My approach to open source involves sharing not only well-developed ideas, but also preliminary sketches, and even experimental configurations like the weirdly confrontational AI bots (ideally deployed as two voice agent configurations engaging in spirited debate).

My name is Daniel Rosehill. I am based in Jerusalem, Israel and have spent most of my career in tech communications. I maintain a strong interest in AI and have been sharing experiments, GUIs, and projects on GitHub since last year.

This GitHub profile serves as a record of my evolving thoughts and explorations. The structure and connections are reflected in indexed repositories that I update periodically.

My primary interest within AI lies in proactively generating personal context data to enhance RAG pipelines for hyper-personalized inference. Currently, I am conducting a private experiment involving feeding contextual information about my life into a vector store, managed by an individual agent. I intend to share results once a substantial repository of context data has been developed and ingested. A working prototype of this workflow is available here.

My core beliefs regarding RAG and context are:

• Even a modest amount of contextual data can yield impressive results.
• RAG offers substantial utility for personal users.
• RAG's potential extends far beyond indexing knowledge bases and enterprise data stores.
• It is possible to consolidate one's life story into a single text document.
• With LLMs now supporting input token windows exceeding one million, numerous intriguing possibilities arise.

Additionally, I am interested in agentic AI, multi-agent networks, MCP, and workflow orchestration, as these complement context-aware applications. Over the past 18 months, I have developed numerous configurations for a network of assistants and agents, covering both personal and professional use cases. My work has progressed from system prompts to assistants with RAG pipelines, and now focuses on workflows and configurations with agentic capabilities.

As a long-term advocate of open-source principles, I firmly believe in AI's transformative potential for positive change. While I find debates surrounding AI ethics and security, and innovations in reasoning models interesting, I am more focused on exploring how we can leverage existing technologies to make meaningful differences in users' lives.

I share the common aversion to meaningless AI hype, but I am excited by the fact that the full capabilities of AI remain undefined. To varying degrees, we are all engaged in a process of discovery. This is a fascinating experiment, with the potential for significant advancements.

Note: I am receiving an increasing amount of unsolicited spam, which is becoming difficult to distinguish from genuine human outreach. If you share interests in AI and tech, please feel free to contact me via the email address listed in the sidebar.

Note: Repositories were reorganized in November 2024. If you previously starred or followed a repository and would like to stay updated, please re-follow. I apologize for any inconvenience.