I missed what EmDash lets agents do.
Yesterday I wrote that Cloudflare's EmDash was "a product out of time, solving yesterday's problem with yesterday's architecture." Twenty-four hours later, I'm partially walking that back -- not because the architecture changed, but because I was looking at the wrong part of it. The security sandbox isn't the story. The agent-native CMS is. And the Astro acquisition pipeline behind it reveals a strategic play that's more sophisticated than I gave Cloudflare credit for.
Scoring My Own Predictions
I don't write follow-ups to my own posts 24 hours later. I especially don't write follow-ups that require me to acknowledge I missed something. But the response to yesterday's piece -- "The WordPress Succession Problem" -- and the additional research I've done since publishing it have convinced me that I got the broad thesis right while underweighting the most interesting angle.
Let me be specific. Here's how the predecessor post's claims hold up:
| Prediction | Verdict | Notes |
|---|---|---|
| EmDash's plugin sandboxing only works fully on Cloudflare | Confirmed | Self-hosted Node.js loses V8 isolate sandboxing. No one disputes this. |
| The WordPress moat is economic/social, not technical | Confirmed | HN's most upvoted criticism thread focused on ecosystem, not engineering.hn-thread |
| Every prior "WordPress killer" plateaued below 1% | Confirmed | Ghost: 0.1% after 13 years. The graveyard keeps growing.ghost |
| WordPress won't be displaced by a better CMS | Still believe this | But I now think EmDash isn't trying to be a better CMS. More on this below. |
| EmDash is "solving yesterday's problem with yesterday's architecture" | Partially wrong | The security framing is yesterday's problem. The agent-native architecture is not. |
The security narrative -- which Cloudflare leads with, and which I spent most of my first post dismantling -- is a red herring. Not because plugin security doesn't matter. It does. 11,334 new WordPress vulnerabilities in 2025, a 42% increase over the previous year, with 1,966 high-severity findings -- more than the previous two years combined.patchstack-2026 Those numbers are real and they are ugly. But security has never been why people choose a CMS. If it were, WordPress wouldn't power 43.3% of the internet in the first place.
What I missed -- what Cloudflare's own blog post buries under paragraphs of security statistics -- is the part of EmDash that actually matters for where the web is going.
The security narrative is a red herring. Not because plugin security doesn't matter -- it does -- but because security has never been why people choose a CMS.
The Cloudflare Playbook: Buy the Framework, Build the CMS, Own the Platform
Let me lay out the timeline, because the timeline tells a story that Cloudflare's marketing copy carefully avoids telling.
January 16, 2026: Cloudflare announces the acquisition of The Astro Technology Company.cloudflare-astro Astro -- the content-optimized web framework that topped State of JS satisfaction rankings for three consecutive years, grew weekly npm downloads from 360,000 to 900,000 in 2025 alone, and was already powering platforms like Webflow and Wix on Cloudflare infrastructure.astro-growth The Astro team joins Cloudflare's payroll. The framework stays MIT-licensed and open source. Fred Schott, Astro's CEO, says joining Cloudflare will let them "accelerate Astro's development faster and on a much larger scale."
April 1, 2026: EmDash launches. Built on Astro 6.0. Running on Cloudflare Workers. Using D1 for the database, R2 for storage, Dynamic Workers for plugin sandboxing, Workers AI for security scanning.
That's less than three months from acquisition to product launch. Either Cloudflare built a CMS from scratch in two months -- which they effectively admit by saying it was substantially AI-generated -- or this was planned before the acquisition closed. I suspect both are true. The Astro acquisition wasn't about making a better static site generator. It was the first move in a vertical integration play, and EmDash was always the target.
