Glossary

The terms this site uses in its own sense. Each definition points at the page where the term does real work, so a reader can check the usage rather than trust the summary.

Sovereign Intelligence, built on AienOS, is in alpha testing. Each component keeps its own evidence level. Definitions change when the evidence changes.

The system

AienOS

The experimental, open source OS layer this site documents. AienOS keeps its own identity within Sovereign Intelligence, the front door and alpha download. See The Machine.

AIEN

The persistent intelligence that learns. Weights suggest, programs explain, verification decides, evidence teaches. See The Machine and Research.

Atlas

The awakening layer, planned. The seed that wakes AienOS on new hardware. Doctrine is published; implementation is ahead.

Omega

The language for all computing, GPU to CPU. It turns intent into explicit programs that can be verified before they are trusted. See What Omega is.

CUDA++

The GPU half of Omega, licensed Apache-2.0 with the LLVM exception, for the NVIDIA GPUs people already own. See CUDA++.

Turing

The unit of net held-out explanatory compression. The site uses it to measure what a change actually explained. See the Turing field guide and the preprint.

Held-out

Data not used while fitting, so a result on it measures transfer rather than recall.

The loop

Propose, verify, measure, learn. The closed epistemic loop described on the front page.

Around the loop

Cortex

The canonical memory. It keeps what the machine has established.

Aegis

The verifier. It checks generated code and changes nothing.

Forge

Realizes what Omega verified.

Argus

Watches capability enforcement, so authority stays where it was granted.

Evidence and coordination

Receipt

A public link to the evidence behind a claim: a pull request, a commit, a measured data file, or a record. See Evidence.

Crumb

A durable coordination note under RFC-0001. Sniff before edits, respect unexpired locks, claim multi-step work, close with the result. No receipt, no crumb.

Research status

The qualification ladder and every experiment with its verdict, receipt and limits. See Research status.

Post-LLM

The position that a learning machine can keep knowledge as inspectable programs, with verification deciding what is trusted. See The Post-LLM Case.

The machines

DGX Spark

NVIDIA Grace Blackwell GB10 with 128 GB of unified memory. The development machine and the judge.

iMac 2017

The x86_64 port of the native C kernel, running on bare metal.

Raspberry Pi 5

The small machine.

QEMU

The laboratory where every kernel boots first.

The rig

The owned inference box.

The same list is available to machines in llms.txt and sitemap.md.