This is the public update surface for research releases, architecture milestones, demonstrations, company announcements, and selected field notes.
Revalidation before execution, revocation, evidence-change handling, and causal decision chains are now first-class runtime behaviors — authority granted at decision time cannot silently persist through changes in evidence, policy, or operational state.
Read the RegOS overview →Each note links to the exact public source revision and the corresponding Hub surface.
The public CLI now carries the 0.2.1 catalog verifier together with explicit read-only public leaderboard reads and opt-in kernel-telemetry sharing. Local inference remains local; private Ask IDA receipts and authority surfaces stay outside the public CLI.
The latest public source update documents the repository boundary while preserving the rule that discovery remains evidence for review, not automatic trust or authorization.
The latest committed public source keeps local model execution, checkpoint handling, integrity, and request admission distinct from the authenticated Hub coordination layer.
The latest public source keeps the simulation and evidence path bounded while leaving hardware, trained-model, and universal household capability claims outside the evidence.
The current public HAVEN source continues the local-first household path with authority-gated actions, provider seams, speech contracts, local model management, and durable household-scoped history.
The public Cartographer source documents a standalone spatial-authoring foundation with generic interchange and bounded proposal workflows while keeping private adapters, anatomy material, and hosted integration surfaces outside the public build.
BUNDLE's public core keeps money, provenance, stale-data handling, deterministic projections, and optional connector admission local and explicit. It does not claim to be a bank connector, adviser, payment rail, or live-balance service.
The public Windows build records deterministic incident progression, data-only scenario packs, replay verification, and an explicit boundary against shell locking or runtime network behavior.
GLYPH publishes an offline reference foundation for visible-pixel data transport with manifests, hashing, codec/FEC seams, and explicit non-claims about optical hardware and conformance.
SIDECHANNEL keeps simulated signals, freshness, privacy, replay, and bounded evidence in a local-first spatial explorer without claiming raw payload retention or physical actuation.
CHAINSIGHT exposes a simulation-first industrial world for planning, governance, bounded execution, verification, replay, and reconciliation, without claiming safety-controller or physical-autonomy authority.
STRATA keeps evidence, uncertainty, branches, provenance, and deterministic scene compilation separate while reconstructing places across time offline.
The public results repository retains a committed numeric-boundary correctness fix. The Hub makes no broader product or live-data claim from its minimal public documentation.
The Hub now indexes thirteen currently public repositories, refreshes their committed source receipts, and leaves private or held-out work out of the public software map. Browse the public directory → Check GitHub visibility →
Bryan Rousey's The Whiteboard After the Film and Gerron Burford's Plymouth Rock Landed on US: The American Dream now have individual Hub overviews and direct SSRN records. Read Bryan's overview → Read Gerron's overview →
The paper examines record gaps, collective memory, credibility, and historical injustice within a conceptual framework. Read the publication overview → Open SSRN →
Gerron Burford's paper examines historical memory, repair, and settlement in relation to the American dream. Read the publication overview → Open SSRN →
The Hub now gives consumer tools, game-engine development, Cartographer, public source receipts, and the Open Canopy boundary their own public pages instead of asking one directory to carry every explanation. Open the library → Browse public software →
The public author surfaces now link Gerron Burford's verified research profile and current writing, while Bryan Rousey's four separately attributed SSRN records are indexed without folding them into the founder-authored corpus. Open the founder profile → Open Bryan's profile →
Cartographer now documents a standalone spatial-authoring foundation with generic interchange and bounded proposal workflows while keeping private adapters, anatomy material, and hosted integration surfaces outside the public build. Open the software directory → Source →
The latest committed Foundry source keeps local model execution, checkpoint handling, integrity, and request admission distinct from the authenticated Hub coordination layer. Read the public boundary → Source →
Gerron Burford's 51-page SSRN paper, written September 4 and posted September 11, 2026, develops a mid-range account of community as part of identity formation and explores expansive convergence, defensive identity, protective withdrawal, procedural convergence, and reciprocal civic belonging. Read the publication overview → Open SSRN →
The paper develops a relational theory of leadership centered on correctability, responsibility distance, and the possibility that consequence can return as meaningful information to authority. Read the publication overview → Open SSRN →
The curated founder-authored index now carries 14 direct Hugging Face article records and 6 SSRN papers. Bryan Rousey's four separately attributed SSRN records remain adjacent to this corpus. Source →
RegOS now treats revalidation before execution, revocation, evidence-change handling, and causal decision chains as first-class runtime behaviors. Source →
A controlled evaluation harness exercises governed systems over long horizons with incomplete information, where ground truth is held by the evaluator. Source →
Signed execution lineage and persistent device identity connect physical state, device continuity, and operator authority to the governed runtime. Source →
Validation work across 4-bit floating-point formats and selective-compute scheduling supports training and evaluating specialized model bodies on native GPU infrastructure. Source →
The company is advancing mission-assurance research that connects governed authority, long-horizon evaluation, and physical trust for mission-critical and regulated operational systems.
A Hugging Face article documenting the port of the IDA native training engine from Hopper-class infrastructure to consumer Blackwell hardware. Source →
A root-cause investigation showing how a fixed activation-row clipping ceiling became the primary cause of divergence in an FP8 mixture-of-experts training run. Source →
