WP AI Progress 🔗 **Source:** [WP AI Progress](https://j.cv/wp-ai-progress/) > [!note]+ Document Notes > tags: [WordPress, AI, Governance] > author: J.CV > source: j.cv > reference: [https://j.cv/wp-ai-progress/] > date created: 2026-02-15 00:00 > > > # 🏷️ Title: WP AI Progress > <!-- deep-read:sections start --> ## 🧭 Table of Contents - [📝 Summary](#-summary) - [📖 Main Content](#-main-content) - [🗒️ Notes](#️-notes) - [📚 References](#-references) - [🗣️ Paraphrase](#️-paraphrase) - [📰 Newsletter](#-newsletter) - [✅ Action Items](#-action-items) - [🧪 Hypotheses](#-hypotheses) - [📊 Methodology](#-methodology) - [❓ Questions](#-questions) - [🔮 Future Research](#-future-research) - [💡 Implications](#-implications) - [🧠 Learning Styles Assessment](#-learning-styles-assessment) - [📊 Accessibility Assessment](#-accessibility-assessment) - [🌍 DEIB Assessment](#-deib-assessment) - [✅ Logical Validity](#-logical-validity) - [🛡️ DARVO Analysis](#-darvo-analysis) - [🎯 Conclusion](#-conclusion) ### 🔗 References - Include links cited in or relevant to the article. --- ## 📝 Summary - What happened: The author catalogs a “week of AI” in WordPress, highlighting rapid progress across core, Gutenberg, plugins, education, and benchmarking for AI and agents. - Why it matters: These are the last core “building blocks” to make WordPress a first-class platform for agentic, LLM-powered workflows. - Key stakeholders: WordPress core contributors, plugin/theme authors, hosting providers, educators, and AI tool vendors (e.g., Automattic, Anthropic). - Risks and tradeoffs: Governance, responsible-AI usage, privacy, performance, and the possibility of fragmented or unsafe agent capabilities if not guided by clear norms like the new AI Guidelines. - Bottom line: WordPress is quickly becoming a practical, open hub for AI agents—through core APIs, plugins, and educational programs—much faster than even optimists expected. --- ## 📖 Main Content - Core arguments: - WordPress has seen “an explosion of visible AI progress” in just one week, spanning core APIs, plugins, and ecosystem initiatives. - A proposal to bring LLM querying directly into WordPress core is described as “the last of the AI building blocks,” positioning WordPress strongly for the “agentic web,” where autonomous agents act on behalf of users. - There is already a “mature and impressive MCP implementation for WordPress,” making it easier to build agents that call WordPress tools. - Automattic is using this stack to ship an official Claude connection with Anthropic, showing real ecosystem partnerships, not just experiments. - The Abilities API (landing in WordPress 6.9 and Gutenberg 24.0) plus early Workflows API progress give WordPress a more formal way to express what agents can do and how. - Educational and evaluation infrastructure is emerging: a WordPress-focused “AI Leaders” micro-credential and a WordPress-specific benchmarking environment / RL playground for models. - An “absolutely absurd amount of AI experiments” are happening via an AI experiments plugin already nearing 1,000 commits, including: - Content guidelines in Gutenberg, enabling sites to encode editorial norms to guide both human authors and agents. - Markdown rendering so agents can request Markdown versions of pages. - Request logging for AI, especially important as a standardized WP AI client comes closer to core. - LLM-powered editor tools like typeahead. - A Service Account exploration to give agents their own user type in WordPress. - Governance and safety are being addressed with: - AI Guidelines (v0) for WordPress contributors to clarify how AI can help the project. - Agent Skills for WordPress, a common repo of skills for tools like Claude, Cursor, and CodeX so agents can interact with WordPress at the code level effectively and safely. - The author emphasizes the pace: even three years ago this rate of progress felt unimaginable; much of it matches what they wished for while running an AI-related startup. - Key concepts and definitions: - Agentic web: An emerging pattern where AI agents perform tasks across sites and services, often autonomously, guided by user intent. - MCP (Model Context Protocol): A protocol for tools/services to expose capabilities to LLMs in a structured way, allowing agents to call “tools” (like WordPress actions) reliably. - Abilities API: A WordPress API to describe what actions are possible, which can be read by agents and tools. - Workflows API: A higher-level system for composing actions into workflows, important for agent orchestration. - Service Account: A special user type meant for automated agents, distinct from normal human users. - Uncertainties / contested areas: - Exactly how LLM querying will be integrated into core and what guardrails will ship by