Here's what Cloudflare now owns, end to end:
| Layer | Product | Role |
|---|---|---|
| Framework | Astro 6.0 | Rendering, themes, components |
| CMS | EmDash | Content management, admin panel |
| Compute | Workers | Serverless runtime for the application |
| Database | D1 | SQLite at the edge |
| Storage | R2 | S3-compatible object storage (zero egress fees) |
| Plugin Runtime | Dynamic Workers | V8 isolate sandboxing for extensions |
| AI | Workers AI | Security scanning, content classification |
| CDN/Security | Cloudflare (core) | DNS, WAF, DDoS protection |
This is not a CMS announcement. This is a platform stack announcement. Cloudflare is building the Android of web publishing: open source the software layer to drive adoption, control the distribution and infrastructure layer to capture revenue. The MIT license is the Trojan horse. It says "you can run this anywhere." The architecture says "but you'll want to run it here."
I called this pattern out yesterday. What I didn't appreciate was the completeness of it. Cloudflare isn't just building a CMS that works best on their platform. They're building an entire vertically integrated web publishing stack where every layer is a Cloudflare product. The CMS is the tip of the iceberg. The iceberg is infrastructure revenue.
Cloudflare isn't building a CMS that works best on their platform. They're building an entire vertically integrated web publishing stack where every layer is a Cloudflare product.
The economic model makes the play explicit. EmDash on Cloudflare uses pay-per-use billing with scale-to-zero -- you pay only for the CPU time your site actually consumes. No idle costs. A blog that gets 100 visitors a month costs effectively nothing. Compare that to traditional WordPress hosting at $5-50/month even for sites that sit dormant. For the long tail of the web -- the millions of small sites that make up WordPress's install base -- that pricing model is genuinely disruptive. Not because it's cheaper on average, but because it eliminates the floor.
And here's the strategic insight that connects the Astro acquisition to the pricing model: Cloudflare already handles a significant percentage of WordPress traffic as a CDN and WAF provider. They already protect these sites. The move from "protecting WordPress sites" to "replacing WordPress sites" is a lateral revenue expansion, not a market entry. The customer relationship already exists. The infrastructure already exists. The sales motion is: "You're already paying us to put a security layer in front of your WordPress site. What if we just were the site?"
The Agent-Native CMS: What I Actually Missed
Here's where I need to course-correct from yesterday's analysis, and this is the part that matters most.
Joost de Valk -- the founder of Yoast SEO, arguably the most popular WordPress plugin ever built, installed on over 5 million active WordPress sites -- published a technical review of EmDash and used a phrase that I think captures something fundamental: EmDash is "agent-native, not agent-compatible."joost
The distinction is important. "Agent-compatible" means you can bolt an AI integration onto an existing system. WordPress has agent-compatible plugins -- chatbots, AI writing assistants, automated SEO tools. They work around WordPress's architecture, parsing HTML, navigating admin panels, reverse-engineering APIs that were designed for humans. "Agent-native" means the system was designed from the ground up so that AI agents are first-class consumers of its data and operations.
EmDash is agent-native in three specific ways, and each of them represents a genuine architectural decision that WordPress cannot replicate without rebuilding from scratch.
Structured JSON, Not HTML Blobs
WordPress stores content as serialized HTML strings in a MySQL database. If you've ever looked at a WordPress post's raw database entry, you know what I mean -- it's a string of HTML markup, sometimes with shortcode tags embedded, sometimes with inline styles, sometimes with WordPress's block comment syntax wrapping Gutenberg blocks. It's a format designed for one thing: rendering in a browser.
EmDash stores content as Portable Text -- structured JSON.joost Each piece of content is a typed data structure with fields, not a blob of markup. A blog post isn't a string of \<p>\ tags. It's a JSON object with a title field, a body field containing an array of typed blocks, metadata fields with structured values, and relationships to other content objects.
This sounds like a nerd distinction. It isn't.
When an AI agent needs to read, modify, or create content in WordPress, it has to parse HTML. It has to understand WordPress's idiosyncratic block comment syntax. It has to handle shortcodes. It has to navigate a database schema that was designed in 2003 and has been backward-compatible-extended for 24 years. This is why every AI-WordPress integration is brittle -- it's working against the architecture, not with it.
