1998-2023: Search 2023-?: Ask
?- : Do
@aravsrinivas
-
Perplexity Accelerates Shipping Speed with Linear Integration
By
–
Perplexity is known for shipping fast, and a big reason is our use of @linear
. Great to see Karri using and benefiting from Perplexity! -

Microsoft Google Apple AI execution and answer engines
By
–
Who is executing well, and who is executing poorly, between Microsoft & Google? What do you think of Apple in AI? Is "answer engine" a similar change to the kind Google's Pagerank brought to web portals 20 years ago? What does it mean to search like Jensen Huang? These are some
-

Next Generation Phones Will Transform Search Technology
By
–
Next generation phones are going to be next generation search too.
-
nanoGPT with MoEs and multimodal tokenization support
By
–
i would love for @karpathy to make a nanoGPT for MoEs with multimodal tokenization support
-
Building at Scale: Why Weekend Projects Never Account for Complexity
By
–
“i could build this in a weekend” is a sign of someone who never actually built anything that works at scale
-
Copilot Unlimited Features: Browse, Generate, Upload Capabilities
By
–
Copilot mode (browses and disambiguates hard questions and does a thorough search), Image upload, File upload, Model choices, Image Generation, API credits… all unlimited…whatever works for you.
-

CNBC Fast Money: Gemini, ChatGPT4, and Perplexity Comparison
By
–
CNBC Fast Money discussing Gemini, ChatGPT4, and Perplexity (though I actually think they were testing the free version, which is a testament to how even free version of Perplexity can rival paid versions of Gemini/ChatGPT4, and the Paid is superior on research and deep dives).
-
Model-Free Policy Training From Offline Model-Based Planning
By
–
sure, if you can train a model-free policy using trajectories collected from model-based planning offline, at scale, that works too.
-
Sora’s Physics Modeling Limitations: Honest Research Assessment
By
–
Seeing a lot of tweets saying @ylecun made incorrect predictions. Reality is Sora, while being amazing, is still not ready yet, to model physics accurately, and the authors of Sora, are intellectually honest researchers, who mentioned this on the tech report section of the blog, https://
t.co/c8WKFoge0f