CKEditor

Ondřej Chrastina caught up with @Pavel Petrov, a full-stack developer at If Insurance Baltic, mid-hustle between sessions at DevWorld 2026! His goal was simple: absorb as much as humanly possible before the clock ran out. What brought him in? Networking, fresh inspiration, and meeting the teams behind the tools he uses daily, including the crews from GitLab and Auth0. He's currently hunting for the right authentication setup for his own projects, plus a lightweight framework for full-stack web dev. #WebDevelopment #DeveloperCommunity #Devworld

A big thank you to CKEditor for supporting TYPO3 Developer Days 2026 as our Room & Lanyard Sponsor! 🚀 From helping attendees find their way between sessions to making sure everyone has their badge close at hand, these small details make a big difference to the overall experience 🧡 We're grateful to have CKEditor as part of #T3DD26 and appreciate your commitment to the open source community. We look forward to another fantastic Developer Days together! 🚀

Your product's AI features are only as available as the model underneath them. If you're a product leader who ships AI-powered features, the Fable 5 ban is a useful stress test. The Commerce Department ordered Anthropic to take Fable 5 and Mythos 5 offline globally with no warning. Less than 12 hours after the ban, they were gone. AI writing assistants, content generation tools, document workflows: all of them, instantly unavailable to every user on earth. Many third-party embedded AI features and application components are built on a single model. When that model goes down, gets deprecated, gets expensive, or gets regulated out of availability, your users feel it. Content workflows break and editing assistance disappears, and the feature you shipped as a differentiator becomes a liability. The product architecture question is: does your AI stack let you swap models without rebuilding the integration and without your users noticing? Increasingly, model capabilities, pricing, and availability are moving targets. The products that handle that gracefully will earn more trust than the ones that don't. For document and content creation in particular, where AI has become embedded in how users actually work, continuity matters. A writer mid-draft shouldn't know or care which model is generating their suggestions. That's your job to handle. Build for model independence. Your users will never thank you for it, because they'll never know it was a problem.

The Fable 5 ban is the model dependency audit you didn't know you needed. Portkey's analysis of 2 trillion tokens shows multi-provider adoption surged from 23% to 40% in just 10 months. Before the Fable 5 shutdown, production teams were already starting to realize that model access can be interrupted by deprecations, pricing changes, capability regressions, and rate limit shifts. The June 12th ban brought that reality home. If your enterprise AI integration is hardcoded to a single model endpoint, this is the scenario that exposes it. That goes for purpose-built applications and for the third-party components you've embedded in your products: AI writing tools, document editors, and content workflows. The architecture answer requires deliberate design: 1. Abstraction over direct integration. Route through a provider-abstraction layer and fallback becomes a configuration change, not a redeployment. 2. Consistent output contracts. Fallback only works if the replacement model produces output your downstream systems can consume without modification. 3. Tested fallback paths. Fallback that has never been exercised is a hypothesis. Run model substitution scenarios before you need them. AI vendors are now subject to the same export control frameworks that govern semiconductor exports and military hardware. Model flexibility is resilience engineering.

Ondřej Chrastina sat down with Mathias Biilmann Christensen, CEO of Netlify, at #DevWorld Amsterdam. His take: the interesting part of AI isn't shipping the same web apps faster. It's building things that weren't possible before LLMs at all. That's what he wants Netlify's primitives to enable. Full episode below!

If your AI model vanishes without warning, what will you do? The US Commerce Department created a governance and continuity nightmare by pulling the plug on Fable 5 at 5:21 PM ET on a Friday. And many organizations weren't prepared. Companies must be ready for situations where one model goes down without notice. At 5:21 PM Eastern, the US Commerce Department issued an export control directive to Anthropic. Less than 7 hours later, Fable 5 and Mythos 5 were offline globally for all customers because selectively complying with a directive barring access by any foreign national was technically infeasible during normal service operations. No warning. No transition window. Just gone. And Fable 5 has returned almost as quickly. This whiplash has exposed a gap that most enterprise risk frameworks haven't caught up with yet: AI model access has geopolitical exposure that standard vendor SLAs don't cover. Executives at Siemens, Renault, Orange, and ChapsVision told Reuters they already use a mix of US, Chinese, and European models specifically to avoid dependence on any one provider. Orange put the stakes plainly: the Fable ban made it "patently clear, if it wasn't before, how important it is for Europe to have access to an AI service that it can control, that will never be switched off on a whim." (Reuters, June 22) For IT leaders, CIOs, and governance teams, add these to your next risk review: 1. Which AI-powered workflows have no alternative if a model is suddenly unavailable? 2. Are the tools your teams use built on a single model, or can they route to alternatives? 3. Are your vendors treating model flexibility as a feature or an afterthought? AI vendor risk now belongs in the same conversation as infrastructure risk. When critical AI capabilities are fully outsourced to third parties with no flexibility or control built in, policy disruptions become operational ones, and business continuity planning needs to account for both.

🎥 Missed our live session on multi-root AI in CKEditor? It's now available on demand. Ondřej Chrastina and Witek break down how CKEditor AI keeps context accurate across multi-root and multi-editor setups, from CMS workflows to page builders with dozens of editable fields. Watch now: https://lnkd.in/daizkyZM

Enterprise teams keep telling us the same thing: AI edits their documents, and they find out something changed only after the fact. Our Senior PMM, Wojtek Cichoń, breaks down why that keeps happening. Frontier models corrupt roughly 25% of content during long edits, and most tools have no checkpoint to catch it before the change lands. He lays out what a governance layer actually needs to do, and why it matters more than picking a better model. Full blog: https://lnkd.in/dh3EtVXT

If your agency has been putting off a content tooling upgrade because procurement felt like its own project, that barrier just got smaller. Through the Carahsoft partnership, CKEditor sits on the same contract vehicles agencies already use, SEWP V, ITES-SW2, NASPO ValuePoint, and OMNIA Partners. For teams working within those vehicles, that's one less procurement cycle standing between them and AI-assisted editing, collaboration, and compliance features like WCAG 2.2, Section 508, and SOC 2. Details on the partnership and contract vehicles in the comments.

Exciting news! Government teams can now leverage CKEditor 5 through Carahsoft contracts, enabling real-time collaboration, enhanced compliance and scalable content creation. This partnership empowers agencies to modernize operations and better serve their communities. Click here for details: https://lnkd.in/ezF5tMMg Kate Bazilevskaya, Wojtek Cichoń