Gradient Bang is a hugely multiplayer, totally LLM-driven sport. Come play Gradient Bang with us. See if you’ll catch me at the leaderboard.
This complete factor began as a result of I sought after to discover a host of items I’m lately obsessive about, in an software of non-trivial dimension, that felt each new and outdated on the identical time.
So … a retro-style area buying and selling sport constructed totally round interacting with and managing a couple of LLMs. Factorio, however as an alternative of clicking, you communicate on your send AI and determine tips on how to earn cash, make pals, and make havoc in your enemies.
One of the crucial issues we’ve been fascinated with as we hack on Gradient Bang:
– Sub-agent orchestration
– Managing very, very, very lengthy LLM contexts, together with episodic reminiscence throughout person classes
– International occasions and big volumes of structured knowledge enter as a part of human/agent conversations
– Dynamic person interfaces, pushed/created at the fly via LLMs
– And, in fact, voice as number one enter
In the event you’ve been development coding harnesses, or writing Open Claw brokers, or doing just about anything else that pushes the limits of AI-native building in this day and age, you’re almost definitely fascinated with these items too!
The sport is totally open supply. So if you wish to see how we constructed it, you’ll clone the repo and get started asking Claude/Codex concerning the code. If you wish to upload a characteristic, publish a PR.
New these days, design your personal company send brokers, run them in a Vercel Sandbox, and produce them into the sport. Suppose you’ll make your pair buying and selling loops quicker? That is going to provide you with a fairly large benefit within the sport. Wish to run with limitless corp send compute the usage of open supply fashions? You’ll be able to do this, now!
See the Vercel Sandbox subagents starter repo right here: https://github.com/pipecat-ai/gradient-bang/tree/major/deployment/vercel