default. - Long-term governance of Agent Skills and AI Guidelines: who maintains them, how they evolve, and how they handle security/abuse concerns. - Performance and cost tradeoffs of embedding AI features into common WordPress setups, especially on low-cost hosting. - How quickly the average WordPress site owner will adopt agentic workflows, versus advanced agencies and power users. --- ## 🗒️ Notes - The author opens by saying that in the past week there has been “an explosion of visible AI progress in the WordPress project,” listing multiple concrete milestones in one place for context. - A proposal to bring the ability to query LLMs directly into WordPress core is framed as one of the “big headlines” and “the last of the AI building blocks,” suggesting that once this lands, WordPress has all primitives needed for agentic behavior. - A “mature and impressive MCP implementation for WordPress” already exists, making it easier for agent frameworks to talk to WordPress. This is crucial because MCP is quickly becoming an important interface standard for LLM tools. - Automattic uses this MCP implementation to build an “official Claude connection in partnership with Anthropic,” indicating that first-party integrations (not just community hacks) are driving adoption. - The Abilities API is landing in WordPress 6.9, with frontend abilities in Gutenberg 24.0, plus some progress on the Workflows API. Together, these give structured ways for code and agents to express what actions are possible (abilities) and how they can be sequenced (workflows). - The first WordPress-focused micro-credential, “AI Leaders,” is introduced to help students build “practical AI skills” in the context of WordPress, signaling a formal educational and credentialing track for AI in the ecosystem. - A benchmarking system is being created for Frontier Labs and hobbyists to measure model performance on “WordPress specific tasks” and to serve as a reinforcement learning environment for improving/fine-tuning models; this is explicitly pitched as an “announcement post.” - The author mentions “an absolutely absurd amount of AI experiments” and calls out favorites, starting with Content Guidelines in Gutenberg. These guidelines are “not AI specific,” but let sites define editorial terms and reasons, helping steer both internal agents and potential future “site delegates.” - Markdown rendering is available so agents can ask for Markdown versions of pages using multiple methods, making content consumption and transformation easier for LLMs. - Request logging for AI is highlighted as especially helpful as the “WP AI client gets closer to core,” giving observability into how agents and LLM calls behave. - Gutenberg sees “a lot of cool editor improvements” using LLMs, with typeahead called out as a favorite. This likely refers to predictive/autocomplete capabilities powered by an LLM. - There’s an exploration of a “Service Account in WordPress” to allow agents to have their own dedicated user type, rather than overloading human users or anonymous roles. - The AI experiments plugin is “rapidly approaching 1000 commits in its short lifespan,” evidence of intense experimentation and community interest. - AI Guidelines (v0) for WordPress are published to give “thousands of contributors a black and white guide on how to use AI to improve the project,” indicating early but explicit governance. - Agent Skills for WordPress is introduced as a repository of skills usable in Claude, Cursor, CodeX, etc., for better code-level interaction with WordPress. - The author closes by reflecting that even three years ago they wouldn’t have imagined this rate of progress, and that these developments line up with what they wanted when they were running a startup in this space; they end with “Let’s keep pushing,” inviting continued momentum. --- ## 📚 References - WP AI Progress (original article): https://j.cv/wp-ai-progress/ - WordPress Core / Gutenberg AI initiatives (general overview): https://make.wordpress.org/core/tag/ai/ - Gutenberg project: https://github.com/WordPress/gutenberg - Automattic: https://automattic.com/ - Anthropic Claude: https://www.anthropic.com/claude --- ## 🗣️ Paraphrase - Paragraph A paraphrase: The author lists a burst of recent AI developments in WordPress—core LLM querying, MCP integration, a Claude connection, Abilities and Workflows APIs, an AI micro-credential, benchmarking tools, and multiple AI-focused Gutenberg and plugin experiments—arguing these constitute the final pieces needed for an agent-friendly WordPress ecosystem. - Paragraph B paraphrase: Alongside the technical progress, WordPress is adding governance and shared resources like AI Guidelines and an Agent Skills repository; the overall pace of change far exceeds what the author