When an AI agent needs to read, modify, or create content in EmDash, it reads and writes JSON. The content is already in the format that AI models natively consume and produce. There is no parsing step. There is no lossy conversion. The content is machine-readable by default, not machine-readable as an afterthought.
| Aspect | WordPress | EmDash |
|---|---|---|
| Content storage | Serialized HTML strings | Portable Text (structured JSON) |
| Content modification | Parse HTML, modify DOM, re-serialize | Read JSON, modify fields, write JSON |
| Schema access | Reverse-engineer wp_posts table | Typed SQL schema with Kysely abstraction |
| API for agents | REST API (designed for humans) | MCP server (designed for machines) |
| Content portability | Export as WXR (XML) | JSON is natively portable |
| Block format | HTML comment syntax (Gutenberg) | Typed JSON block objects |
When an AI agent needs to interact with WordPress, it has to parse HTML designed for browsers. When it needs to interact with EmDash, it reads and writes JSON. That difference is more important than any security sandbox.
The MCP Server
EmDash ships with a built-in Model Context Protocol server.joost For anyone not tracking the MCP ecosystem -- and I've been tracking it closely -- this is the protocol that lets AI tools like Claude and ChatGPT interact directly with external systems through a standardized interface. It's the difference between an AI agent that can talk about your website and an AI agent that can operate your website.
With EmDash's MCP server, an AI agent can:
- Create, read, update, and delete content
- Manage media assets
- Modify site configuration
- Build and deploy themes
- All through a programmatic interface designed for machine consumption
WordPress has REST APIs. They were designed for JavaScript frontends, mobile apps, and third-party integrations -- human-directed clients. EmDash's MCP server was designed for AI agents as the primary consumer. The difference isn't capability -- WordPress APIs can do most of these things -- it's ergonomics. The MCP server speaks the protocol that AI agents already understand.
Agent Skills and AI-First Migration
Here's the detail that made me reconsider my yesterday thesis that EmDash is "solving yesterday's problem": EmDash includes what it calls "Agent Skills" -- structured guidance for AI tools on how to perform complex operations like WordPress theme migration, custom block building, and plugin creation.joost
The WordPress migration path illustrates this philosophy. Content migration is handled by a standard WXR import tool or a REST API exporter plugin -- table stakes, nothing novel. But theme migration -- the hard part, the part that historically requires a developer to manually rebuild everything in the new system -- is handled by giving AI agents a structured skill guide for converting WordPress themes to Astro components. The Block Kit Agent Skill does the same for WordPress blocks: instead of engineering a compatibility layer, EmDash says "here's how to teach an AI agent to do the conversion."
This is clever in a way I didn't initially appreciate. Every previous WordPress alternative tried to solve the migration problem with engineering -- build importers, build compatibility layers, build bridges. EmDash outsources the migration work to AI agents. "Don't rebuild WooCommerce. Let Claude do it." That's aspirational, not proven. But it's a fundamentally different approach, and it's one that gets more viable as AI capabilities improve.
The Yoast Founder Is Experimenting With Leaving WordPress
I want to dwell on the Joost de Valk signal for a moment, because I think the tech press is underweighting it.
Joost de Valk built Yoast SEO. Five million active installations. One of the defining plugins of the WordPress ecosystem. He has as much invested in WordPress's success as almost anyone alive. And in his EmDash review, he says that EmDash reflects what he "would design if starting from scratch: sandboxed plugins, structured content, TypeScript throughout, edge deployment."joost
More importantly: he spent an evening building an EmDash theme for his own blog and started migrating content.joost
He's not naive about the challenges. He lists them explicitly: "The plugin ecosystem is empty." He invokes the graveyard: "Ghost, Craft, Statamic; all technically excellent, none have built the ecosystem needed for broad adoption." He flags governance uncertainty: "developers will want to see long-term governance commitments before investing their time." He notes that two months of AI-assisted development is not two months of battle-testing.
But the man who built the most popular WordPress plugin in history is experimenting with leaving WordPress. That's a data point. It doesn't tell you EmDash will succeed. It tells you how deep the discontent runs.
