Skip to main content
These are the tutorials worth reading front-to-back. Each one takes 20–40 minutes and ends with a working system you can keep. Pick the one closest to what you’re building — the patterns translate.

Agent swarms

Fork one golden VM into N parallel agents. Build a full-stack app (server + client + landing) in ~60 seconds by running three agents simultaneously.

Parallel scenario testing

Drive a real webapp to a stateful point, branch the live VM, then run every checkout scenario in parallel against isolated copies of the state.

Database state testing

Snapshot a seeded database, branch per migration scenario, run them in parallel, compare outcomes. Reproducible schema changes at the speed of git checkout.

Building pi extensions

Extend the pi coding agent with custom tools inside Vers VMs — a pattern that generalizes to any agent framework.

​
The shape of a Vers workflow

Every tutorial here follows the same arc:
1

Provision once

Boot a root VM, install everything your workload needs, get to a “golden” state.
2

Commit

Snapshot the golden state so you can restore it later, anywhere, forever.
3

Branch per task

Every task, test, or agent trajectory runs in its own branch — parallel, isolated, microseconds to spawn.
4

Collect and discard

Pull results out, keep the branches that matter as commits, let the rest go.
Once you’ve read one tutorial, the next will feel familiar — the primitives are the same, only the workload changes.

​
Want to contribute a tutorial?

The docs source lives at github.com/hdresearch/vers-docs. Tutorials live in tutorials/, follow the existing structure, and welcome PRs.