The Transformation of Documents: Repositories Are the New Unit of Knowledge Work
Traditional documents like reports and spreadsheets are limited by human cognitive constraints and are no longer sufficient for AI-driven workflows. Source code repositories, with their support for version control, decomposition, and stable references, are becoming the central medium for collaboration between humans and AI agents. Documents still exist but are increasingly treated as output artifacts generated from repository-stored source, similar to compiled binaries in software development.
- ▪AI agents require stable references, change tracking, and decomposability, which traditional documents lack but repositories provide.
- ▪Platforms like GitHub offer the infrastructure for repositories to serve as shared workspaces between humans and AI agents.
- ▪In the AVA project, human-written policies are converted by AI agents into executable invariant packs, with all collaboration and artifacts managed within a repository.
- ▪Repositories function as the source of truth, while documents become rendered outputs rather than primary work units.
- ▪The repository-centric model is expected to generalize beyond software development to all complex knowledge work processes involving intent, elaboration, and validation.
Opening excerpt (first ~120 words) tap to expand
The Transformation of Documents: Repositories Are the New Unit of Knowledge Work by Peli de Hallex, Don Syme, Ben Zorn on May 15, 2026 | Tags: agents, AI, productivity software, software development, software engineering Summary: Traditional documents — reports, slide decks, spreadsheets — are artifacts shaped by human biology: linear, single-file, snapshot-oriented. AI agents need something different: stable references, change tracking, decomposability, etc. Source repositories provide all of these natively. This post argues that the repository is emerging as a primary interaction medium between humans and agents — a workspace where humans declare intent, agents elaborate it into artifacts, and both iterate toward quality.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at SIGPLAN Blog.
Discussion
0 commentsMore from SIGPLAN Blog
