AI was front and center at WordCamp Europe, starting with a Contributor Day that brought together WordPress contributors from around the world to learn about and collaborate on the project, followed by two days of talks, workshops, and one-on-one discussions about the current and future state of WordPress and AI.
This post summarizes the team’s collective experience and feedback, with an eye towards increasing the impact we can have at future flagship events.
Event Recap
Contributor Day
The AI Team spanned 8 tables, with more than 70 participants joining in person over the duration of the event. @justlevine and @gziolo served as the table leads assisted by @psykro and @artpi in person, while @gajendrasingh acted as digital table lead providing support for folks on the ground and at least 6 digital-only participants.
To accommodate the variety of participant skills and interests – paralleled by the broad scope of #core-ai work – the team focused primarily on onboarding contributors to the various areas they were interested in and helping them identify meaningful ways to contribute. By lunchtime, the majority of participants were onboarded, leaving the afternoon free for collaboration and discussion among new and veteran contributors.
By the numbers, participants contributed 8 new GitHub Issues and 22 new PRs across 5 Core AI repos, Core, and Gutenberg. Beyond the numbers, contributors tested AI and MCP plugin features, evaluated Agent Skills, provided detailed feedback on usage, designed their own proof-of-concept plugins for various explorations, and even kicked off a new initiative to address the current gap around user-facing documentation.
WordCamp Europe
The opportunities to contribute span far beyond the Contributor Day tables. Team members held conversations with more than 80 individual WordPress Contributors and Committers, Speakers, plugin developers and product leads, SEOs, WCEU organizers and volunteers, site owners and end users about their needs, understandings, and use cases around WordPress and AI, alongside feedback about the Core AI team’s current and planned work.
Highlighted Discussions
- Guidelines & Context
- Participants discussed ways to balance the need to explore agentic context with future-compatibility, security, and the fast-moving evolution of context types.
- Participants agreed globally-scoped context made the most sense for the existing shared CPT, although a more self-documenting name could lend itself to expand to other holistic patterns.
- Agentic memory is evolving quickly and has to handle both globally-scoped learnings and user-scoped PII. While no consensus was reached on storage implementation, agreement was reached to experiment in parallel with Guidelines rather than block the MVP on it.
- Artifacts were distinguished from other forms of context and given a functional shape: a database-stored reference to other bit data. What remains is deciding what sorts of agentic work should be considered an Artifact, which can also be explored in parallel to globally-scoped Guidelines.
- Update: The Merge Proposal is live and ready for feedback!
- DESIGN.md
- Participants experimented with a working proof of concept and discussed paths towards inclusion as an AI or Gutenberg experiment and leveraging theme.json as the technical source of truth.
- User-facing docs
- Participants noted a lack in both end-user documentation and a way to contribute to such documentation, and kickstarted a new initiative to address both.
- MCP Client
- Multiple participants and attendees expressed interest in and explored potential use cases around a WordPress MCP client, including making external MCP tools available for use inside of WordPress and using it in conjunction with MCP Adapter as a centralized passthrough.
- Ability categorization, filtering, and slug fragments
- Contributors discussed the differences between structural hierarchy (slug fragments) and attributive classification (category metadata), as well as the programmatic implications in relation to Ability filtering, discovery and autocomplete, and progressive disclosure. Participants made little progress on making Ability Category metadata taxonomically useful, but agreed on the independent value of slug fragments and to continue discussing 7.1 inclusion on Trac.
- Security & Data Encryption
- Security was one of the most popular topics among contributors and WordCamp attendees, with many curious and some misinformed about the considerations for WordPress 7.0 and beyond.
- Discussions included: the damage radius of AI or AI credentials, the impact of PII and database grepping, supply chain mitigation like author provenance and efforts like Gandalf, and the needs and considerations of introducing a Data Encryption layer to Core. Team members got a demo and discussed the implications of Angie’s “Super Admin Mode” feature that can write plugins on the fly and read keys from wp-config.php.
- Education and AI Enablement
- Several conversations focused on the needs to educate users about using and developing with AI. While programs like WordPress Credits primarily target students, there is an even bigger need for people already in the ecosystem and those whose jobs AI is likely to disrupt.
- Participants discussed ways to “upskill” other contributors and committers without gatekeeping, and shared their varied experiences with company “Enablement” efforts. One suggestion was to create educational Handbook pages and AI Team blog posts that cover more foundational topics than the team’s own work. Next step: gather feedback from the Design and A11Y teams on their experiences.
- Friendlier times for non-US contributors
- Many contributors and attendees noted the prohibitive AMER scheduling of Core AI Team meetings, even as the majority of contributors and team members are not AMER based. Contributors offered numerous suggestions, from additional meeting slots to better async comms, which the team is now exploring.
What went well
- Lots of new Contributors and contributions. Participants reported walking away feeling comfortable about contributing in the future.
- The Contributor Day post successfully helped new contributors onboard and orient themselves (once they were made aware of it).
- No wifi issues with the venue.
- The event was an amazing opportunity to sync in person.
- The Table Leads deep familiarity with the Core AI Team efforts helped participants find contributions they were specifically interested in, in lieu of prepared tickets or a specific focus for the day.
Challenges
- Most attendees weren’t aware of the Contributor Day post, and needed to be personally directed to it.
- The low Table Lead -> participant ratio made it hard to track work and collaborate around specific efforts.
- The lack of a clear agenda or task list for the day made it difficult for participants to know what to work on:
- The breadth of participant experience and interest – and lack of pre-day insight into them – combined with the broad umbrella of #core-ai work made it hard to plan work in advance.
- Most #core-ai repos do not currently triage for Good First Issues or Help Wanted, and WordPress/ai’s tracker was mostly cleared out before the start of the day.
- While the personal direction for in-person contributors was net-positive for newcomers, digital contributors didn’t have that opportunity and left quickly.
- Along with with the lack of broader #core-ai roadmap, folks couldn’t just “crush tickets” or join in, but needed the specific guidance of the Table Leads.
Recommendations for the future
- Distill the Contributor Day post into a Handbook entry, and announce it to all participants at the start of the session.
- Prepare and share a Feedback Form for participants to complete before they leave.
- Collaborate with Contributor Day Organizers to try to get more insight regarding participant interests and needs, to tailor both onboarding efforts (a group session, sharing pre-recorded videos or links, etc) and particular focuses.
- Scale the number of Table Leads to the number of participants. Not all table leads require deep Core AI lore, just the ability to help onboard and facilitate participant contributions.
- Ensure there is an adequate number of GitHub issues and Trac tickets from which participants can self-select – especially from non-AI plugin repositories.
- Create a Core AI booth at WordCamp where attendees can get their broader questions answered and provide invaluable feedback on the team’s efforts. Contributor Day isn’t the right avenue for these discussions.
- Update: This is approved for WordCamp US – learn more and sign up to volunteer!
Props to @artpi @gajendrasingh @gziolo @karmatosed and @psykro for sharing their learnings and feedback and to @estelaris for prepublish review.