Æxtend

The Project uses a series of AI agents to Mimic your skills, to make the model much better than it was at that skill, which also make it better than leading models at that skill while still being a mid-tier model. We first feed an AI model some documents you might have, it makes conclusions, which show us its thinking process and what its learning. From there you can edit that according to what you need it to be then test it with that takeaway/thought-process, then tweak it, and repeat the cycle till what you have is the perfect one. From there use it on real life projects.

Usually Gemini 1.5 < GPT-4o For us, Gemini 1.5+Æxtend > GPT-4o

There is also Yddgrasil mode, which is several Gemini 1.5+Æxtend skills put togther, this is >4o, weaker than OpenAI O1, but can reason with many more diverse skills and over much longer sizes (1M tokens), giving it some advantages over even the strongest models

This project was made with my favorite fullstack setup, Sveltekit+Firebase, Then I used the Google GenAI api and its Model Gemini 1.5 for the very long very fast AI capabilities. From there I also used the Clip model to measure text similarity, and I used @Coinbase's wallets for all the transactions between my agents. Overall everything connects in the front end, but I had to make a key operation into a backend part and now its really clean+open source jump point for AI agents