Skip to content

Reading Tracks

Each track is a curated reading sequence for engineers entering a specific domain or problem space. Essays within a track build on one another — earlier essays establish the conceptual foundation for later ones. Start at the top of whichever track matches your current problem.

Modern Data Architecture

For engineers designing or inheriting analytical data systems: data lakes, warehouses, lakehouses, and the pipelines that feed them.

  1. Parquet vs CSV, ORC, and Avro — The file format layer; why columnar storage exists and when it is appropriate
  2. The Physics of Storage Systems — The physical constraints (latency, throughput, IO amplification) that all storage choices operate within
  3. Lakehouse vs Warehouse vs Database — The architectural models for analytical data at scale, and their trade-offs
  4. The End of the Data Warehouse? — Whether the warehouse abstraction is mutating or being displaced, and how open table formats reshape the landscape
  5. Why Most Data Pipelines Fail — The organizational and technical failure modes of data pipelines, independent of tooling
  6. Metadata as Infrastructure — Why metadata is a control plane, not documentation, and how enforcement differs from registration
  7. Why Most Data Lakes Become Data Swamps — How governance failures produce data swamps, and the structural practices that prevent them
  8. The Hidden Cost of Metadata Debt — The compounding economic cost of catalog drift, ownership vacuum, and lineage gaps

Distributed Systems & Scale

For engineers evaluating whether to distribute, and how to reason about the costs of distribution.

  1. Why Most Microservices Should Be Monoliths — The organizational and technical pre-conditions that microservices require
  2. Appropriate Use of Microservices — When service decomposition is architecturally correct, and how to know the difference
  3. Distributed Systems and the Myth of Infinite Scale — CAP theorem, coordination overhead, data gravity, and the limits of distribution
  4. Why Most Kubernetes Clusters Shouldn't Exist — The orchestration overhead that most organizations adopt before they need it
  5. The Myth of Serverless Simplicity — How serverless relocates rather than eliminates infrastructure complexity
  6. The Hidden Cost of Real-Time Systems — The cost of streaming at every layer: infrastructure, observability, economics

Embedded Systems Doctrine

For engineers selecting hardware and protocols for embedded, IoT, and low-power systems.

  1. ESP32 vs Raspberry Pi — Microcontroller determinism vs Linux convenience; when each class is appropriate
  2. MQTT vs HTTP in IoT Systems — Protocol selection for constrained environments; connection models, power, and security
  3. LoRaWAN vs Raw LoRa — The protocol stack trade-off for long-range, low-power radio communication

Observability & Operations

For engineers building or improving the operational visibility of production systems.

  1. Observability vs Monitoring — The conceptual distinction and why it matters for how you instrument systems
  2. The Economics of Observability — The cost structure of observability at scale: ingestion, storage, cardinality, and human interpretation
  3. Why Most ML Systems Fail in Production — The observability gap in ML systems and what production monitoring requires beyond accuracy metrics
  4. The Human Cost of Automation — How automation accumulates abstraction debt and reduces the operational knowledge that observability depends on

Infrastructure Economics

For engineers and engineering leaders making infrastructure investment decisions.

  1. Container Base Image Philosophy — The security and operational cost trade-offs in container image selection
  2. Infrastructure as Code vs GitOps — How infrastructure automation models differ and when each is appropriate
  3. The Economics of GPU Infrastructure — GPU scarcity, utilization economics, training vs inference cost structures
  4. The Myth of Serverless Simplicity — Per-invocation vs reserved capacity economics, and when serverless is genuinely cheaper
  5. Why Most Kubernetes Clusters Shouldn't Exist — The full operational cost of cluster ownership
  6. The Human Cost of Automation — Skill atrophy, complexity transfer, and the organizational cost of automation

Spatial Systems

For engineers building geospatial data pipelines, spatial analytics systems, or location-aware applications.

  1. Geospatial File Format Choices — COG, GeoParquet, Shapefile, and the format decisions for raster and vector spatial data
  2. The Physics of Storage Systems — The latency and throughput constraints that spatial queries operate within
  3. Polars vs Pandas for Geospatial Data — The dataframe computation layer for spatial analytics
  4. The Operational Geometry of Spatial Systems — Spatial indexing models (quadtree, R-tree, H3), partitioning strategies, and routing system architecture
  5. DuckDB vs PostgreSQL vs Spark — Query engine selection for spatial analytical workloads