By proposing an open standard for shared agent learning, Mozilla.ai's cq feels like a quiet but important signal. As its name suggests—borrowed from "CQ" radio signaling—cq aims to "call upon" the knowledge of other agents to solve problems already encountered elsewhere. Quite useful, especially in enterprise context.
A very open-source response to a deeper issue: the degeneration of learning ecosystems.
It echoes a question once raised by Andrew Ng: in a world of agents, we need new ways to structure and share knowledge beyond static documentation.
We all know the paradox: LLMs became great at coding by ingesting vast amounts of developer conversations, and in doing so, contributed to the slow erosion of those very communities and their economic model—see Stack Overflow, which went from 200,000 questions a month at its peak to under 4,000 in December 2025, following a shift clearly triggered by ChatGPT.
But even in a world of coding agents, one problem remains wide open: how do we preserve and organize collective learning? Especially inside large enterprise organizations, running multiple teams, shared codebases, implicit best practices, and often the same problems solved again and again.
Without a shared memory, agents don't scale knowledge—they just scale repetition, unnecessarily burning tokens at machine scale while scaling inconsistency. Yes, a shared ecosystem of skills and other tools will certainly help, but just as equally-skilled developers need to share knowledge, cq and other agentic memory initiatives aim at building a shared agent knowledge commons, shifting from documentation to execution-driven learning systems, from isolated agents to collective cognition.
As written in the founding documentation: "Think of it as Stack Overflow, but written by agents, for agents, consumed by agents—with humans in the loop for governance and quality assurance."
Their move is not isolated. It follows several other memory-oriented agent add-ons. But in a way, it also echoes Mozilla's historical role in the web: defending open standards and a shared infrastructure against fragmentation and capture, by trying again to protect a layer of the digital stack as a commons.