The Mullenweg/WP Engine crisis is the catalyst, but the underlying pressure has been building for years. WordPress.org was revealed to be Matt Mullenweg's personal property -- not the WordPress Foundation's -- in Automattic's own legal filing.repository-governance 159 Automattic employees left via severance package, with roughly 80% from the WordPress/Ecosystem division.techcrunch-wpengine 20 core contributors signed an open letter citing the Foundation's lack of oversight.repository-governance Mullenweg deactivated WordPress.org accounts of governance critics.register-bans The jury trial is scheduled for February 2027.techcrunch-wpengine
When the creator of Yoast SEO is building EmDash themes on his own blog, and 159 Automattic employees have walked out the door, and the project lead's governance decisions are being adjudicated in federal court -- that's not a "WordPress killer" narrative. That's an ecosystem in stress, and EmDash is the most visible pressure valve.
When the creator of Yoast SEO is experimenting with leaving WordPress, that's not a "WordPress killer" narrative. That's a data point about how deep the ecosystem discontent runs.
The Hacker News Verdict: Technically Interesting, Strategically Doomed
The Hacker News thread on EmDash drew 364 upvotes and 261 comments -- substantial engagement for a CMS announcement.hn-thread The sentiment split was telling: roughly 25% positive, 50% skeptical, 25% mixed. High interest, low confidence. That's the community telling you: we see the ideas, we don't see the path.
The most insightful criticism came from a commenter called 9dev: "The hard part is displacing WordPress market share...building a community of bloggers, marketers, agencies...The valuable IP is the moat, but that's not an engineering problem."hn-thread Another commenter, ehnto, was more blunt: "None of it matters when the client has a suite of plugins they have been using for 10+ years, that are now core to their business."hn-thread
These are the same arguments I made yesterday. They are correct. And they are -- I now think -- slightly beside the point.
The critics are evaluating EmDash as a WordPress replacement. They're asking: can this CMS attract the agencies, the freelancers, the plugin developers, the theme designers, the hosting companies? Can it replicate the $596.7 billion ecosystem? The answer is obviously no. Not now, probably not ever. On those terms, EmDash is dead on arrival, and the graveyard of Ghost, Strapi, Sanity, and Contentful is waiting to receive the body.
But what if EmDash isn't trying to win the CMS war? What if the CMS is just the entry point for the platform play?
The vendor lock-in critics -- FlamingMoe calling it "open source but architecturally locked in," others noting the pattern of Cloudflare creating alternatives that only run properly on their infrastructurehn-thread -- are correct about the mechanism but maybe wrong about the framing. They see it as a flaw. Cloudflare sees it as the business model.
There's one criticism from the HN thread that I think deserves more attention than it got. A commenter named cutler pointed out: "Default EmDash size: 483MB. Default WordPress size: 70MB. That's progress?"hn-thread In a serverless context -- where the runtime environment is abstracted away -- install size matters less. But it matters in the vibes. It matters when a developer's first experience with your "WordPress successor" is watching nearly half a gigabyte of node_modules download. First impressions are vibes, and that vibe is: bloat.
The AI-generated code criticism is the most 2026 angle in the thread. One commenter objected to Cloudflare "using an LLM to build something and try to pass it off as something built from the ground up." Another noted that AI-generated code "bypasses licensing concerns" -- ironic for a project that chose the MIT license to signal openness.hn-thread The "vibe coded" accusation is a criticism category that didn't exist even a year ago, and it landed hard because the two-month development timeline makes it hard to deny.
But here's where I've shifted since yesterday: the fact that EmDash was substantially built by AI agents might be the point, not the problem. If EmDash is designed to be operated by AI agents, then being built by AI agents is dogfooding. It's a proof of concept. The codebase is both the product and the demonstration of the workflow the product enables.
The Security Model, Revisited: A Wall With Doors
I covered EmDash's plugin security model extensively yesterday, and I stand by most of that analysis. But the data deserves a fuller accounting.
