Projects where the AI's output is the work — research, policy, argument — rather than a tool that does something for a user. Part of the Israeli AI Ecosystem resource list.
Projects where the AI's output is the work — research, policy, argument — rather than a tool that does something for a user. Part of the Israeli AI Ecosystem resource list.
Everywhere else in this index, the deliverable is software: an agent that books a table, a server that queries a government API, a skill that files a form. Here the deliverable is a document, and the software exists only to hold the process to a standard. The repository is the thinking, in public, including the parts that are unfinished and the numbers that have not been checked yet.
What makes one of these worth listing rather than being a folder of AI-generated text is that the standard is written down and enforced: sourced claims, a named failure mode when a claim cannot be sourced, and a pipeline that refuses to advance a half-evidenced argument to the next stage.
An Israeli national political manifesto built the way one ought to be — first-person testimony, CBS data, comparative policy from other countries, costed proposals with named losers. Ships a Claude Code plugin that enforces the evidence standards at every stage
The municipal sibling: a Jerusalem mayoral platform. Adds a gate the national version does not need — every proposal traced to a named instrument and classified by whether the municipality can actually deliver it, or is merely campaigning on it
Both are experiments rather than campaigns, and neither is a real candidacy. See each repository's own disclaimer.