expected a few years ago, and they urge the community to keep building. --- ## 📰 Newsletter In the last week, WordPress quietly accelerated into the AI fast lane. We’re seeing a proposal to bring LLM querying into core, a maturing Abilities API (coming in 6.9) plus frontend support in Gutenberg 24.0, and even early Workflows API work. Combine that with a solid MCP implementation and Automattic’s official Claude connection, and WordPress is positioning itself as a first-class host for AI agents. On the experiments side, the AI Experiments plugin is racing toward 1,000 commits, shipping features like content guidelines inside Gutenberg, AI-powered typeahead, markdown rendering for agent-friendly content, and request logging for AI calls. There’s also an exploration of a dedicated Service Account role so agents can have clearly scoped identities. Education and evaluation are catching up fast: the new “AI Leaders” micro-credential aims to give students practical, WordPress-specific AI skills, while a WordPress-focused benchmarking environment will help both Frontier Labs and hobbyists measure model performance on real WP tasks. If you build on WordPress, this is the moment to: - Track the Abilities and Workflows APIs in core. - Experiment with the AI Experiments plugin in a staging environment. - Start thinking about how your plugins and themes can expose “agent skills” safely. The agentic web is arriving—and WordPress is increasingly ready to host it. --- ## ✅ Action Items #### DevRel - Developer advocacy: Prepare talks and demos showing how to use the Abilities API, MCP implementation, and the upcoming WP AI client; measure success via adoption in example plugins and PR references. - Education: Design workshops around the “AI Leaders” micro-credential focusing on building AI-powered blocks and safe agent workflows; run office hours for agencies. - Developer experience: Create starter repos and quickstart templates that wire up Claude (via the official connection) and the Abilities API; improve docs with code examples. - Developer marketing: Publish case studies of early adopters using agent workflows; run a content series on “Agent-ready WordPress” with clear funnels to tutorials. - Community management: Facilitate discussions around AI Guidelines; gather contributor feedback on pain points with AI tools and surface them to core teams. - Internal/external/cross-project: Coordinate with MCP and tooling communities (Claude, Cursor, etc.) to keep WordPress Agent Skills well-maintained and discoverable. #### OSPO - Policy & Governance: Draft internal policies on using the new AI Guidelines for contributions; clarify decision rights around AI features in shared WordPress projects. - License Compliance: Verify MCP, Agent Skills, and AI-related tooling align with WordPress GPL norms; schedule periodic scans for AI-related dependencies. - Security Management: Establish a process for reviewing new agent skills and Abilities definitions for privilege escalation or abuse risks. - Software Inventory: Track which WordPress sites and plugins in the organization are using AI/agent features; maintain an updated SBOM including AI components. - Developer Support: Create guidelines and regular office hours to help teams adopt AI features responsibly, with legal and security guardrails. - Community Engagement: Consider funding or contributing to WordPress AI benchmarking infrastructure and the AI Guidelines evolution. #### Community - Maintainers: Run roadmap clinics on how Abilities and Workflows APIs affect plugin architectures; encourage small, incremental AI feature additions with clear change logs. - Contributors: Offer “first issue” AI tickets (e.g., adding logging or content guidelines support) to help new contributors get involved safely. - Sponsors: Align funding to high-value AI infrastructure—benchmarking, governance, and docs—rather than only flashy demos; publish transparent reports. - Ecosystem: Coordinate with other open-source CMSes on agent safety norms, benchmarking standards, and shared tools to avoid duplicated effort. #### WordPress - Core: Prioritize the LLM querying proposal and WP AI client integration, with clear performance and safety criteria for inclusion. - Core Performance: Profile AI-related endpoints and editor features; set budgets for latency and error rates for AI calls. - Core AI: Develop responsible AI guardrails, evaluation sets, and pilot projects around Abilities, Workflows, and Service Accounts. - **Core Program Team:** Set regular check-ins across Core, Gutenberg, Hosting, Polyglots, Docs, and Security to align on AI timelines and communication. - Mobile: Explore how Abilities and agent workflows surface in mobile apps; design offline-friendly behaviors and clear AI UX. - Playground: Use WordPress Playground to provide a browser-based lab for trying AI features and agent skills without local setup. - Community: Provide meetup and WordCamp kits for running AI-in-WordPress sessions; support speaker mentorship on responsible usage. - Training: Build lesson plans that map directly onto the “AI Leaders” micro-credential; consider certification pilots for agencies. - Marketing: Shape narratives around WordPress as an open, agent-friendly platform, emphasizing governance and user control. - Support: Update knowledge base content on AI features; create clear troubleshooting guides for AI errors and misbehavior. - Design: Audit AI-related interfaces for clarity, consent, and user control (e.g., which actions agents may perform). - Documentation: Document Abilities API, Workflows API, AI Experiments plugin features, and Service Accounts with examples-first patterns. - Photos: Consider how media AI tools (e.g., alt-text generation) integrate with WordPress Photos; maintain moderation standards. - TV: Publish talks and walkthroughs on AI initiatives; ensure captions and transcripts for accessibility. - Polyglots: Coordinate AI-assisted translation workflows while maintaining glossary and style consistency; define where automation is acceptable. - Accessibility: Review AI features for WCAG compliance, focusing on clarity of AI vs human content and controls. - Meta: Improve .org tooling for discovering AI-enabled plugins/themes; add filters and documentation flags. - CLI: Add WP-CLI commands to inspect abilities, workflows, and AI logs; ship plugin scaffolds that include AI best practices. - Hosting: Work with hosts on sane defaults for AI features, including rate limits, logging, and telemetry to track reliability. - Tide: Extend scoring to consider AI-related security practices and logging; loop in plugin authors with reports. - Openverse: Explore AI-powered search and moderation while preserving strong provenance; document data use transparently. - Themes: Encourage theme authors to test with agent-driven content flows; add testing matrices for AI-heavy sites. - Plugins: Improve review guidance for AI features, particularly around data sharing, logging, and permission scopes. #### The WPOCC - Programs: Create fellowships focused on AI-related core components (Abilities, Workflows, AI Guidelines, Agent Skills). - Projects: Sponsor documentation and governance sprints around AI Guidelines and Service Accounts. - Funding experiments: Pilot microgrants for maintainers of AI infrastructure plugins, benchmarking tools, and educational content. ### FAIR - Technical independence: Encourage local logging and model metadata tracking for AI interactions; explore signing and provenance for Agent Skills. - Policies and covenants: Promote funding transparency for AI initiatives and sunset clauses for experimental AI APIs where needed. - Productized governance artifacts: Turn AI Guidelines, agent role definitions, and skills review criteria into reusable charters and scorecards. --- ## 🧪 Hypotheses - H1: Introducing LLM querying into WordPress core will significantly increase the number of plugins offering agentic workflows within one year. - H2: Standardized Abilities and Workflows APIs will reduce security incidents related to ad hoc AI integrations. - H3: A shared Agent Skills repository will improve developer productivity when adding AI features to WordPress plugins/themes. - H4: The “AI Leaders” micro-credential will increase learner engagement with WordPress compared to generic AI courses. - H5: WordPress-specific benchmarking will reveal that general-purpose LLMs underperform on WP tasks without fine-tuning or skills. --- ### 📊 Methodology - For H1: Track the number of plugins declaring AI/agent capabilities in the plugin directory before and after core LLM support; control for general AI trend growth. - For H2: Collect and categorize reported security issues related to AI plugins over time; compare projects using Abilities/Workflows vs those with custom approaches. - For H3: Survey developers using Agent Skills; measure time-to-ship for features versus teams building from scratch. - For H4: A/B test “AI Leaders” against a control group taking a non-WordPress AI course; compare completion rates and subsequent contributions or projects. - For H5: Use the WordPress benchmarking environment to evaluate base models vs fine-tuned/skill-augmented models on standardized tasks (e.g., block development, content workflows). --- ### ❓ Questions - How will WordPress core balance flexibility for agents with strong defaults around permissions and logging? - What governance model will sustain the Agent Skills repository over time? - How do non-technical site owners consent to, understand, and revoke agent