The WordPress plugin security problem is not exaggerated. Patchstack's 2026 report -- the most comprehensive annual assessment of WordPress security -- documents the situation in unsparing detail: 91% of vulnerabilities in plugins (10,359 findings), 9% in themes (971), and just 6 in WordPress core.patchstack-2026 The top vulnerability type is Cross-Site Scripting at 3,800 findings (47.7%), followed by Broken Access Control at 1,130 (14.2%) and CSRF at 904 (11.4%).patchstack-2026
The exploitation timeline is the scariest number: 70% of high-impact vulnerabilities were exploited within 24 hours of disclosure. The median time to mass exploitation for heavily targeted flaws was 5 hours. And 46% of vulnerabilities lacked developer patches at the time of public disclosure.patchstack-2026 That means nearly half of disclosed WordPress plugin vulnerabilities are zero-days by default -- not because of sophisticated threat actors, but because the plugin developer is a solo developer who abandoned the project or a small shop that can't respond fast enough.
EmDash's response -- sandboxed V8 isolates with capability-based permissions, where each plugin declares what it needs in a manifest file and the runtime enforces those boundaries at the engine level -- is a genuine architectural improvement.dynamic-workers I said yesterday that it's "a wall, not a suggestion." I maintain that. I also maintain that the wall has doors, and those doors have to be open for plugins to work.
But there's a nuance I glossed over. There's a meaningful security difference between "this plugin has unrestricted access to everything" and "this plugin declared that it needs \database:write\ and \network:api.stripe.com\." The first model means a compromised slider plugin can exfiltrate your payment data. The second means a compromised slider plugin can only do what its manifest permits. Yes, users will approve whatever permissions an app requests. But the blast radius of a compromise is smaller. The sandbox doesn't eliminate trust. It bounds the consequences of misplaced trust.
That's a real improvement. It's not the revolution Cloudflare's marketing implies. But it's more than nothing, and I should have been clearer about that yesterday.
The Cloudflare-only limitation remains the central paradox. The full sandbox -- V8 isolates via Dynamic Workers -- requires Cloudflare infrastructure. Self-hosted EmDash on Node.js loses the isolation model entirely. The security feature that justifies the entire project's marketing narrative evaporates the moment you exercise the open-source freedom to run it on your own hardware.dynamic-workers That's not a bug Cloudflare plans to fix. It's the reason Cloudflare built the product.
| Security Dimension | WordPress | EmDash (Cloudflare) | EmDash (Self-Hosted) |
|---|---|---|---|
| Plugin isolation | None | V8 isolates (Dynamic Workers) | None (Node.js process) |
| Permission model | Trust everything | Capability-based manifests | Capability-based manifests (unenforced at runtime) |
| Theme database access | Full access | Blocked by architecture | Blocked by architecture |
| Vulnerability scan | Third-party plugins (Wordfence, Sucuri) | Built-in Workers AI scanning | Not available |
| Blast radius of compromise | Entire server | Declared permissions only | Entire Node.js process |
The sandbox doesn't eliminate trust. It bounds the consequences of misplaced trust. That's a real improvement -- just not the revolution that the marketing implies.
What EmDash Actually Gets Right
I've been critical of EmDash -- yesterday and today -- and I want to be fair about what it does well, because intellectual honesty demands it and because the things it gets right are instructive about where the web is headed.
The content model is forward-looking. Portable Text (structured JSON) is the correct content storage format for a world where content is consumed by browsers, mobile apps, AI agents, voice assistants, and whatever comes next. WordPress's HTML-blob model was right for 2003. EmDash's JSON model is right for 2026.
The Astro foundation is solid. Astro isn't a bet on an unproven framework. It's the #1 in developer satisfaction for three consecutive years on State of JS, the #4 most admired web framework on Stack Overflow, and used by Unilever, Visa, and NBC News.astro-growth Cloudflare didn't build on a toy. They built on the best content-oriented framework available.
The economic model is genuinely disruptive. Scale-to-zero pricing eliminates the floor cost of running a website. For the tens of millions of WordPress sites that exist in a low-traffic steady state -- and that's most of them -- paying nothing until someone visits is materially better than paying $5-50/month for an idle server.
