Sprints Are the Wrong Unit for AI Delivery
MCP and LLM tooling make delivery, reporting, and testing more continuous. That changes what teams should optimize for.
ai / llms / software deliveryRecent writing
Practical articles about software and technology.
Software architecture, data systems, AI, reliability, and the decisions that come up in real engineering work.
2026
Indexes are not free
Adding an index speeds up reads. It also slows down writes, increases storage, and adds a maintenance burden that compounds over time.
postgres / database / architecture / software engineeringThe Cost of Making Everything Real Time
Real time sounds like an obvious improvement until you price it. The useful question is not how fast the data can arrive, it is how often anyone actually acts on it.
data / architecture / product engineering / software engineeringWhy RAG Fails in Production
Most RAG systems fail at retrieval, not generation. The fix is usually better evidence, ranking, and product logic.
ai / llms / architecture / software engineeringWhat Zero-Downtime Deployment Actually Requires
Zero-downtime deployment depends on overlapping versions, safe traffic shifts, and database changes that work before and after release.
devops / architecture / database / software engineeringEvals Are Not Tests
Unit tests tell you when code breaks. Evals tell you if your LLM is still answering the right question. These are different problems, and treating them the same way is how AI products degrade silently.
ai / llms / software engineering / braintrustIdempotency Is Not a Nice-to-Have
Every retry, webhook, and payment system depends on idempotency. Most engineers treat it as an afterthought. Here is what happens when they do, and what it actually takes to build operations that are safe to repeat.
architecture / distributed systems / software engineeringEvent-Driven Architecture Without Kafka
Most teams reach for Kafka too early. Here is how to build a production event bus with Celery and Redis, the patterns that make it reliable, and the four conditions where Kafka actually earns its place.
architecture / kafka / software engineering / redisPer-PR preview environments on Docker Swarm
How to give every pull request its own live URL using Docker Swarm, Traefik, and a wildcard DNS record, without running a Kubernetes cluster.
docker / devops / github-actionsWhat a Data Pipeline Actually Is
ETL has three letters. Most people can explain two of them. The third is where teams make decisions that shape who owns data, when it is trustworthy, and how hard it is to change.
data / architecture / engineering / software engineeringThe Difference Between a Queue and a Stream
Queues move work forward. Streams preserve a history consumers can read from. That difference changes ownership, replay, and failure handling.
architecture / kafka / redis / software engineeringAI Tools Are Not Magic Bullets
LLMs are not magic bullets. They are powerful translation engines for messy data, especially when you use them to summarize, structure, and move information between systems.
ai / llms / data / software engineeringContext Engineering Is Not Prompt Tuning
Prompt engineering focuses on the question. Context engineering focuses on the world you hand the model before the question. That distinction changes how you design reliable LLM systems.
ai / llms / software engineering / platform engineeringAI Scalability: What It Actually Means for Software Architecture
AI can generate working software quickly, but AI scalability depends on architecture, constraints, and disciplined engineering after the proof of concept.
architecture / scalability / ai / software engineeringSafe Database Migrations at Scale
A migration that passes in staging can lock a busy production table for minutes. The difference is not the migration. It is the assumptions behind it.
postgres / database / architecture / software engineeringWhy Search Is Such a Hard Thing to Talk About
Search sounds like one feature, but it is really a bundle of product decisions, customer expectations, matching strategies, ranking tradeoffs, and business context.
search / postgres / product engineering / software engineeringObservability Is Not Logging
Adding more logs does not make a system observable. Observability is the ability to ask questions about a running system that nobody anticipated when the system was built.
observability / platform engineering / software engineering
Areas of expertise
Browse by topic
Start with a subject, then follow the related articles through the decisions behind it.
Architecture & Distributed Systems
Practical writing on software architecture, distributed systems, messaging, scalability, and the tradeoffs behind reliable system design.
View topicData Engineering
Guides to data pipelines, PostgreSQL, search, indexing, migrations, and the engineering decisions that make data systems trustworthy.
View topicAI Engineering
Engineering-focused writing on LLM systems, context, evaluation, AI delivery, and using models as reliable parts of larger products.
View topicPlatform Reliability
Practical platform engineering on observability, deployments, preview environments, delivery workflows, and operating software reliably.
View topic