Dasdan said Dropbox has over 700 million users and tens of exabytes of data.
“We have over 700 million users, you know, tens of exabytes of data, even our code is millions and millions of lines of code”
Person
Ali Dasdan
1 episodes · 16 quoted claims
Dropbox built its own search engine before the AI era because available tools could not handle its scale.
“For example, we had to build our own search engine. This was before AI even because we had to.”
Dropbox does not build foundation models and instead adopts the best available models.
“we don't do foundation models, we try to adopt, you know, the best models that are out there.”
In early 2025, Dropbox's adoption of AI coding tools was below about 40%, according to Dasdan.
“So 2025 early we were around, I think, less than 40% adoption”
Dropbox reached almost 100% of its developers using AI coding tools, according to Dasdan.
“And we were able to reach almost 100% of the people, not only using actually these AI coding tools”
Dash includes a container called a stack that groups multiple pieces of content into one unit that AI can be pointed at.
“We have a container called a stack. You can group multiple, let's say, content pieces into one unit”
Dash connects third-party data sources only after users agree to and initiate the connection.
“as long as you connect these data sources, users have to agree and they have to initiate that.”
Dash pulls connected content in real time while honoring governance and security permissions.
“Then we pull the content in real time and also make sure that we honor all the governance, you know, security permissions.”
Dasdan cited Stanford research showing large language models may focus on the beginning and end of their input and ignore the middle.
“They showed, for example, LLMs might be focused on, you know, the first data, as well as the last part, middle parts might be ignored.”
Dasdan said Stanford did research around 2023 on how language models use long context.
“There is some research, for example, I think 2023 or so that Stanford did that.”
Dasdan said Dropbox is fundamental to many businesses, especially those handling lots of media files such as construction and advertising companies.
“Dropbox being a file sharing site, you know, and storage site, it's very fundamental to so many different businesses”
Dasdan said customers with large video libraries want to find specific frames within videos, which generative AI now makes easier.
“It's not only sometimes finding the video, but finding a certain frame within a video.”
Dasdan described users' work data as spread across platforms such as Salesforce, Slack, email, Notion and Confluence.
“They have, let's say, sales data in Salesforce, they are using Slack, they have email, they have I don't know, Notion or Confluence”
Dropbox still has humans perform code reviews, including for AI-generated code, with AI providing context.
“code reviews are still done by a human, even though, you know, we try to provide as much context as possible using AI.”
Dasdan said AI can be used to check code that other AI generated.
“you can use AI actually to check other AI code, AI generated code.”
Dasdan said some foundation model labs use their own AI tools to write new models.
“even in some of these foundation labs, actually, they're using their own tools to actually write these new models.”