Local memory layer for AI agents, survives restarts, no embeddings
Binary Lattice beats vectors: 19μs lookups, no embeddings, survives agent restarts.

2ms startup beats Lua, but 2x-5x slower at runtime—unfocused tradeoff.
Systems engineers, AI agent developers, embedded scripting use cases needing fast startup
Lua · Wren · QuickJS
I built it because I needed a specialized runtime for high-frequency, short-lived tasks (like AI agent loops) where Python/Lua startup latency and GC overhead become actual bottlenecks.
Core Architecture:
Arena-restart: No Tracing GC. It prioritizes allocation speed (~2x faster than Lua). Memory is wiped instantly between tasks via a restart() mechanism.
Fast Compilation: 100k classes compile in ~888ms (Single-pass).
Lightweight: The VM is designed for a ~2ms startup target.
The Trade-off: Execution is currently 2x-5x slower than Lua as I'm still optimizing the bytecode dispatch (if-chains).
I’d love to hear your thoughts on using Arena-restart for autonomous scripts and any advice on optimizing dispatch without bloating the startup time.
Binary Lattice beats vectors: 19μs lookups, no embeddings, survives agent restarts.
O(1) arena cleanup with handles making dangling pointers unconstructable by default.
Recovers 95% critical facts when switching GPT-4 ↔ Claude with real benchmarks.
Product Hunt meets roast culture with gladiator-style startup duels.
Local-first agent memory with SQLite graphs, but requires a license key.
Identity protection through compression cycles is clever, but agent memory is a crowded space.