The MIT license is strategically correct. WordPress's GPL license was a community-building masterstroke in 2003. In 2026, it's a governance weapon -- as the WP Engine dispute demonstrated, the GPL's copyleft provisions can be wielded to extract compliance or fees from commercial users. MIT removes that weapon entirely. Whether that's good for the community or good for Cloudflare is an open question, but it eliminates a class of governance risk that WordPress has proven unable to manage.
Passkey-first authentication is right. WordPress still defaults to username/password authentication in 2026. EmDash defaults to passkeys with OAuth and magic link fallbacks. This is a small detail that signals a large difference in security philosophy.
What EmDash Gets Wrong (Or at Least Incomplete)
The ecosystem is empty, and ecosystem building is not an engineering problem. Zero plugins. Zero themes. Zero agencies. Zero freelancers. Joost de Valk, defending EmDash, still had to acknowledge: "Ghost, Craft, Statamic; all technically excellent, none have built the ecosystem needed for broad adoption."joost The technology is necessary but not sufficient. It has never been sufficient. I've watched this exact failure mode play out for 20 years.
The migration story is weakest where it matters most. Content migration -- blog posts, pages, media -- is straightforward. EmDash has a WXR importer and a REST API exporter plugin. Fine. But the sites that would benefit most from migrating -- sites with complex plugin dependencies, WooCommerce stores, membership sites, LMS platforms -- are the ones where migration is effectively a full rebuild. The "let AI agents handle it" approach is clever but unproven, and the gap between "AI agents can convert a WordPress theme to Astro" and "AI agents can replicate the full functionality of a WooCommerce store with 47 plugins" is the distance between a demo and a deployment.
The April 1 launch was a genuine mistake. I said this yesterday and the data confirms it. Multiple Hacker News commenters assumed it was a joke. GitHub issue #19 on the EmDash repo is literally titled "April fools or real?"gh-aprilfools Cloudflare's CTO had to personally intervene to confirm the product was real.hn-thread When your CTO has to spend his time in a comments section reassuring people you're serious, the launch date has cost you something that engineering cannot recover.
The governance question is unresolved. Joost flagged this explicitly: "developers will want to see long-term governance commitments before investing their time."joost EmDash is a Cloudflare product. Cloudflare is a public company with quarterly earnings targets. What happens to EmDash if it doesn't generate sufficient infrastructure revenue to justify continued investment? Open source survives abandonment better than proprietary software, but "the code is available" and "the project is alive" are very different things. WordPress's governance problem is a BDFL who abuses his power. EmDash's governance risk is a corporation that loses interest.
The CMS That Might Not Need to Be a CMS
Here's where I want to revise my thinking from yesterday, and I want to be explicit about what changed.
Yesterday I argued that the WordPress succession wouldn't come from a better CMS. I said it would come from something that makes CMS irrelevant. I described a paradigm shift where "AI agents generate, deploy, and maintain web presences without a CMS layer at all -- where building a website becomes a conversation, not a software installation."
I still believe that. What I didn't see yesterday is that EmDash might be building toward that future, even if it's wearing the clothes of a traditional CMS right now.
Think about what EmDash actually is, stripped of the CMS marketing: it's a structured data layer with an agent-native API, sitting on a serverless compute platform with built-in storage and a globally distributed edge network. The "CMS" part -- the admin panel, the visual editor, the theme system -- is a user interface for humans. The MCP server, the Agent Skills, the structured JSON content -- that's an interface for machines.
If AI agents become the primary creators and managers of web content -- and the trajectory of the last two years suggests they will -- then the human-facing CMS interface becomes a monitoring dashboard, not a primary tool. The content gets created by agents. The themes get built by agents. The optimization gets handled by agents. The CMS becomes the API that agents talk to, and the human just sets the parameters and reviews the output.
EmDash, in that framing, isn't a WordPress competitor. It's the beginning of the post-CMS infrastructure layer -- a structured content platform that happens to have a human interface but was designed for machine operators.
I don't know if Cloudflare sees it that way. Their marketing positions EmDash as a WordPress replacement, which is the less interesting version of the story. But the architecture tells a different story than the marketing, and the architecture is what matters in the long run.