abilities on their sites? - Can WordPress become a reference platform for open agent safety norms, not just tooling? - How will hosting providers adapt pricing and resource limits to AI-heavy WordPress sites? --- ### 🔮 Future Research - Model evaluation focused specifically on CMS and site-management tasks. - Best practices for representing site-specific editorial guidelines as machine-readable policies. - UX research on how users perceive and control agents operating in their dashboards. - Comparative studies of agent frameworks integrating with WordPress vs other CMSes. - Longitudinal impact of AI micro-credentials on open-source contribution patterns. --- ### 💡 Implications - Practice: Plugin and theme authors need to think in terms of abilities, workflows, and agent skills, not just UI widgets. - Policy: Projects should adopt AI Guidelines-like documents to guide contributor behavior and tool usage. - Security: Service Accounts and Abilities APIs could become critical security boundaries; misconfiguration could be risky. - Education: WordPress-specific AI curricula may draw new contributors into both AI and open-source. - Funding: Infrastructure like benchmarking, guidelines, and shared skills is a strong candidate for philanthropic or corporate funding. --- ### 🧠 Learning Styles Assessment - Reading level estimate: Approx. Flesch-Kincaid Grade 9–11 (conceptually advanced but written in relatively straightforward language). #### 👀 Visual learners - Present: Mainly text; some implicit structure via bullet points and lists of features. - Improve: Add architecture diagrams for Abilities/Workflows, flowcharts of agent interactions, and before/after editor screenshots. #### 🔊 Aural learners - Present: No native audio. - Improve: Provide a short audio summary or podcast-style discussion of the week’s AI progress. #### 🗣️ Verbal learners - Present: Clear, concise prose; repeated use of key terms like “abilities,” “agents,” “guidelines.” - Improve: Add a glossary of AI terms as used in WordPress; more analogies for non-technical readers. #### 🧩 Kinesthetic learners - Present: Implied calls to experiment with plugins and APIs. - Improve: Step-by-step tutorials, labs using WordPress Playground, and checklists for enabling specific AI features. #### 🧠 Logical learners - Present: Implicit timeline and causal narrative (APIs + skills + guidelines → agentic web readiness). - Improve: Explicit tables comparing capabilities before/after these changes; decision trees for when to use which tools. #### 🧑‍🤝‍🧑 Social learners - Present: Reference to wide contributor base and community guidelines. - Improve: Discussion prompts for meetups, community calls on AI Guidelines, and cohort-based learning around “AI Leaders.” #### 🧘‍♀️ Solitary learners - Present: A short, dense article suited to individual reflection. - Improve: Reflection questions at the end (e.g., “What AI workflow would you build on WordPress first?”). #### 🌿 Naturalistic learners - Present: No direct environment/sustainability framing. - Improve: Discuss energy and cost implications of AI features and how to design greener workloads. --- ### 📊 Accessibility Assessment #### Screen reader semantics - Present: Simple article structure with headings and lists is likely screen-reader friendly. - Improve: Ensure semantic HTML5 landmarks and descriptive link text throughout. #### Keyboard navigation - Present: Standard blog layout likely works with basic keyboard navigation. - Improve: Confirm visible focus states, skip links, and logical tab order in theme. #### Contrast and readability - Present: Depends on site theme; text content is brief and scannable. - Improve: Verify contrast ratios and line height; offer a high-contrast mode if theme doesn’t already. #### Text resizing and reflow - Present: Responsive modern theme likely supports zoom and reflow. - Improve: Test at 200% zoom; ensure no clipping of code or lists. #### Media and captions - Present: Mostly text, minimal media. - Improve: If future posts embed videos of AI demos, ensure captions and transcripts. #### Forms and interactions - Present: Limited to standard site forms (e.g., comments). - Improve: Ensure labeled fields, accessible error messages, and adequate target sizes. #### Standards reference - Map gaps to WCAG 2.2 AA: particularly 1.4.x (contrast), 2.4.x (navigation), 3.3.x (input assistance). --- ### 🌍 DEIB Assessment #### Cultural breadth - Present: Global WordPress community is mentioned implicitly, but examples are fairly generic/Western. - Improve: Highlight AI use cases from diverse regions and linguistic communities. #### Inclusive language - Present: Neutral, technical language; minimal jargon beyond AI/WordPress terms. - Improve: Add plain-language summaries and avoid assuming