Discord has been catching heat over its new verification rules. You’ve probably seen the threads everywhere. At some point I stopped scrolling and asked myself, Why does hanging out with friends online suddenly feel like paperwork? So I started looking for better alternatives to discord that don’t care who you are. They respect your privacy and still deliver a Discord-level experience. Some are great for gamers who want low-latency voice and familiar server layouts. Others are better for smaller communities that want tighter control or even self-hosting instead of ranking them from best to worst, I’ve grouped them by what they’re actually good at.

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.

Andrej Karpathy doesn't make random career moves. He co-founded OpenAI in 2015, left to build Tesla's self-driving program, came back to OpenAI for a year, then left again in 2024 to start an AI education company. Every transition has been deliberate and every one of them has turned out to be worth paying attention to. On Tuesday he posted on X that he's joined Anthropic. "I think the next few years at the frontier of LLMs will be especially formative," he wrote. "I am very excited to join the team here and get back to R&D." The "get back to R&D" part is the signal. Karpathy has spent the last several years teaching, building, and explaining. Now he's going back to the frontier. And the specific place he's going says something about where the most important work in AI actually is right now.

When I started looking for alternatives, I wasn’t searching for a better chatbot. I was searching for one that can simply work on my machine while being useful for me. Most AI apps on Android are just front-ends. You type something. It leaves your phone. A server processes it. A reply comes back. That's not what I call Private AI. MNN Chat does something different. It is an Open Source Android App that runs LLMs directly on your device. You download a model inside the app, and your phone handles the rest. The prompts don’t leave or gets processed by any server. It’s just your device doing the work. Under the hood it uses an engine optimized for CPU inference, which matters more than people think. Phones don’t have desktop GPUs sitting around waiting for 70B models. Efficiency is the difference between “interesting demo” and “actually usable.”

Tired of feeling like a tenant on your own PC? These 7 open source tools add missing features of your OS, fix everyday annoyances and give you more control.

Nvidia just dropped a 120B model that only uses 12B parameters at a time. Take a second with that. You get the reasoning depth of a 120B model. You pay the compute cost of a 12B one. That gap is not a rounding error or a marketing trick. It is the whole point of what Nemotron 3 Super is built to do. This is not another chatbot release. Nvidia built this specifically for AI agents — systems that plan, call tools, check their own work, and run for hours without a human in the loop. The use case is different. The architecture is different. And if you are building anything with agents in 2026, the timing of this release is hard to ignore. It's already live. Weights are on HuggingFace. Let's get into what actually makes it interesting.

Most multimodal models are text models with image handling bolted on. A vision encoder reads the image, converts it into tokens the language model understands, and the two systems communicate through that translation layer. It works. It's also where things break down when text and image content need to stay tightly in sync. SenseNova-U1 takes a different approach. Released by SenseTime under Apache 2.0, it removes the visual encoder and VAE entirely. No translation layer or separate systems. Pixel and word information modeled together from the start. The technical report isn't out yet and the A3B variant is still pending. But the 8B weights are available now.