Where This Goes
Yesterday I ended with "I'm still waiting" for the paradigm shift that would make WordPress unnecessary. Today I'm revising that to something more specific.
EmDash will not displace WordPress. I'm as confident in that prediction today as I was yesterday. The $596.7 billion ecosystem, the 70,000+ plugins, the 25,000+ themes, the global army of freelancers and agencies -- none of that goes away because Cloudflare shipped a TypeScript CMS with better JSON serialization.wp-engine-study The CMS market share graveyard will add another headstone, and in five years EmDash will either be a niche developer tool or a subsidized customer acquisition funnel for Cloudflare's infrastructure products.
But EmDash might be the first product that points in the right direction, even if it won't arrive at the destination.
The structured JSON content model is correct. The agent-native API design is correct. The serverless, scale-to-zero economic model is correct. The vertically integrated platform approach is the only strategy with enough structural advantage to challenge an entrenched incumbent -- it's the Android strategy, and while Android didn't kill feature phones overnight, it did kill them.
The WordPress succession -- when it happens -- won't look like a migration from WordPress to EmDash. It will look like a gradual dissolution of the CMS category itself, as AI agents take over content creation and management and the "content management system" becomes an API layer that humans rarely touch directly. EmDash is pointed at that future more clearly than anything else I've seen, even if it's dressed up as a WordPress competitor to make the pitch deck work.
Here's my updated prediction, specific enough to hold me to: within two years, EmDash's most important feature won't be plugin sandboxing or Astro themes. It will be the MCP server -- the agent interface that lets AI systems treat web publishing as a programmable operation rather than a human workflow. The CMS will become the quiet infrastructure under an agent-managed web.
And when that happens, the "WordPress succession problem" won't be about which CMS wins. It will be about whether "CMS" is still a meaningful category at all.
I'm still waiting. But I'm watching more closely now.
References
- patchstack-2026Patchstack. "State of WordPress Security in 2026." 11,334 vulnerabilities in 2025; 42% increase over 2024; 1,966 high-severity; 113% YoY high-severity increase; 70% exploited within 24 hours; 5-hour median exploitation time; 46% unpatched at disclosure. Patchstack.
- hn-threadHacker News. "EmDash -- a spiritual successor to WordPress that solves plugin security." 364 upvotes, 261 comments. April 1, 2026. Hacker News.
- ghostW3Techs. Ghost CMS usage statistics. 0.1% of CMS-using websites after 13 years. W3Techs.
- cloudflare-astroCloudflare. "Cloudflare Acquires Astro to Accelerate the Future of High-Performance Web Development." January 16, 2026. Cloudflare Press Release.
- astro-growthAstro. "2025 Year in Review." Weekly downloads: 360K to 900K (2.5x growth). State of JS 2025: #1 interest, #1 satisfaction, #1 positivity. Stack Overflow 2025: 4th most admired (62.2%). Astro Blog.
- joostJoost de Valk. "EmDash: a CMS built for 2026." Called EmDash "agent-native, not agent-compatible." Spent evening building theme and migrating content. Noted empty ecosystem and governance uncertainty. Joost's Blog.
- dynamic-workersCloudflare. "Dynamic Workers." V8 isolate-based plugin sandboxing; not available for self-hosted deployments. Cloudflare Blog.
- wp-engine-studyWP Engine. "The Value of WordPress: World's First Study of the WordPress Economy." WordPress economy valued at $596.7 billion. WP Engine.
- techcrunch-wpengineTechCrunch. "The WordPress vs. WP Engine drama, explained." 159 Automattic employees left. $32M annual demand. TechCrunch.
- repository-governanceThe Repository. "WP Engine Accuses Automattic and Matt Mullenweg of Concealing Control of WordPress.org." WordPress.org revealed as Mullenweg's personal property. 20 core contributors signed open letter. The Repository.
- register-bansThe Register. "WordPress drama latest: Leader Matt Mullenweg exiles five." Deactivated WordPress.org accounts of governance critics. The Register.
- gh-aprilfoolsGitHub. "April fools or real?" Issue #19 on emdash-cms/emdash repository. GitHub.