OpenAI has acquired Multi Software Co., a startup with a videoconferencing platform for software development teams.

Multi co-founder and Chief Executive Officer Alexander Embiricos announced the deal in a blog post published today. Embiricos, who was previously a product manager at Dropbox Inc., founded the company in 2019 with former Google LLC engineer Charley Ho. TechCrunch cited a source as saying that the deal is an “acqui-hire” that will see most of Multi’s five-person team join OpenAI.

Multi’s namesake videoconferencing platform is built on Zoom Video Communications Inc. infrastructure. It includes a screenshare tool that allows developers to make their displays viewable in real-time to colleagues while they’re writing code. Users can open new applications and files on one another’s computers, as well as highlight items of interest on a shared screen with an embedded sketching feature.

The platform also includes a set of artificial intelligence tools. The platform can automatically summarize a meeting, create a list of action items and provide links to the code files that were used by the participants.

Multi says software teams use its platform for pair programming. This is a development approach wherein one developer writes code while another checks the newly written code for bugs and suggestions improvements. Multi says its installed base includes application teams at VMware LLC, the Mozilla Foundation and other major tech industry players.

The company will shut down its platform on July 24 and delete users’ data. In today’s blog post, Embiricos detailed that customers may request a one-month extension if they need more time to switch to a different service. Multi plans to help users find competing applications that can meet their requirements. 

“What if desktop computers were inherently multiplayer? What if the operating system placed people on equal footing to apps? Those were the questions we explored in building Multi,” Embiricos wrote.

The deal comes a few days after OpenAI announced another startup acquisition. The ChatGPT developer last week acquired Rockset, a database provider backed by more than $100 million from Sequoia, Greylock and other institutional backers.

Rockset is built to power applications that require the ability to quickly carry out analytics tasks. A product recommendation engine, for example, could use the database to quickly turn newly obtained data about a customer’s buying preferences into a shopping suggestion.

Another core selling point of Rockset is that it simplifies the process of creating indexes. An index is a collection of shortcuts that allows applications to find specific records in a database significantly faster than would otherwise be possible. The ability to locate specific pieces of information quickly is particularly important for language models, which must often retrieve large amounts of data to generate prompt responses.

OpenAI last week disclosed plans to integrate Rockset’s technology into the “retrieval infrastructure” that underpins its products. Rockset’s website, in turn, states that one of the use cases to which its database can be applied is retrieval-augmented generation, or RAG. This is a machine learning technique that allows large language models to incorporate information from external sources such as a company’s internal applications into prompt responses. OpenAI’s flagship GPT-4 model supports RAG. 

Image: Multi