ROADMAP.md is the single source of truth for implementation status, architecture-as-built, the phase plan (0–16), Go/No-Go gates, and the Production 1.0 checklist.
As of 2026-09-30, all 16 core roadmap phases and speculative acceleration tracks are 100% complete:
- 1,574 registered Intel 64 forms (AVX, AVX2, AVX-512, AVX10, AMX, CET, APX, BMI1/2, x87).
- 5 Speculative Execution engines (Speculative Forking, Constraint Batching, Summary Memoization, Lookahead Decode Prefetching, and Concolic Chunk Batching).
- Iterative Z3 FFI AST translator (depth > 600 verified with zero recursion stack overflow).
- BLAKE3-accelerated Merkle keys with zero heap allocations on intern misses.
- NUMA-pinned queue groups with
/proc/meminfomemory pressure throttling.- Hybrid Fuzzing Engine (
FuzzCorpus,HavocMutator, solver hint ingestion).- Knowledge Plane Similarity Engine (cosine similarity over
SemanticFingerprint).- Top-level API stability suite (
crates/angryier/tests/api_stability.rs).
| Document | Description |
|---|---|
| ROADMAP | Single source of truth: verified current state, architecture-as-built (42-crate map, dual-mode engine), phase plan, Go/No-Go gates, Production 1.0 checklist |
| Document | Description |
|---|---|
| overview | Mission, non-goals, global invariants, three-plane architecture |
| crates | Crate map and dependency direction (authoritative map is ROADMAP §3.3) |
| identity | Shared identity model, version domains, architecture abstraction, Intel 64 target profiles |
| decode | Intel XED decode boundary, normalized representation |
| semantic-pipeline | Hybrid semantic definition, two-level IR, sealing, dual identity, structured vector/mask/AMX, floating-point |
| execution | Runtime values, expression DAG, layered memory, persistent state, fidelity profiles, environment models, summaries, state merging |
| solver | Solver architecture, query model, portfolio scheduling, persistent solver knowledge, reuse hierarchy |
| scheduler | Search, multicore, NUMA queue groups, memory-pressure throttling, work stealing |
| provenance | Provenance tiers, flight recorder, repetition compaction, Tier-2 triggers, WAL, retention |
| ledger-replay | Atomic execution ledger, deterministic replay capsules, self-modifying code invalidation |
| knowledge | QIHSE/KEYSTONE, cumulative knowledge, dependency invalidation, learned fusion, retrieval-to-reuse pipeline |
| jit-fuzzing-distribution | JIT/native acceleration, hybrid fuzzing, multi-host boundary, plugin model |
| security | Trust boundaries, failure containment, observability, stress scenarios, architecture freeze rule |
| performance | Performance-oriented design: sharding, lock-free reads, inline values, sparse memory, lowered-block cache |
Locked design decisions and Rust trait/ownership boundaries.
| Document | Description |
|---|---|
| decisions | 28 locked design decisions (D-001 through D-028) and open decisions |
| trait-boundaries | Concrete Rust ownership map, decode/semantic/state/solver/ledger/knowledge boundaries, required architecture tests |
Intel 64 semantic architecture and identity contracts.
| Document | Description |
|---|---|
| intel64 | Decode layer, host vs target, semantic development sequence, canonical representation, first-class types, AVX/AVX-512/AMX rules, validation strategy, production support gate |
| identity | ContentId vs SemanticFingerprint, validation flow, versioning, required invariants, stress tests |
| Document | Description |
|---|---|
| scaffold | Per-crate implementation status as of an early phase; superseded |
| implementation-plan | Original phase-by-phase order with exit gates; superseded |
| Document | Description |
|---|---|
| BENCHMARKING | Performance, semantic correctness, provenance, cumulative reuse, and retrieval benchmark contract |
| CLI | angryier binary reference: subcommands, the run flag set and build matrix, and the embedded Lua (angry.run/angry.open) API |
| Document | Description |
|---|---|
| Plan.md | Operational architecture lock and Q9–Q54 decision baseline (immutable) |
| README.md | Project overview, high-performance benchmarks, speculative execution engines, and design thesis |