Where Nexus Is Going Now

This is not the end. We have so far to go. The basic engine needs to be developed more for advanced features, extend context, increase capability, usage cases, our own AI model development. Check it out.

Meet the Pocket Agent

Imagine if you will for a moment that you face the unimaginable, a broken computer. You've googled it, you've checked out the you tube videos that have nothing to do with your specific problem, you have no idea, it's just broken.

Now imagine you have a USB drive with an AI model that specializes in fixing broken computers, your internet is broken and that's OK, because Nexus doesn't need internet.

But wait, it's more than just broken internet, it won't even boot, it just pops up an error and shuts off or restarts every single time you try! But wait... Nexus doesn't care about that either.

So you pop Nexus into your computer that won't even boot. When the computer starts this time, you're able to choose Nexus Live environment, It boots into a custom Linux with API functionality specifically with your AI in mind.

You tell it to diagnose the broken system, and it goes to work and starts scanning the file system for errors, automatically processing results and moving on after every step.

You walk away, have some dinner, watch some TV. Come back later to a message, I found the issue and fixed it. Please restart and test that it's working right.

Thanks to Nexus this won't be a dream much longer.

Less Exciting Stuff

It's hard to throw out anything that compares to the pocket agent except variations of that, your own actual work agent in your pocket with a work environment designed for AI collaboration. We are doing back end engine stuff like dissecting and reassembling KV to increase memory and allow for project organization like file versions as a project progresses. We want to try adjust attention to see how far we can push it while still retaining the pretty well perfect context we have now. There is also the advanced UI that can tie separate AI models together, you can literally have two computers with two separate AI models collaborating together. We also want to try and build a shared KV for multiple models on the same system to allow them to communicate without having to process text into tokens every time. Plus we want to develop our own AI models trained to make the most of this new engine.

I am personally excited about the opportunities Nexus exposes in all sorts of directions, not needing servers really helps when you can still have usable AI.