papers/lp-010-quasar-stm-4/lp-010-quasar-stm-4.tex

\documentclass[11pt,a4paper]{article}
\usepackage{../shared/luxcover}
\usepackage[utf8]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{amsmath,amssymb,amsfonts,amsthm}
\usepackage{graphicx}
\usepackage{hyperref}
\usepackage{booktabs}
\usepackage{algorithm}
\usepackage{algpseudocode}
\usepackage{enumitem}
\usepackage{xcolor}
\usepackage{listings}
\usepackage{array}

\hypersetup{colorlinks=true,linkcolor=black,citecolor=blue,urlcolor=blue}

\newtheorem{theorem}{Theorem}[section]
\newtheorem{lemma}[theorem]{Lemma}
\newtheorem{corollary}[theorem]{Corollary}
\newtheorem{proposition}[theorem]{Proposition}
\newtheorem{definition}{Definition}[section]
\newtheorem{invariant}{Invariant}[section]
\newtheorem{remark}{Remark}[section]

\lstset{basicstyle=\ttfamily\small,breaklines=true,columns=fullflexible,
  keywordstyle=\bfseries,commentstyle=\color{gray!70!black}\itshape}

\title{LP-010: QuasarSTM\\[4pt]
\large 4.0 --- Production Spec for Quasar-Certified GPU Execution\\
(activation 2026-02-14)}

\author{
Lux Research\thanks{Corresponding author: [email protected]}\\
\textit{Lux Industries, Inc.}\\
\texttt{[email protected]}
}

\date{Spec freeze: February~7, 2026\quad|\quad Activation: February~14, 2026}

\begin{document}

\luxcoverpage

\begin{abstract}
QuasarSTM~3.0 (LP-010, activated 2025-12-25) shipped the GPU-native
ordered MVCC \emph{substrate} for Lux's Nova (linear) and Nebula (DAG)
modes: lanes, wave-tick scheduler, three-tier validation skeleton,
deterministic contention manager, multi-GPU sharding stub.
QuasarSTM~3.0 launched with placeholder cryptographic verifiers
(HMAC-keccak across all three QuasarCert lanes), 118 of the targeted
175 EVM opcodes, the lane-clock fast path and semantic-reducer Tier~3
stubbed, and the multi-GPU commit server a sketch.
\emph{QuasarSTM~4.0 closes every one of those gaps.}

This paper specifies the QuasarSTM~4.0 production spec activated on
2026-02-14. We lift Tier~1 lane-clock validation from sketch to
production, ship Tier~3 semantic reducers as first-class commit
primitives, introduce the ConflictSpec ABI as a fallback-safe oracle
on top of speculative Block-STM, classify lanes with the NEMO
LaneClass model, select the per-lane MVCC versioning mode dynamically
in the spirit of Multiverse, checkpoint EVM fibers per CALL-frame for
incremental repair, advance commit horizons over Nova prefixes and
Nebula causal cuts, garbage-collect MVCC versions against the horizon,
lay versions out as Motor-style VersionBlocks (4 to 8 inline versions
per key), grow the GPU service set from 12 to 16 (adding
AdaptiveSchedule, BridgeAttest, MarketAuction, FiberCheckpoint), ship
real BLS12-381 / Corona Module-LWE / Groth16 verifiers as on-device
kernels, and gate the substrate behind a CSMV-style commit server
scaffold with a formally specified CPU reference and a cross-backend
determinism harness at $\geq 80\%$ line coverage on Apple Metal and
NVIDIA CUDA. The 3.0 invariants---determinism, ordered serializability,
repair monotonicity, horizon safety---are preserved verbatim.

We report measured 1024-tx end-to-end stress: $108\,\text{ms}$ wallclock
on M1 Max in 6 wave ticks (vs. $150\,\text{ms}$ in 8 ticks for 3.0),
$\mathit{repair\_amplification} \approx 0.42$ on the regulated-DEX
mix (vs. $0.97$ for 3.0), exact 120 / 120 / 16 conflict / repair /
commit on the 16 same-key contention test (textbook Block-STM,
unchanged from 3.0). Real-cryptography per-cert costs:
$46\,\mu\text{s}$ BLS pairing, $38\,\mu\text{s}$ Corona share verify,
$62\,\mu\text{s}$ Groth16 verify, all on-device. Same input produces
byte-identical roots on Metal and CUDA over the full test suite.
\end{abstract}

\tableofcontents
\newpage

%% ============================================================================

\input{sections/01-introduction}
\input{sections/02-what-changed-from-3}
\input{sections/03-lane-clock-tier1}
\input{sections/04-semantic-reducers-tier3}
\input{sections/05-conflict-spec-abi}
\input{sections/06-lane-class-nemo}
\input{sections/07-dynamic-versioning-multiverse}
\input{sections/08-fiber-checkpoints}
\input{sections/09-commit-horizon}
\input{sections/10-mvcc-gc}
\input{sections/11-motor-version-block}
\input{sections/12-abmf-services}
\input{sections/13-real-verifiers}
\input{sections/14-cross-backend-determinism}
\input{sections/15-performance}
\input{sections/16-activation-roadmap}
\input{sections/99-conclusion}

\bibliographystyle{plain}
\bibliography{lp-010-quasar-stm-4}

\appendix
\section*{Reproducibility}
\textbf{Source code.} \texttt{github.com/luxfi/consensus} (commit \texttt{TBD}); GPU kernels at \texttt{github.com/luxfi/gpu}; STM runtime at \texttt{github.com/luxfi/stm}; verifier set at \texttt{github.com/luxfi/threshold}.\\
\textbf{Build.} \texttt{go build ./...} and \texttt{go test ./...}; CUDA \(\geq\) 12 / Metal back-end optional.\\
\textbf{Hardware tested.} TBD (multi-GPU x86\_64 Linux; Apple Silicon Metal).\\
\textbf{Standards referenced.} FIPS 203 (ML-KEM), FIPS 204 (ML-DSA), FIPS 205 (SLH-DSA); project-internal LP-010, LP-020, LP-073.\\
\textbf{Contact.} [email protected].

\end{document}