What is Ato?
Ato is a way to treat software as a continuation: the computation that is still possible from a given point, rather than a service endpoint, a repository, or a snapshot.
The point is not to replace containers, package managers, or virtual machines. Those can prepare and run the environment. Ato's layer is what makes the computation running on top of them recordable, savable, branchable, transferable, and continuable.
The problem
Most software is delivered as access to a service: an interface that talks to someone else's server and database. What you receive is the screen, not the software. Three consequences follow.
- Work in progress cannot move. A session stops being usable when the provider's environment changes, the service shuts down, or you lose access.
- The past is not a place you can return to. A saved document is not the same as the program state that produced it.
- The environment decides what can run. The same project may work on one machine and fail on another because a runtime, a library, or a service is missing.
LLM-assisted development makes this more visible: software is produced faster and in more variations, so the ability to keep, compare, move, and resume a working computation matters more than the ability to keep the machine it first ran on.
The model
Computation ──▶ seal ──▶ Capsule ──▶ transfer ──▶ materialize ──▶ Run ──▶ continue
- A Computation is the work remaining from the current point. It changes when an interaction occurs.
- A Capsule is a sealed, immutable, addressable point in that computation. It can be preserved, passed, resumed, composed, and forked.
- A Run is a live evaluation resumed from a Capsule. Runs advance; sealing a point produces another Capsule.
- A Record is evidence that an interaction was observed. It explains how a Run reached a point but is not the point itself.
- A Materialization is a physical way to realize a Capsule: replay, workspace reconstruction, or a future snapshot mechanism.
These distinctions are the core of the model. Core concepts defines each term and explains how they relate.
What runs today
Ato is experimental. The repository includes computation objects, lifecycle commands, adapters, portable bundle handling and a static verification example. The portable application profile separates Contract K (observable conditions) from Derivation D (a declared realization path); a receipt records actual observations. The profile specifications are implementation-track drafts. Host capability and authorization determine which route can run. General heterogeneous-host recovery, process-memory restoration and autonomous AI Formation are research or development work. About Ato · Current development guide.
What Ato is not
- Not a general-purpose environment provisioner that replaces Docker, Nix, or VMs.
- Not a reimplementation of package managers or toolchains.
- Not a replacement for Git or a container registry.
- Not a guarantee that every process can be replayed deterministically.