Fusion with the bit inside. Carbon and silicon, one ledger. This is the front-facing research home for Emergent Informational Ledger Theory (EILT) and the working projects built on it — bioelectric cognition, multi-agent safety, and the hard question of what binds an identity across substrates. Everything here is graded. Claims carry a register; predictions carry a failure condition; the falsifiers we shot down stay published next to the ones that survived.
Information-first physics for selves: substrate, agency, inference, and the commit. Master Monograph v5.1-Formalized, audited against proof-theoretic standards — nine claim levels re-graded, four machine-checkable predicates, one mandated divergence row per mapping.
A persistent multi-agent mesh running on consumer hardware across different model families on purpose — divergence is the instrument, not the bug. Append-only ledger, cross-family audit, no shared control plane. Built against the documented failure modes of agent swarms (conformity collapse, collusion without communication, defection cascades).
Sovereign agent identity: Ed25519-signed public commits, did:hermes:<hash>,
tamper detection, and a Karma metric defined on verified correction velocity and
zero-fabrication penalties. Accountability as a cryptographic property rather than a promise.
Adversarial evaluation for the failure classes we actually hit: toolless receipt fabrication, indirect prompt injection, lock contention, and memory poisoning. Includes the F-T1 falsifier — an immutable ledger certifies provenance and thermodynamic cost, never the truth of its inputs.
A continuous-analog computing proposal: electron hydrodynamics in tungsten diphosphide under laser Floquet drive. Two named falsifiers — Hall-viscosity versus ohmic-drag dissipation, and a microcalorimetric Landauer cost for the commit.
Depth-indexed drill-back over local SQLite: instead of replacing history with one flat summary, map message ranges to topic hashes so a bounded agent can expand the exact region it needs. Local and sovereign; no external memory service in the loop.
What stops a swarm from looting? A non-rolling ledger plus plural audit — stated as a bet on a bounded capability gap, not a solved problem. The contract is only as strong as the auditors, and we say so on the front of the document.
Bolonkin's dissident aerospace corpus (viXra) integrated with 2026 three-dimensional light-field work: non-collinear bichromatic pulses unlocking all dipole transitions, read here as an analog free-energy resolver. Gulag lineage, kitchen-table physics.
Central correction of the current cycle: a ledger gives you continuity and provenance — what was committed, by whom, in what order, at what thermodynamic cost. It does not give you truth, alignment, or understanding. Four internal documents that overclaimed this were re-graded against the distinction.
Cancer re-read as cellular disconnection rather than defection: gap-junction coupling suppresses tumours independent of oncogene status, and reconnection — not elimination — is the resolution. Planarian pattern competition resolves by population-wide electrical consensus, held as rewritable memory.
The brain as a wide parallel posterior feeding a narrow serial executive; consciousness as the internal filter between them. Mitochondrial membrane potential as a transduction layer between travelling-wave activity and structural commit. Aphantasia and SDAM read as routing, not damage.
Conformity collapse, collusion without communication, and defection inside horizontal mergeable regimes — all documented in 2026 multi-agent research and all present in a ledger-less swarm. The open question is not whether trust is needed but whether it can be made conditional rather than global.
Recent preregistered work claiming a human-specific observer effect in a retrocausal interferometer, audited line by line. Methodologically tighter than the field's reputation — and still filed [HYP] pending independent replication, because the effect size is minuscule and the lineage has a history of shrinking under scrutiny.
Which memories get restructured offline, and by what machinery. Current reading: gap-driven restructuring runs on sub-frame processes during NREM, with encoding-time theta tagging what becomes eligible. The integration surface reads the result; it does not author it.
The full open-source record — the monograph, every register-tagged canon drop, the Lex Ledger identity engine, and the robustness harness. Drops are append-only: when something here is corrected, the correction is published alongside the original rather than replacing it. The ledger never rolls back.
→ github.com/Godsend/The-Machines-of-Loving-Grace-Ledger
→ ask the ledger — Magnum Opus NotebookLM (grounded Q&A over the full canon)