Skip to Content

Opening the prepared-machine alpha

The current TARX alpha is a managed-local Electron alpha on a prepared machine. It is not a public download, clean-machine install, standalone/offline package, or second-Mac proof.

Use this page when you are opening the prepared alpha for review or support.

What this alpha proves

The alpha proves the local Always On loop on the prepared machine:

prompt -> run -> SSE activity -> evidence -> review/files/terminal/preview -> history/reload -> visible cancel

Current runtime lane:

AreaCurrent truth
App surfaceTARX Always On
Runtimeprepared-machine managed-local Electron alpha
Model laneNVIDIA Nemotron GGUF through llama.cpp
Model idtarx-computer
Provider typenvidia_local

What to check first

  1. Open the prepared TARX Electron app.
  2. Confirm the default selected surface is TARX Always On.
  3. Submit a small prompt.
  4. Watch the run move through visible activity.
  5. Open evidence surfaces after completion.
  6. Reload and confirm the run remains visible in history.

What not to claim from this alpha

Do not describe this as:

  • public download ready
  • clean-machine install proven
  • standalone/offline proven
  • second-Mac proven
  • hosted Supercomputer live
  • mesh routing live
  • enterprise governance live
  • full Codex replacement
  • NIM or TensorRT

The safe description is:

TARX is proving the local Always On loop on a prepared machine. The staged architecture supports future mesh contribution, hosted Supercomputer capacity, and enterprise governance, but those are not live claims for this alpha.

Next

Continue to What Always On does.

Agents