AIG is developing a governed RF analysis and edge experimentation substrate for lawful signal workflows, bounded modulation/FEC triage, GRAND-inspired lab experiments, provenance, and human-gated decoder orchestration.
This research lane is adjacent to public interest in adaptive communications programs such as DARPA's Lightweight Universal Codec (LUC), but it does not claim LUC implementation, universal codec capability, protected payload recovery, or program association.
Read Capability Note Research CorpusSIGIntAgentOS began as a fieldable signals-intelligence cyberdeck. The RF research lane extends that origin story into governed edge experimentation: not just collecting or classifying signals, but preserving provenance, exposing uncertainty, bounding decoder paths, and keeping active authority human-gated.
DICE governs multi-agent coordination. Governed RF experimentation applies the same control-plane doctrine to signal interpretation workflows at the edge.
Offline dry-run plans rank lawful candidate decoder paths, label uncertainty, preserve provenance, and block protected/encrypted payload hints.
Synthetic lab fixtures, SigMF-style metadata, source SHA256 chains, and pure-Python IQ feature extraction support auditable RF workflows.
Bounded toy experiments identify structural FEC/code-family hints without emitting recovered payload bits or claiming universal decoding.
GRAND-inspired Hamming-weight and reliability-ordered noise-guess experiments run only on tiny synthetic binary symmetric channels and toy codes.
The SIGIntAgentOS RF Workbench gives operators a local view of features, broker rankings, classifier outputs, FEC/GRAND summaries, and audit logs.
Controlled decoder wrappers remain dry-run by default and require explicit authorization gates before external tool execution can be considered.
Source code, repository access, non-public artifacts, operator prompts, and evaluation packets are not published from this page.