My vision of open source involves open sourcing not only my good ideas, but also my bad ones, my idea sketches, and joke configurations I write at three in the morning for weirdly confrontational AI bots (best deployed as two voice agent configurations bickering down the ether at one another).

My name is Daniel. I live in Jerusalem, Israel, and I've spent most of my career to date working in tech communications.

I'm fascinated by AI, and since about last year, have been sharing some experiments, GUIs, and projects on Github.

This Github profile is, in a sense, an imprint of my thinking. The contours and points of connections are mapped out in indexed repositories which I try to update every few months.

My main interests within AI include the idea of proactively generating personal context data fed into RAG pipelines for hyper-personalized inference. To that end, I'm currently running a private experiment by which I feed contextual information about my life into a vector store, backed by an individual agent. I'll share results once I've developed and ingested a meaningfully large repository of context data. A working prototype of the workflow, and its repo, is here.

My RAG/context "beliefs": you really don't need much of it to do some very impressive things; it can offer vast utility for personal users; it is extraordinarily useful (and powerful) for so much more than just indexing knowledge bases and enterprise data stores. You can fit (quite literally) your life story into a single text document and with LLMs with one million plus input token windows widely available, that opens up a lot of intriguing possibilities.

Also of interest and in a very complementary way (because context is only useful when we can do things with it): agentic AI, multi-agent networks, MCP, and workflow orchestration. Over the course of the past 18 months, I've written a lot of configurations for a network of assistants and agents, covering both personal and professional use cases. My first "creations" were system prompts, then assistants with RAG pipelines, and now I'm working on workflows and configurations with agentic capabilities.

Besides being a long-term advocate of open-source, I'm a huge believer in the transformative potential of AI for good. While debates around AI ethics and security interest me and advances in reasoning models are exciting, I'm more interested by the open-ended question of how we can best leverage what we have already to use the technology to make meaningful differences in users' lives.

Meaningless AI hype irks me as much as it does every semi to highly cantankerous techie on the internet, but I love the fact that the textbook on what AI can do hasn't been written yet. To greater and lesser extents, we're all free-styling it. It's a crazy experiment, with potentially great results.

Note: I receive a daily and ever-worsening barrage of unsolicited spam which is sadly (okay, maybe I blame AI here) is getting harder and harder to distinguish from real genuine human-to-human outreach (which I also get some of!). If you're also into AI, tech, and want to chat, drop me a line at the email in the sidebar.

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