Latest field note
Sidekiq Queue Latency: What It Tells You and What It Does Not
How to separate queue delay, job runtime, retry pressure and downstream contention before changing Sidekiq concurrency or adding workers.
Read the articleInsightsField notes from production systems
Engineering Insights
Rails performance, reliability, modernization and practical AI systems — field notes from diagnosing, building and operating production software.
Filter by topic
14 articles shown
Earlier articles
What a One-Week Rails Assessment Should Actually Produce
The evidence, decisions and implementation boundaries a focused Rails assessment should leave behind — without pretending one week can explain an entire system.
Rails Upgrade Planning: Build a Risk Map Before You Touch Versions
A practical way to inventory compatibility, data, deployment and operational risk before turning a Rails upgrade into a sequence of reversible changes.
How to Diagnose a Slow Rails Request Without Guessing
A measurement-first workflow for separating application, database, cache and downstream latency before you change code or infrastructure.
Why Ruby Is the Better Language for LLM-Powered Development
Ruby uses 42–45% fewer tokens than TypeScript and 28–32% fewer than Python across every major LLM tokenizer. Here are the numbers, the code, and the argument.
Training Rails LLMs, part 2: LoRA training, GGUF quantization, and local deployment
The second half of our fine-tuning project — training three models on a single A100 GPU for $75, fighting disk quotas during export, quantizing 30 billion parameters into a 17 GB file, and deploying everything locally with Ollama.
Training Rails LLMs, part 1: from 45 repos to 111K training samples
How we built a 15-step dataset pipeline from our internal Rails projects to train three domain-specific language models that write code our way. The dataset engineering mattered more than the model.
Migrating Off Heroku: A Rails Developer's Guide to DigitalOcean, Hetzner, Hatchbox, and Kamal
Heroku got expensive and the free tier is gone. Here are the hosting alternatives we actually use for Rails apps — DigitalOcean, Hetzner, Hatchbox.io, and Kamal — with real config examples and honest trade-offs.
What's New in Ruby 4.0: ZJIT, Ruby Box, and What It Means for Your Apps
Ruby 4.0 shipped in December 2025 with two headline features — the ZJIT compiler and Ruby Box. Here's what actually changed, what's worth paying attention to, and whether you should upgrade.
How to Audit a Legacy Rails Codebase Before Taking It Over
A practical checklist for evaluating an inherited Rails application. What to look at first, which red flags actually matter, and how to estimate the real cost of taking over someone else's code.
How to Build and Publish a Ruby Gem from Scratch
A practical walkthrough of building a Ruby gem from an empty directory to a published package on RubyGems.org. No magic, no hand-waving — just the actual steps and the files you need.
Sidekiq Patterns We Actually Use in Production Rails Apps
Battle-tested Sidekiq patterns from years of running background jobs in production. Covers job design, error handling, idempotency, and the configuration choices that keep things from falling apart at 3am.
How to use langchainrb and llama3 to generate blog posts
A quick walkthrough on setting up LangchainRb with Llama 3 via Ollama to generate blog posts using retrieval-augmented generation in a Rails application.
Deploying a Shopify App on DigitalOcean: A Comprehensive Guide
Discover the straightforward process of deploying a Shopify app on DigitalOcean in this concise guide. Learn how to connect your app with a GitHub repository, configure crucial environment variables, and seamlessly integrate your app with Shopify. Ideal for developers looking to leverage DigitalOcean's robust cloud platform for efficient Shopify app deployment.