Reason Machines has open-sourced Aspen, an Apache 2.0 Git server written in Rust designed to handle massive concurrent push traffic from AI agents by scaling writes per branch rather than per repository. Aspen independently implements the architecture OpenAI described for its internal Pando system. It uses PostgreSQL or SQLite to store branch history as version chains updated via compare-and-swap, allowing pushes to different branches to proceed without mutual blocking, while S3-compatible object storage holds immutable pack files. Aspen validates and stores pushes as-is without unpacking, serving fetches by streaming stored packs—recompressing nothing—and supporting CDN delivery via signed URLs. Trunk-only advertisement keeps negotiation small even with millions of branches, and background log-structured compaction merges packs 16 at a time. On a single 8-core server with S3 Standard and a 4-core PostgreSQL instance, Aspen sustained 567 pushes per second for five minutes against a git.git clone (4,500 files, 75,000 commits), with median push latency of 208 ms and p99 of 314 ms; a secondary client observed new pushes 30 ms post-acknowledgement at median. Depth-1 fetch storms consumed only 1.029x the bytes of stock git upload-pack. Cursor reports 120 pushes/second on S3 Standard and over 300 on S3 Express One Zone for its Origin service. Deployment options include a local `aspen dev` mode with SQLite, Docker Compose with PostgreSQL and MinIO, Helm charts, and Terraform for AWS, GCP, Azure, and Hetzner, plus a mirror mode for GitHub/GitLab. The project remains pre-alpha with no releases, web UI, or guaranteed 24-hour durability.