A Laravel-heavy batch with some broader engineering thinking woven in.
Christoph Rumpel’s The Agentic Artisan is the most grounded take I’ve read on what it means to be a developer right now. His distinction between right shortcuts (scaffolding migrations, writing test setups, generating boilerplate) and wrong shortcuts (letting AI write an entire feature without review, generating your conference talk without your perspective) is precise and practical. The throughline: direction is now the scarce resource. You still need to know what good code looks like and understand architecture and trade-offs, because you’re reviewing AI output all day long. “Don’t outsource the soul” is the sentence I keep coming back to.
The Ray blog post on automatic Laravel output is short but worth five minutes. Ray’s Laravel integration can auto-intercept slow queries, duplicate queries, exceptions, and native dumps without manual ray() calls scattered through your code, you just flip a few flags in ray.php or your .env. If you use Ray and haven’t configured auto-interception yet, this covers the specific settings you want enabled during development.
Stefan Zweifel’s Laracon EU 2026 Recap is my favourite way to catch up on conferences. The talks he highlights: Dan Harrin on description-over-instructions as an abstraction philosophy for Filament, Ryan Chandler on unhappy path handling in Laravel (launching lesserknownlaravel.com right after), Tobias Petry on TimescaleDB aggregates for billion-row datasets, Simon Hamp and Shane Rosenthal on NativePHP’s native iOS/Android rendering, and John Drexler on shipping to production on day one. Good notes throughout.
Simon Willison’s Shopify/Liquid post documents a fascinating engineering story: Shopify CEO Tobias Lütke used an autoresearch pattern, having a coding agent run ~120 automated experiments to find performance micro-optimizations, to achieve a 53% improvement in Liquid template parse+render time and a 61% reduction in allocations. The agent had a benchmarking script as its feedback loop and a test suite of 974 unit tests as its safety net. The real insight is structural: a robust test suite is what makes this kind of automated research possible at all.
Tighten’s Blaze Under the Hood is the longest article in this batch and earns every word. It explains how Blaze accelerates Blade component rendering by shifting work from runtime to compile time, then proves the point by building two toy versions: Smol Blaze (constant folding only) and Lil Blaze (full pipeline with tokenizer, parser, AST walker, function compiler). The function compiler is the key: instead of the 20+ lines of ComponentAttributeBag overhead Laravel generates per component, Blaze emits a simple require_once + function call. If you have a Blade-heavy application, install it this week.