The development team behind the FFmpeg project thanked AI startup Cursor.ai on Sunday in a message on X for contributing tokens to several developers on the project. It was a bit of unexpected good news from the AI community.
The amount of the credits was not disclosed.
“These credits will support their ongoing work on FFmpeg, including development and code review,” stated the message from the official FFmpeg project. “We greatly appreciate Cursor’s support for FFmpeg and its community.”
Lee Robinson, the Cursor head of AI education, responded with his own X message: “Thank you for creating and maintaining high quality software!”
Cursor’s donation of credits seems like a turning point for AI companies to view open source projects such as FFmpeg as more fundamental to their well-being, rather than just a target to test their latest models.
A Codec Framework For All
FFmpeg is well known even in the open source software community for the quality of its code, which is almost universally used across the Internet for encoding and decoding audio and video files. It is an open source command-line framework for processing multimedia files, largely written in the C programming language and assembly.
“If anyone deserves free credits, it’s the FFmpeg crew,” responded one observer on X. “Honestly that’s such a cool way to help those dedicating their own time and energy to make the internet better.”
Fabrice Bellard originally created FFmpeg in 2000, primarily as a way to provide a standard for what was then a wildly varying quality of competing proprietary video codecs online, from the likes of Microsoft (WMV and WMA), Apple (.MOV) and RealNetworks (RealAudio).
FFmpeg is one of 48 open source projects financially supported by the New York-based Software in the Public Interest, Inc. (SPI), a US 501(c)(3) non-profit organization. SPI was originally founded in 1997 to give the Debian Linux distribution a funding mechanism. Other projects supported by SPI include Arch Linux, Gentoo Linux, Haskell, Libreboot, LibreOffice, PostgreSQL and systemd.
SPI gets its money primarily from donations earmarked for specific projects, either from individuals or organizations, though donations can also include equipment and services. It usually receives between $400,000 and $700,000 in contributions across all the projects.
FFmpeg has no full-time employees; rather, a small group of about 20-30 maintainers oversees the maintenance of various subsystems.
AI Brings New Demands
FFmpeg was one of the first projects to suffer from a deluge of AI-generated bug reports. A year ago, Google’s Big Sleep, an agentic AI system built on Project Zero and Google DeepMind to automatically create security reports on projects, unveiled a number of obscure bugs within FFmpeg, leaving the core team 90 days to develop fixes. The developers were not pleased.
“We take security very seriously but at the same time is it really fair that trillion dollar corporations run AI to find security issues on people’s hobby code? Then expect volunteers to fix,” the project retorted on its X account.
On a Lex Fridman podcast, core maintainer Kieran Kunhya said the action “opened up a lot of interesting fissures.” Google has “substantial resources working on the security of ubiquitous open source projects, and they’ve used a lot of compute and very expensive and very capable security researchers to do that. And their viewpoint is they are contributing by doing that, but I think that’s where opinions differ,” he said.
The debacle presaged Anthropic’s Claude Mythos project, which in April–before its official release–unveiled thousands of bugs across different open source software projects, leaving project maintainers scrambling. This led to the creation of Project Glasswing, which will provide more coordination with maintainers as well as funding and AI credits for open project maintainers to fix newly revealed bugs.