and which can be deployed for a fraction of the cost because they have far fewer parameters than general models.
0
0
3
A wise man who quotes Seinfeld too often once told me that the company that makes the typing noise when you’re on a call with a support agent does millions of dollars in revenue. Async interactions matter…a LOT; but to obfuscate latency and dodge the uncanny valley.
Deepseek is a UI breakthrough - show me the latent space! With speech to speech models, we’ll want to see what the model is thinking while we’re talking. Today’s models just generate silence here. Don’t like a latent thought? Change the conversation
0
0
4
Environment design will matter a lot in the next several years. And not just world models. The hard stuff - like curating data - too.
For friends of open source: imo the highest leverage thing you can do is help construct a high diversity of RL environments that help elicit LLM cognitive strategies. To build a gym of sorts. This is a highly parallelizable task, which favors a large community of collaborators.
0
0
8
Given a take on Deepseek; you can almost perfectly separate the grifters from the real deals
1
0
7
@jxnlco Ooh; this technique often works for finding good restaurants internationally - read reviews in local language to find more recent openings, hot spots, authentic cuisine, etc.
0
0
1
For months, I’ve been opining that so few AI-enabled tools even have good autocomplete. I whined but I understood that autocomplete isn’t really possible with higher cost/latency inference. But now it’s possible. And people will build it. And so I’m not betting against NVIDIA.
2
0
10
RT @arimorcos: What DeepSeek's $5.5M cost demonstrates is that the common narrative that training and customization necessitate 9 figure co…
0
5
0
We will continue to need human data to align with fickle and heterogenous human preferences and likely to unlock new frontier capabilities in domains outside math and programming (that cannot be easily verified)
1
0
3
BYOC, I can't quit you.
What does Taylor Swift and Brokeback Mountain have to do with BYOC (Bring Your Own Cloud)? Apparently, everything! 😋 @nikhilbenesch (@MaterializeInc), spencer kimball (@CockroachDB), @EdoLiberty (@pinecone) -- @AmplifyPartners + @sarahcat21
1
0
2
Vendors must evaluate new deployment models as their customers' needs change and new cloud primitives present opportunities to reduce costs and management overhead. This panel with Edo, Spencer, Nikhil and Grace was a banger. Thanks for letting me join @badgalgge !
BYOC (Bring Your Own Cloud) – everyone is talking about it, and nobody is quite sure if they should support it. So I got 3 leaders in databases together – @EdoLiberty (@pinecone), Spencer Kimball (@CockroachDB), and @nikhilbenesch (@MaterializeInc) – to break down what’s going on:
0
1
9
It’s only bitter because we need to keep re-learning it every time a new SOTA model comes out…
0
0
1