deterministic-computing

启元甲言 (Qiyuan Jiayan) - 中文自举编译器：自举链全量编译 Git，五代 SHA256 全等。中文编译器/汇编器/链接器，零外部依赖直出 x86-64 PE。注：非同名 NLP 工具包。

⚠️ MOVED: See constraint-theory-core, constraint-theory-python, constraint-theory-web, constraint-theory-research

Governance framework defining institutional authority boundaries and policy contracts for AI-assisted workflows.

Open-source, offline-first reference model for transparent and auditable driver worktime calculation based on real operational data.

Follow-up to int-llm: TinyLlama-1.1B weights rounded to mixed F11/F12 int16, matching the Q16.48 integer oracle on all 632 evaluated greedy decisions. CPU, MCU, and CUDA evidence.

Reversible coordinate permutations for Llama checkpoints with exact Q16.48 validation and direct SafeTensors support.

High-performance incremental computation engine for hierarchical data structures with real-time query capabilities

Visual anatomy of a Q16.48 integer GPT — four self-contained offline HTML pages: weight atlas, bit-exact inference trace, 32-D embedding space, 3-D weight explorer. Every displayed number is exact (BigInt in-browser, verified against the C oracle). Fourth in the int-llm series.

SPL-G1 is a hardware causal-audit Trusted Compute Unit (TCU) — a dedicated security primitive, not a general-purpose CPU/GPU. Built on a 2D in-memory-compute (PIM) array with true IEEE-754 FP16 and the splcc C-to-microcode compiler, it anchors every operation to an immutable causal chain via a 256-bit SBC fuse and audit constraints. Verified in RTL

Oracle-verified experiments in mathematics and physics ML — exact integer training, zero-calibration quantization, and an append-only ledger that books nulls and retractions beside the wins.

Formal verification of HCSP using Coq

Recorded evidence and manuscript for a single-fixture packed-AWQ route-block arithmetic mechanism study

Real-time AI based anomaly detection & L2 MACsec control loop on legacy infra without GPUs. Powered by CXL 3.1, DPDK SmartNICs, and Ultra Ethernet.

The official repository documenting Abdulrahman Al-Alawi's foundational framework, theorem, USDL law, and the HCSP core for deterministic computing.

VECTORIA is a deterministic, cross-platform computational kernel framework focused on transparent execution, strict validation, and architectural honesty.

Official TLA+ specifications and Why3 formal verification artifacts for the HCSP Sovereign Deterministic Core, proving 100% complete logic correctness (19/19 goals) against zero-day exploits and cloud-freezing deadlocks.

Deterministic AI safety framework for medical & autonomous systems. Physics-based cancer treatment support, fail-closed architecture. NRA-IDE,律環公理,内包性動力学