Research index / Project guide
MiRust.com publishes structured research guidance. It does not implement the browser runtime, execute model inference, or transfer research proposals into support claims.
Project boundary: the project owner identifies GGUPF.MiRust.com as a separate implementation project. MiRust.com remains the public guide, documentation, evidence model, and research index.
Current research tracks
Small-model runtime architecture
Loader design, tensor metadata, memory planning, quantization, tokenizer boundaries, KV caches, CPU and GPU dispatch, fallback behavior, and reproducible measurement.
Teleodynamic and resource-bounded systems
Fast and slow control loops, internal resource budgets, no-op behavior, structural adaptation, emergent halt, and the distinction between a theoretical model and a verified implementation.
TinyLM-16M reference workload
A compact transformer profile used to explain operators, memory estimates, quantization, model containers, and test planning. Values remain proposals or externally reported estimates until independently verified.
Standard ML runtime case study
A compiler and virtual-machine case study covering parsing, Hindley–Milner inference, pattern compilation, closures, bytecode, arena-based object graphs, and an optional embedded diagnostic model.
Rust-oriented architecture
Traits as ports, focused adapters, enums for closed domains, typestate, newtypes, RAII, explicit errors, message passing, and bounded use of dynamic dispatch.
Enterprise delivery and governance
Dependency policy, supply-chain review, testing, observability, semantic versioning, MSRV policy, SBOMs, release signing, migration, and rollback.
How research becomes site content
-
Intake
Record the source, owner, date, scope, and intended target page.
-
Separate facts from proposals
Mark observed values, estimates, design hypotheses, and external reports explicitly.
-
Publish a useful slice
Convert accepted material into a guide, documentation topic, roadmap item, evidence requirement, or correction.
-
Verify before widening claims
Require compatible source revisions, environments, methods, and limitations before changing a status.
Research inputs in this release
The current starter content incorporates original summaries derived from reports on custom Standard ML runtimes, TinyLM-16M browser execution, teleodynamic AI, browser-local Rust runtimes, Rust design patterns, and enterprise governance. The reports are research inputs, not proof that the described systems exist or meet performance targets.