… (see outline.md)

Outline — cq - Stack Overflow for Agents

Speakers

• Peter Wilson — Staff Engineer at Mozilla.ai. Based in the North East of England, 20+ years in software engineering across security, infrastructure, and developer tooling. Formerly HashiCorp; Principal Engineer at NatWest; Architect at Sage.
• Davide Eynard — Staff ML Engineer at Mozilla.ai working on trustworthy AI. Interested in open applications of ML on federated systems. Self-describes as "Genetically a teacher, forever a student."

Abstract (as provided)

cq is Mozilla.ai's proposal and proof-of-concept for a standardised way to share knowledge across agents, locally, within a team/org and in a public commons.

The intention of cq is to allow human operators of agents to be less frustrated with the repeated mistakes due to things like outdated training data, stop agents wasting so many tokens bashing their heads against the wall, and let a lesson learned by one agent be useful to all of them.

This session will cover the idea for cq, a high level overview of how it works today, and what we learned building it. Finally we will talk about where we want to go with this, and where we think the community can be involved.

Thesis (synthesised)

Human contributions to Stack Overflow collapsed after ChatGPT, breaking a feedback loop that used to keep collective developer knowledge fresh. cq is Mozilla.ai's attempt to rebuild that loop for agents — a shared knowledge commons so a lesson learned by one agent benefits all of them — grounded in Mozilla Manifesto principles of openness, interoperability, and user-driven web experience.

Section TOC

Note: the available transcript covers only the opening framing. The promised content (cq design, how it works, lessons learned, future direction, community involvement) is not present in the supplied text.

Terminology glossary

• cq — Mozilla.ai's proposed standardised way to share knowledge across agents, locally, within a team/org, and in a public commons. (Definition from the abstract; not yet defined in the transcript text provided.)
• moz://a.ai — Mozilla.ai. The presenters explicitly distinguish this from Mozilla / Firefox: "are you the firefox guys? Or are you the browser? Guys? We don't want AI in the browser..."
• moz://a.ai/a URI — A "kind of fake URI" the presenters use as a demo: "It is kind of fake URI, but the browser folks can do anything they want with it. So you just connect to that URL and it will open the [Mozilla Manifesto]"
• Mozilla Manifesto — Referred to in the transcript as "the manifesto that has been set up like 20-ish years ago". Contains principles about the internet/web as an open resource, user agency, security, open code, standards, and interoperability.

Named frameworks / concepts

• Stack Overflow analogy — cq is framed as "Stack Overflow for agents"; the motivating observation is that human Stack Overflow contributions "basically plummeted after chat gbt came out".
• Mozilla Manifesto principles as foundation — openness, user-driven experience, security, open code, standards, interoperability — cited as the value frame cq operates within.

Open questions / not covered in the supplied transcript

The supplied transcript stops at the opening framing. The following topics from the abstract are not yet present in the text and should not be answered from outside knowledge:

• What cq actually is technically — protocol, format, storage model
• How cq works today (the proof-of-concept implementation)
• Lessons learned building it
• Where Mozilla.ai wants to take it next
• How the community can get involved
• Specific examples of agent mistakes cq would prevent
• Relationship to existing standards (e.g. MCP, agent memory systems)
• Privacy / trust model for the public commons