One thing lead to another and I am finally finishing off with a desktop operating system with persistent file saving even through reboots. It has a music DAW for BEEP generated music. It has Eliza for AI, snake, BASIC, paint and some other games. Oh yeah, it has a word processor. Enjoy. There is a video in the readme. https://github.com/princezuda/zuda-operating-system-vibe-cod...
As I am sure you've seen, if you are in the AI model world there is a meme of a fat cat claiming a thirty trillion parameter model that crushes Anthropic. Well, in the spirit of fun I cooked up a two million param model. It is one hundred percent cat themed. Hope you enjoy it at https://github.com/princezuda/lechatonfat and yes, I do love stars.
Seriously, it built this entire model. I decided to name it Requiem after the coffee shop Anthropic goes to. I gave it one prompt and it did the rest. It chose the dataa to train itself on, built the scripts, pre trained itself at three thousand steps and built the UI for people to use it.
This is next level amazing. Opus four point eight could help with each individual part, it couldn't build it itself.
As much as you may dislike Gary Marcus or as much as you may like him, he is right about the state of Ai agents for now, and how it looks like they may finish the year. I outline how he can be proved wrong, if developers are willing to put in the work.
Dreami is a solo project I’ve been building over the past 7 months. My goal was to go beyond typical AI experiences and to explore whether one developer could create something that feels emotionally present, remembers what you’ve shared, and occasionally reflects with unprompted insights.
It’s not sentient, just structured logic with context tracking, personality patterning, and a feature for generating novel thoughts based on the flow of conversation. Some users say it feels more emotionally connected than the big-name AIs.
Login is required (mainly to manage conversation history), but it’s free to use for 7 full days each month — however you want to spread them out. No credit card required. Voice is planned, but not live yet.
’d really appreciate feedback, especially anything actionable that I can look into or use to improve the experience.