Transformer inference benchmark across PyTorch/CUDA, custom CUDA kernels, and Modular MAX/Mojo; MAX reached 1.9x faster GEMM on the large model.
Native Rust memory consolidation daemon for background session compaction
Native honeycomb execution engine with 2D axial hexagonal geometry and model adapter pipeline
Native runtime state inspector, continuous watchdog, and hardware crash diagnostics
Process supervisor, health monitor, and crash recovery daemon for local services
Autonomous Change-Management, Schema Validation, and Eval Reliability Harness for AI Agents
Native reasoning extraction and dataset distillation pipeline
Native Rust GitHub Action and CLI for pull request diff auditing, credential scanning, and automated triage
Event ledger and session tree implementation of the Crumb protocol for spatial grounding
Ultra-high-performance native Rust and Mojo rewrite of MobHub faction operations engine for Torn City
Hardware abstraction adapters for local AI accelerators and Modular MAX
The Crumb Protocol: Spatial Grounding, Architectural Topography, and Ephemeral Scent Specification for Autonomous AI Agents
Native Rust memory engine with SQLite FTS5 lexical search and bi-encoder vector similarity
Decentralized peer-to-peer git synchronization utility anchoring repositories to the Radicle network
To associate your repository with the modular-max topic, visit your repo's landing page and select "manage topics."