Your dedicated Akasha instance
Understand what belongs to a Mind and what belongs to the platform.
A Mind is an instance of Akasha with its own data, configuration, and endpoint. It is the place your applications connect to for memory and engine operations. Minds, the platform around it, manages account ownership, provisioning, lifecycle, usage, and billing.
The Free and Pro contract
Both Free and Pro are specified to launch a dedicated Firecracker VM–based Akasha instance. Free is capped; it is not a shared-memory product. Pro expands the capacity of a dedicated Mind. The core engine API surface should be available under the same instance model, subject to authority and actual implementation status.
A service configuration can limit running workloads and cost. It should not silently change the meaning of a plan or present a disabled service as a nonexistent product feature.
What still needs alignment
The initial audit found Render provisioning and resizing, alongside a separate Kubernetes management path. The latest local customer flow now delegates to the Minds ensure service and disables direct Render creation. The engineering coordinator reports the hosted application still uses the earlier path. The replacement has not yet been demonstrated through a deployed instance-to-VM trace and authenticated user journey.
The local upgrade path returns Pro capacity metadata without sending an explicit resize instruction. A response alone therefore does not prove a capacity change. Some engine controls also remain restricted to internal callers; customers need a scoped instance-owner path for those controls, while host administration remains separate. These gaps have been reported for repair.
Two kinds of isolation
Instance isolation separates one runtime and its data from another. Learning isolation separates incoming information from accepted knowledge within an instance. They solve different problems. A dedicated VM does not make an untrusted source true, and a quarantine workflow does not replace runtime isolation.
Learning isolation · Platform provisioning · Instance authentication