I've used Figma. I've used Adobe XD. And for most design work they do the job fine — if you're okay with paying for them and okay with your files living on someone else's server. I wasn't looking for a replacement. I just stumbled across OpenPencil while browsing GitHub one evening and the one thing that caught my attention wasn't the canvas or the components. It was the MCP server built directly into the tool. An AI agent that can read, create and modify your design files from the terminal. That's not a plugin. That's a different way of thinking about design tools entirely. I installed it, connected it to Claude Code, created a sample design and spent some time with it. Here's what I actually found.

Give an AI agent a hard problem and it usually figures out the easy wins fast. After that, more time does not help. It just sits there, trying the same things. ZhipuAI ran GLM-5.1 on a vector database optimization problem and let it go for 600 iterations. It did not run out of ideas. At iteration 50 it was sitting at roughly the same performance as the best single-session result any model had achieved. By iteration 600 it had reached 21,500 queries per second. The previous best was 3,547. That gap is not incremental improvement. It is a different category of result. GLM-5.1 is open source, MIT licensed, and the weights are on HuggingFace right now. It works with Claude Code, vLLM, and SGLang. If you are building anything that runs agents over long tasks, this one is worth understanding.

OpenAI just launched Daybreak, a new cybersecurity initiative built around one uncomfortable reality, AI is speeding up vulnerability discovery faster than most companies can patch the damage. Earlier this year, HackerOne temporarily paused parts of its bug bounty program because maintainers were getting flooded with AI-assisted vulnerability reports. Some were valid. Some were hallucinated. Either way, humans still had to read them all. And that’s the change happening underneath all the AI hype. Finding bugs is getting cheaper. Faster too. What used to take weeks of manual research can now happen in hours with the right models and enough compute. Security teams are starting to deal with something closer to triage overload than a tooling shortage. OpenAI seems to think the answer is more AI, but aimed at defenders instead of attackers. That’s where Daybreak comes in. The company says Daybreak combines its latest models, Codex Security, and a group of security partners like Cloudflare, CrowdStrike, Cisco, and Palo Alto Networks to help security teams identify vulnerabilities, validate fixes, generate patches, and monitor risky code before attackers get there first. What makes this launch interesting is that it arrives just weeks after Anthropic introduced Mythos, its own cybersecurity-focused AI system. Both companies are chasing the same problem. But they’re handling access very differently.

Microsoft's MAI Image 2 just ranked third globally on Arena.ai. Here is what it genuinely does well, where it falls short, and what this launch actually signals about Microsoft's direction

A few years ago, running a small site felt simple. You wrote something useful, Google sent people your way, and a handful of those people stuck around. That loop is broken now. AI search tools don’t send visitors. They take your words, compress them into an answer box, and move on. No click. No context. If you’re a small publisher, it feels less like competition and more like extraction. What surprised me is this: even as AI search started eating the web, our site didn’t collapse. Sessions dropped in some places, sure, but the people who did arrive stayed longer, clicked deeper, and actually cared. After digging into our 2026 data, I realized why & it has nothing to do with ranking #1 anymore.

Getting an AI agent to start a task is easy. Getting it to finish one properly is a different story. Most agents fall apart somewhere in the middle. A tool returns unexpected output, the model misreads it, and everything that follows builds on that mistake. By step thirty you are looking at something that has completely lost track of what it was supposed to do. The five AI models here were built with that specific problem in mind. They handle complex multi-step tasks, real browser control, deep research and coding workflows. All open source & self hostable.

The assumption has always been that serious AI runs on serious hardware. Your phone gets the watered-down version, good enough for a demo but not for real work. MiniCPM-V 4.6 is a direct challenge to that assumption. 1.3 billion parameters. Runs on iOS, Android, and HarmonyOS. Needs 4GB of GPU memory or 2GB on CPU via GGUF. And on the Artificial Analysis Intelligence Index it scores 13 against Qwen3.5-0.8B's score of 10 at 19x lower token cost, and against Qwen3.5-0.8B-Thinking's score of 11 at 43x lower token cost. These parts matter when it comes to a model which runs on your phone.