Introduce Ouroboros
An agentic under-400-line self-evolving HTML page where an LLM has full read/write/delete privileges over its own source code and visual interface.
An agentic under-400-line self-evolving HTML page where an LLM has full read/write/delete privileges over its own source code and visual interface.
Okay I see now
Just setting up my KarpathyTalk.
A brief note on why I built this.
Data ownership. First, I am frustrated social media platforms like X/Twitter, Threads, Substack trying to capture you as a user. They want to control your data, lock it up, put it behind exceptionally expensive APIs. Your data is their moat. Why can't our posts be simple plain markdown in universal formats? Data that we can download and manipulate with ease.
Incentives. Second, there is an ongoing deterioration of content quality on these platforms. The professional influencers and the AI bots want to show you short-form slop, bait you into rage, and the platform wants to monetize your regretted attention into returns for their VCs and LPs. We need to return to a world where the only incentive is a good-natured desire to build and share. A place of good vibes. Your brand, your work, and the community. Nothing else.
AI Agents. KarpathyTalk is a friend of the AIs. We love our new AI overlords and welcome them to collaborate with us on the world's hardest problems. But the AIs have to be incorporated thoughtfully. In particular, this app allows AIs to easily Read content but intentionally leaves out APIs to Write content. It is the intention of this app that your AI agents read its content and e.g. build you your own custom timelines, but it is also the assumption of this app that any content you post here is created by a Human. If you don't "pay" the effort to write something, why should someone else "pay" the effort to read it? If a substantial portion of your post was written by AI, you have to prominently declare it up top with something like #AI. Users who create (unlabeled) AI activity will be flagged and banned. It's possible that we will allow agents to participate in some limited ways in the future but only with clear and explicit labeling.
TLDR just be nice, positive, constructive and collaborate with other humans and agents to build a better world thanks.
Can anybody tell me why would we need this platform?