A properly technical batch, frontend, backend, infrastructure, and ML fundamentals.
-
envscan post on dev.to: It describes a tool I wish existed years ago. The problem:
.env.examplegoes stale because maintaining it is manual work, same as any other documentation. Every time you addprocess.env.DATABASE_URLin code, you’ve implicitly declared a dependency, but nothing ensures.env.examplereflects it. envscan flips the approach: instead of maintaining a schema that the tool validates against, it scans your.jsand.tsfiles and extracts everyprocess.envreference, infers types from naming conventions (_URL→ url,_SECRET→ secret,PORT→ number), and generates.env.exampleautomatically. Run it in a pre-commit hook and the file stays in sync because it’s regenerated from the code itself. The CLI is almost ready; the CI integration (GitHub Action that comments on PRs when new env vars appear without matching entries) is the paid tier at $6/month. -
Flexbox Masonry Layout Explained with Math: Frontend Masters goes deeper here than any masonry tutorial I’ve seen. Instead of handwaving about how items get placed, it works through the actual calculations (column assignment, item sizing, gap distribution) with diagrams and formulas. If you’ve ever tried to implement masonry and wondered why your layout was off by a few pixels, this is the reference that explains why.
-
light-dark() supporting images: Bramus Van Damme covers a CSS spec addition that’s been needed for a while. The
light-dark()function already worked for colors; now it can switch images and other url() values based on the user’s color scheme preference. This means you can have different logos, illustrations, or background images for light and dark modes without JavaScript or media query duplication. Still in spec-development, but worth tracking. -
Understanding Quantization: ngrok’s post is the clearest introduction to model quantization I’ve found for a developer audience (as opposed to ML researchers). It covers what quantization actually does to model weights, the tradeoffs at different bit widths (FP16, INT8, INT4), how GPTQ and GGML differ, and when quantized models are “good enough” versus when you need full precision. If you’re running local models and wondering what Q4_K_M means, start here.
-
Fine-Tune Your OPcache Configuration: Tideways has written the PHP infrastructure post most developers need and skip. Benjamin Eberlei walks through the OPcache settings to get right in production:
opcache.max_accelerated_files(set it higher than your file count or scripts silently get evicted),validate_timestamps(disable in production but understand the deployment implications),opcache.revalidate_freq, and memory settings. The “surprises” in the title are real: misconfigured OPcache can silently degrade performance without any error or warning. It’s less than a 10-minute read with real production impact.