WordPress has a problem it cannot fix from the inside. Not a performance problem. Not a features problem. A structural one. 96% of its security vulnerabilities come from plugins, and the reason is simple. Every plugin gets access to everything. The database, the filesystem, the entire execution context. That is how it was built in 2003 and that is how it still works today. Cloudflare looked at that and decided patching was the wrong answer. EmDash is their attempt to start over. Built in TypeScript, Its serverless & powered by Astro & MIT licensed. No PHP, legacy architecture or plugins that can silently access your entire database. I want to be straight about what this is right now. It is a v0.1.0 developer preview. You are not migrating your production site today. But the architecture decisions behind it are serious enough that if you build on WordPress, run a plugin business, or host WordPress sites for clients, you should understand what Cloudflare just shipped.

For the past year, every serious AI conversation has circled back to the same dependency: Nvidia. If you wanted frontier performance, you needed their chips, If you wanted scale, you needed more of them. Then GLM-5 dropped & suddenly, benchmark charts that usually move inch by inch started shifting. There’s also a growing buzz online claiming GLM-5 may have been trained independently of Nvidia hardware, some even speculate about alternative stacks like Huawei’s. Nothing official confirms that. But the fact that people are even asking that question tells you how disruptive this release feels. Because the real reason people are talking isn’t just the size. It’s what GLM-5 is capable of. It is designed for longer, more demanding tasks where the model has to think in steps, plan ahead, and stay consistent instead of just giving a clever one-shot answer. It can handle multi-step workflows. It doesn’t lose track halfway through long contexts. And on Vending Bench 2, it ran a simulated business for an entire year and ended with a $4,432 balance. I’ve seen plenty of open models get close to the big closed systems before. But rarely do they feel balanced across everything. GLM-5 is one of the first open models in a while that doesn’t feel “almost there.” It feels like it’s actually in the same arena. And that’s why it’s suddenly everywhere.

Meta has a new AI model and for the first time in years it is not called Llama. Muse Spark launched yesterday under Meta Superintelligence Labs, a new internal division Meta quietly formed by bringing together researchers from Google DeepMind and other frontier labs. It is natively multimodal, supports multi-agent reasoning, and is available right now at meta.ai. It is also not being released as open weights. That last part is worth sitting with for a second. Meta built one of the most trusted brands in open source AI through Llama. Developers built on it, researchers published with it. Muse Spark continues none of that. No weights, no HuggingFace release, private API preview only. What you get instead is a genuinely capable multimodal model with some benchmark numbers that are hard to ignore and a new reasoning mode called Contemplating that puts it in conversation with Gemini Deep Think and GPT Pro. Whether that trade is worth it depends entirely on what you were using Meta AI for in the first place.

The open source model space has genuinely caught up. There are models today that genuinely rival GPT-5 and Claude Opus level performance and you can download their weights for free. The problem is running them. A 70B model at full precision wants an A100. Most developers aren't working with that. They're on an M2 MacBook Pro, an RTX 4060, maybe a gaming PC with 16GB of VRAM. That's exactly the hardware gap these five models are trying to close. All open source and capable enough to handle real coding work, and runnable on mid-range consumer hardware

Mistral has been shipping specialized models for a while now. One for coding. One for reasoning. One for chat. Each one doing its thing separately and requiring a different deployment decision. Medium 3.5 ends that confusion. One 128B dense model, one set of weights, handling instruction following, reasoning, and coding together. Mistral didn't just release a new model, they retired three existing ones to make room for it. Devstral 2, Magistral and even Medium 3.1 is gone. Medium 3.5 is what replaced all of them. That's either a sign of real confidence or a very expensive consolidation bet. Looking at the benchmarks, it's starting to look like the former.

Most open source models that claim agentic capability are really just instruction-tuned models with tool calling bolted on. They can call a function. They cannot think across ten steps, remember what they decided three tool calls ago, and course correct when something breaks mid-task. This is where Trinity-Large-Thinking comes into picture. Arcee AI released it this week. 398 billion total parameters, but only 13 billion active during inference. That MoE architecture means it runs closer to a 13B model in practice while carrying the knowledge of something nearly 30 times larger. And unlike most models where reasoning stops between steps, Trinity keeps its thinking tokens alive across the entire agent loop. Every decision it makes is informed by everything it reasoned through before it.

Text rendering in open source AI image generation has been broken for a long time. Ask most models to put readable words on a poster, lay out a comic panel, or generate anything where the text actually has to make sense and only few models can do it accurately and from rest you get something that looks like it was written by someone who learned the alphabet from a fever dream. ERNIE-Image is Baidu's answer to that specific problem. It's an 8B open weight text-to-image model built on a Diffusion Transformer and it's genuinely good at dense text, structured layouts, posters, infographics and multi-panel compositions. It can run on a 24GB consumer GPU, it's on Hugging Face right now, and it comes in two versions, a full quality model and a turbo variant that gets there in 8 steps instead of 50.