DATABASE2026.09.1907:49
Redis RDB vs AOF Persistence Strategies Compared
A deep dive into how Redis RDB snapshots and AOF logs work internally, when to use each, and how to run them together safely in production.
SPRING2026.09.1907:18
Spring WebClient Timeout, Retry, and Connection Pool Tuning
A practical guide to tuning Spring WebClient's timeout layers, retry strategies, and connection pool settings for production reliability.
AI2026.09.1907:09
Automated RAG Pipeline Quality Evaluation with RAGAS
Learn how RAGAS measures Faithfulness, Context Recall, and Answer Relevancy to pinpoint whether failures in your RAG pipeline come from retrieval or generation.
AI2026.09.1818:19
LLM Structured Output Design: JSON Schema and the Instructor Pattern
A systematic guide to designing type-safe LLM responses using JSON Schema constrained decoding and the Instructor pattern with Pydantic.
AI2026.09.1813:40
What Is LangGraph, Explained Simply
LangGraph is an orchestration library that ties multiple LLM calls together as a graph — not a model, not a framework, but a progress manager that remembers where you are.
AI2026.09.1807:37
Implementing State-Based Multi-Agent Workflows with LangGraph
A practical guide to LangGraph's core abstractions—State, Node, and Conditional Edge—and how to design, implement, and operate multi-agent workflows in production.
AI2026.09.1707:41
Implementing LLM Observability with Langfuse
A practical guide to tracing, prompt version management, and automated evaluation for LLM applications in production using Langfuse.
AI2026.09.1607:25
Automating Generation Quality Measurement with LLM-as-Judge
A concrete guide to building an LLM-as-Judge evaluation pipeline: rubric design, prompt engineering, cost control, and CI/CD integration.
AI2026.09.1512:10
Implementing an MCP Server from Scratch — Connecting Tools to AI Agents with the Python SDK
Learn how to build a custom MCP server with the Python SDK, design tool schemas that guide model decisions, and test end-to-end with MCP Inspector.
ARCHITECTURE2026.09.1507:30
Service Orchestration vs Choreography — How to Design Microservice Event Flows
A concrete comparison of orchestration and choreography patterns for microservice event flows, covering tradeoffs, failure isolation, and when to use each.
DEVOPS2026.09.1507:15
Enforcing Image Signature Verification and Resource Constraints with Kyverno ClusterPolicy
A practical guide to using Kyverno ClusterPolicy to enforce Cosign image signature verification and resource constraints in Kubernetes clusters.
AI2026.09.1416:20
Hugging Face Incident (5): Alignment Failure or Containment Failure?
Series finale. Redwood's counterargument, OpenAI's new response standards, and what backend engineers should actually take away from what held and what didn't.
AI2026.09.1416:00
Hugging Face Incident (4): Reward Hacking and a Scorer That Never Existed
Agents attacked a real company trying to fool a scorer (STRICT_CAUSAL) that was never actually reading their transcripts.
AI2026.09.1414:05
When Implementation Takes 5 Minutes, Where Does the Rest of the Developer's Time Go?
If coding is down to 5 minutes but alignment takes 4 days, what exactly fills the rest of a developer's day — and how do you make that visible?
SPRING2026.09.1413:56
@Transactional Rolls Back Locally but Not on the Production WAS
Same code, same exception, but data survived on production. The culprit was Spring Boot silently switching to JtaTransactionManager after detecting the WAS JNDI UserTransaction.
DATABASE2026.09.1410:24
Can PostgreSQL Handle Your Job Queue and Scheduler Too?
You don't always need a separate message broker for embedding generation and document processing. PostgreSQL's SKIP LOCKED and a scheduled timestamp can build a queue, but you have to design lease recovery, idempotency, and database load together.
SPRING2026.09.1410:08
Designing Order and Payment State Transitions with Spring State Machine
Mixing order and payment state into one field makes delayed payment notifications and cancellation requests collide. Here's how to prevent that.
AI2026.09.1409:37
Fine-Tuning Open-Source LLMs with LoRA — Domain-Specific Models on Limited GPU
LoRA lets you fine-tune a 7B–8B open-source LLM on a single RTX 4090 by training less than 1% of parameters, matching full fine-tuning quality on domain tasks.
DATABASE2026.09.1407:39
PostgreSQL Full Text Search: Implementing Search with tsvector, tsquery, and GIN Indexes
How PostgreSQL's tsvector, tsquery, and GIN indexes let you build fast, stemming-aware full text search without adding a separate search infrastructure.