Tools

Search

Explore

Videos Channels Figures

Atmrix

About

Tools

Search

Explore

Videos Channels Figures

Atmrix

About

IF

Isa Fulford

05/08/25

@ Sequoia Capital

The 03 model is effective at searching because it is trained with the same tools and browsing datasets used for deep research.

Video

SC

How OpenAI Built its Groundbreaking Deep Research Product ft. Isa Fulford

@ Sequoia Capital

How OpenAI Built its Groundbreaking Deep Research Product ft. Isa Fulford

05/08/25

Related Takeaways

IF

Isa Fulford

05/08/25

@ Sequoia Capital

Deep research is designed to be the most comprehensive tool for search queries, while the 03 model is better suited for medium-level searches.

SC

Sequoia Capital Cast

02/25/25

@ Sequoia Capital

The model that powers Deep Research is a fine-tuned version of the O3 model, specifically trained on challenging browsing and reasoning tasks, enhancing its online search and analysis capabilities.

SC

Sequoia Capital Cast

02/25/25

@ Sequoia Capital

The quality difference among Deep Research products is evident, stemming from the effort put into constructing the datasets and the optimization capabilities of the O Series models.

IF

Isa Fulford

05/08/25

@ Sequoia Capital

To develop Deep Research, we created reinforcement learning tasks to teach the model browsing and data analysis capabilities, and provided it access to tools for training.

a

a16z Cast

02/07/25

@ a16z

The release of DeepSeek's reasoning traces is significant as it allows for better training of smaller models quickly and cheaply.

LF

Lex Fridman Cast

02/03/25

@ Lex Fridman

DeepSeek's recent models, including V3, highlight the importance of balancing training and inference compute resources to optimize performance and efficiency.

NL

Nathan Lambert

02/03/25

@ Lex Fridman

DeepSeek-R1 is a reasoning model that builds on the training of DeepSeek-V3, enhancing discussions around AI capabilities and accelerating industry conversations.

YC

Y Combinator Cast

02/06/25

@ Y Combinator

DeepSeek's R1 model builds on V3 by applying various algorithmic improvements to enhance its reasoning capabilities, achieving performance similar to OpenAI's models on complex reasoning benchmarks.

SC

Sequoia Capital Cast

02/25/25

@ Sequoia Capital

Deep Research's ability to train end-to-end allows it to adapt flexibly to live web information, making it capable of creative searches and generating insightful responses.