Research
This section publishes public-facing engineering notes from American Dev Corp. It is designed as a readable research journal rather than a marketing feed: concise where necessary, careful about boundaries, and focused on governed systems work that can be discussed safely in public.
Wizard is the operator-facing layer for American Dev Corp's governed AI work: a local interface meant to help a person inspect, steer, approve, and understand what the system is doing. iLiC sits underneath Wizard as the governed cognition core, exploring deterministic control, memory continuity, and advisory AI behavior under explicit state-machine constraints. These notes are meant for technically serious readers who want architectural clarity without access to private repositories, private memory contents, or sensitive implementation details.
iLiC Notes 001
A public architecture note on deterministic control, memory continuity, and AI as advisory infrastructure.
A first public framing of why governed cognition is a different design problem from generic assistant behavior, and why explicit runtime control matters.
iLiC Notes 002
A draft note comparing survivable memory structures with familiar assistant-style memory features.
A public distinction between memory continuity as system substrate and memory as a convenience layer in ordinary AI products.
iLiC Notes 003
A runtime-governance note on mutation trust, overwrite visibility, and explicit preservation semantics.
A focused paper on why continuity systems must distinguish creation from mutation and make overwrite behavior visible before commit.
iLiC Notes 004
A research note on distributional convergence, recursive reinforcement, and why governed cognition remains important.
A public argument that recursive probabilistic systems may over-optimize toward consensus replication unless bounded by clearer governance structures.
iLiC Notes 005
A public note introducing Wizard before iLiC and clarifying governed autonomy under user authority.
A framing note for Wizard as the visible operating layer where user consent, memory, telemetry, and governed action come together.
iLiC Notes 006
A public architecture note on evidence provenance, challengeable claims, and governed knowledge admission.
A continuation of the iLiC research arc, separating memory from knowledge and framing validation as an auditable governance layer.
iLiC Notes 007
A public engineering note on structured knowledge catalogs, role boundaries, and reducing repeated AI runtime cost.
A practical framing of why categories, sources, scope boundaries, confidence levels, and permission rules can make AI systems cheaper, cleaner, and easier to inspect.
iLiC Notes 008
A public engineering note on source material, indexed evidence, retrieval, validation, belief traces, and governed knowledge promotion.
A continuation of the iLiC research arc, framing Personal Internet source material, Bookworm indexing, belief evaluation, and Wizard review as one governed learning loop.
iLiC Notes 009
A public engineering note on separating fluent conversation from usable knowledge through indexed evidence, citations, and governed source use.
A continuation of the iLiC research arc, showing how Personal Internet Library retrieval can support answers while keeping memory and KB promotion under human-approved governance.
iLiC Notes 010
A public engineering note on iLiC as bounded cooperating subsystems and conversation coherence as a governed evaluation layer.
A continuation of the iLiC research arc, explaining Wizard, Bookworm, Candidate Shelf, Claim Ledger, memory, telemetry, and validators as cooperating subsystems, then showing why conversation quality needs repeatable evaluation.