Complete procedures with a way back.

Build an educational causal language model from randomly initialized weights using a licensed, provenance-tracked corpus. Train a dedicated tokenizer, define a roughly 34-million-parameter decoder architecture, establish random and unigram baselines, checkpoint and resume one bounded epoch, evaluate held-out loss and qualitative limitations, run local inference, and publish only an honest research artifact.

Release OCI images from protected source through a disposable digest-pinned BuildKit builder, native SPDX and SLSA provenance, optional CycloneDX inventory, Cosign keyless or KMS signing, exact identity verification, digest-only admission, and evidence-backed recovery.