Beyond Facts: Formal AI Hypothesis About the Ontology Itself
How ontology and formal semantics let an AI go beyond inferring facts and formulate inspectable hypotheses about the types, relations, and rules of the theory itself.
Notes on data platforms, streaming systems, and the engineering trade-offs behind them.
Recent posts
From Full Loads to Incremental Processing: Computing Changes Is Not Enough
How monotonicity, coordination, and the information preserved by a pipeline explain the boundaries between full loads and incremental processing.
Journey to Real-Time Analytics - Apache Doris: One Engine to Rule Them All
Apache Doris consolidates responsibilities that a Trino architecture distributes across several components. This comparison examines the operational gains, the trade-offs, and what a load test revealed.
Journey to Real-Time Analytics - A Streaming-First Lakehouse
Among Iceberg, Delta Lake, Hudi, and Paimon, which Lakehouse format has the mental model most aligned with CDC and continuous streaming — and why that still doesn't solve real-time serving.
Journey to Real-Time Analytics - Lakehouse Is Not the Solution
Why support for updates and streaming ingestion does not turn Apache Iceberg into a real-time analytics layer.
SQL Execution Model
How a SQL query becomes an AST, a logical plan, an optimized plan, and execution.
Data Platform — Solution for Ingestion (Part 3)
An autonomous, near real-time ingestion layer built on Kafka, Avro, and object storage.
Data Platform — Solution for Consumption (Part 2)
A decoupled architecture for analytical data storage, cataloging, and processing.
Data Platform — The Challenges (Part 1)
The scalability and autonomy challenges involved in building data platforms.