Notes
Notes
Short, systems-oriented writing on durable memory, versioning, and reproducible search.
- Anomaly Detection Belongs in Your Database →
Why we built SIMD-accelerated isolation forests directly into Stratum's SQL engine — and why exporting to Python is the wrong default.
- Versioned Analytics for Regulated Industries →
How immutable snapshots, copy-on-write branching, and cross-system consistency solve audit compliance, reproducibility, and scenario analysis in regulated environments.
- Memory That Collaborates →
How Datahike's distributed index space lets independent processes share and join databases through storage alone.
- Stratum: SQL that branches →
How we built a SIMD-accelerated columnar SQL engine on the JVM with copy-on-write branching - faster than DuckDB on 35 of 46 queries via the Java Vector API.
- Why We Built Datahike →
A personal story about functional values, long-lived systems, and the memory layer AI needs.
- Yggdrasil - Branching Protocols →
A protocol stack that brings Git-like branching to any storage system.
- The Git Model for Databases →
Copy-on-write, structural sharing, and branching - applied to your data.
- Why Search Needs Versioning →
Immutable search indexes for reproducible retrieval and systems that can explain themselves.