deep AI background knowledge. #### Neurodiversity support - Present: Short, list-heavy structure can help some readers. - Improve: More consistent subheadings, icons + text, and predictable patterns across related posts. #### Representation - Present: No imagery or specific personas referenced. - Improve: When adding examples or images, represent diverse roles and identities. #### Psychological safety - Present: Encouraging, non-alarmist tone about AI. - Improve: Clear guidelines on acceptable AI usage and where to raise concerns about misuse. --- ### ✅ Logical Validity #### Strong reasoning - Inductive: The author cites multiple concrete initiatives (APIs, plugins, benchmarks, guidelines) to support the claim that AI progress in WordPress is accelerating rapidly; the examples are specific and varied, strengthening the inference. - Causal: Suggests that having querying, abilities, workflows, and skills in place will position WordPress for the agentic web; the causal chain is plausible given how agents operate (they need stable APIs and capabilities). #### Fallacies - None obvious; the piece is descriptive rather than argumentative. The main risk is mild survivorship bias (highlighting successes without discussing failed experiments), but that’s acceptable for a short progress report. #### Tone of voice - Optimistic, concise, and slightly informal: phrases like “absolutely absurd amount of AI experiments” and “Let’s keep pushing” convey enthusiasm and momentum. --- ### 🛡️ DARVO Analysis - Deny: No denial or minimization of harms; the post doesn’t attempt to downplay risks. - Attack: No attacks on critics or alternative approaches; it’s focused on progress, not debate. - Reverse victim/offender: No attempt to cast any party as a victim of criticism or regulation. - Overall: The article is informational and promotional rather than defensive; DARVO patterns are not present. --- ## 🎯 Conclusion WordPress is rapidly assembling the last major pieces needed to support AI agents: core LLM querying, structured Abilities and Workflows APIs, mature integrations (like MCP and Claude), active experimentation, and emerging governance and education (AI Guidelines, AI Leaders, Agent Skills, benchmarking). Practical next steps for readers: - If you’re a developer, start exploring the Abilities API, AI Experiments plugin, and Agent Skills to prepare your plugins/themes for agents. - If you’re a community organizer or educator, consider how “AI Leaders” and AI Guidelines can shape local meetups, workshops, and contributor onboarding. - If you’re in leadership or hosting, begin planning for governance, logging, and performance around AI features on your WordPress infrastructure. --- ## 🧱 Refactor Appendix (append after 🎯 Conclusion) ### 📚 Literature Split Pad - Purpose: Stage Konik CEE units that become per-source Literature notes. - Split level: **H3**. - Include verbatim quotes under 25 words and a [view highlight] link. ### [H1] 5–10-word present-tense claim > [!quote] [WP AI Progress](https://j.cv/wp-ai-progress/) by [[J.CV]] via [[j.cv]]. Published [MISSING]. Accessed 2026-02-15. > “In the the past week there has been a explosion of visible AI progress in the WordPress project” Two lines: context and use. Cite data if relevant. [view highlight](https://j.cv/wp-ai-progress/#H1) #xref #articleseed ID:rw_[MISSING] ### [H2] 5–10-word present-tense claim > [!quote] [WP AI Progress](https://j.cv/wp-ai-progress/) > “Even three years ago I would have never imagined this rate of progress in this project.” Two lines: context and use. [view highlight](https://j.cv/wp-ai-progress/#H2) #xref ID:rw_[MISSING] --- ### 🔖 Atomic Split Pad (Zettel Candidates) - Purpose: Promote reusable claims into Atomic notes. - Split level: **H4**. - Each item: one idea, present tense, specific, testable. #### [Z1] WordPress rapidly becomes agent-ready platform - Scope/condition in one line. - Pointer to page/time or dataset. - Links: [[WordPress-AI]] [[Agentic-Web]] #### [Z2] Shared agent skills improve WP AI development - Scope/condition in one line. - Pointer to page/time or dataset. - Links: [[Agent-Skills]] [[Developer-Experience]] --- ### ✏️ Fleeting Queue (Extract by Selection) - Purpose: Capture quick actions or sparks for Fleeting notes. - Extraction: **Select one line** → “Refactor: Extract selection to new note (content only).” - Keep each line short and self-contained. - [F1] Explore Abilities and Workflows APIs for plugin roadmap — context/https://j.cv/wp-ai-progress/ #fleeting #inbox - [F2] Design meetup session on WordPress AI Guidelines — context/https://j.cv/wp-ai-progress/ #fleeting #inbox - [F3] Prototype Agent Skills for my existing plugin — context/https://j.cv/wp-ai-progress/ #fleeting #inbox --- <!-- deep-read:sections end -->