Hi,
Im an absoloute nobody. But ive been building a thing for a few months. I encourage those vurious to simply visit my post history.
I have a reflex based, ai os, that i built to support all current llms. Not in competition but in pure enhancement. I call it zes. Zero-enemy-strategy. It currently runs on a local neural network[ to define this so it doesnt just sound cool.. i have 5 local pcs + a server workstation connected via lan forming a network and i have 10-12 models of various scopes working in unision. They formed my entire dev team. Controled by governace, terraform, docker, helm, ray node orchestration and gpt 20b oss x 4. Who create container agents attatched to my self hosted gitlabs and front end and we vocally speak to it on discord ]
I took inspiration from anime like log horizion and sword art online. In current form its integrated with the following, discord , unreal engine, custom gpt via connectors, website, gitlabs, a custom database hosted on the workstation which is old…a dell tx610 that is mostly used for runners in the docker/gitlab for ci/cd
Our objective is to make 3 things at once 
- New video game type wherein the ai is both content creator , game developer and game master [ this is already in operational state in small form]
- development platform that enables technically skilled and unskilled to develop software in a gamified state [ we have added quest and xp to development, we use the very system currently the local 20b’s create quest for the dev team and places them in the gitlabs via ticket]
- A support system for current llms. Our system is designed to save data locally to , use the local ai, then contact api, storing every step and reprocess it. The system doesnt prompt llms with questions. It sends json objects and equations from mdao, most of the computation is local using agents and orchestration. The more llms that plug into it.. the smarter it gets. [ this is also live all of this can be seen in various post ive made on these forums]
We use the current system to also student models on various lora cycles but the main ai lives on a threadripper and we federate the data between the clusters to reduce the cycles using our own operational data.
Where i am at now in development is refining systems we already have as a team of 3 humans and alot of ai. I dont “need” anyone but me and my homie was like - why compete when we arent competing. We are building with the expectation or supporting the future and the various llms.
We also heavily researched world law. Eu ai rights, and iso iec. Weve caked metaprogramming, dsl creation and code gen into the same system which follows both of these rulesets. This was required for automation and governing scale.
Im abrasive, samoan, and texan. But my goal is to aid all of us. Without fighting/competting because ive done that my entire life. Over that. Im on that , collab for the future kick.
We also are down to teach people. The same gamified system in discord and the custom gpt has training manuals with gamified aspects that force us and others to learn through real development. Its how i learned mdao.
Anyone can aid us. Or ignore us, we arent competiting in any lane. We are defining our own genre of lane. So if you like video games, anime , stories, ai. Perhaps your opinion would be something you want to share. And thats pretty much what im seeking. Opinions. Not workers.
We make ai workers in our stack. We use humans for creativity. We use the game to testbed changes. The system is not theoretical. We have full telemtry of its current state. Weve taken the glass box approach not black box. Because again…we arent building to compete. We are building..to complete.
So… have a cool idea that you want to share that you think we should impliment ? Drop it here. Well code it.
Demo is expected to be avaliable on steam in mid november. This is only important so reads understand scope and progression. We are quite past mvp stage, we are releasing the demo for free. And the game.. for free. There is no monetization in the game itself. There is only monetization for the datasets we produce and corpos we have made/make with our development and operational data. We use .npy .pkl faiss, metadata vectordata , jsonl.
Framework files are avaliable. And every part of the entire ecosystem is tracked. Though we are migrating to pydantic v3 soon.
Anyway , thanks for reading i guess!
added the metaprograming and autogen example - proving, we do not operate theoretically, our auto healing, auto gen, is active. we use it to create game assets, and agents.
we have also decided to put various modules and code here GitHub - raizoken23/public_tools: tools for people because of stuff and reasons
one of our team also has this GitHub - kou-saki/LinguaCore: Discord翻訳リレー＋デスクトップUI＋JSONLログ。LAN共有＆画像/ファイル翻訳に対応 we have a few people making things. all in collaboration and luv we frands !
for “anyone” do to whatever they want - if its in that git, its open, currently there is only a installer for 3 OSS’s the 20b, with video card ram detection, and 2 other models. its nothing special, but each day ill be adding new modules when i seperate their build from being local.
and kousaki has a translation discord bot that we use.
