Prism Engine is independently developed by Julian Torres. The system is open source. Collaboration takes five shapes. The evidence class is recorded either way.

Prism Engine is built by Julian Torres. The architecture is a sustained argument that the inference stack's abstractions are inadequate, and that the right response is a different one. The argument is not a vision document; it is a compiler, a runtime, an artifact format, and a constitutional model for canonical state.

The collaboration surface is named. The contracts are open. The collaboration takes five shapes:

• Hardware validation. Bring a target. Run the qualification harness. Publish the record. The target's evidence class moves with the run.
• Datacenter deployment. Bring a workload and a machine. Compile it. Run it. Compare the plans. Publish the receipts.
• Engineering. Bring a system worth understanding. Build the bridge between Prism and the system. The bridge is open source.
• Research. Bring a hypothesis the engine can test. The test is a receipt. The result is a milestone, a Lab Note, or both.
• Edge and robotics. Bring a constraint that the cloud assumes away. The plan is the constraint. The artifact is the handoff.

The system is open source under the project's license. The contribution path is the repository. The collaboration surface is the issue tracker and the discussion channels. The proprietary models remain confidential where an engagement requires it. The evidence class is recorded either way.

This site is Prism Observatory v1: the evidence-bound public projection of Prism Engine. It mirrors Prism's subject, projection, and receipt concepts. It is not a transaction system. A browser selection is not a constitutional event. The vocabulary is preserved because it is meaningful, not because every action it describes happens in the browser.

The site is the first public client of the runtime used to observe the runtime. The author wrote the engine; the engine produced the projection. The projection is the site.