Blog

A short experiment with the Caveman skill, which makes AI coding assistants respond in short caveman-like sentences. It did save some tokens, but less than expected.

I spent the large part of two days reviewing a front-end that was built almost entirely by AI for a real product. Short verdict: locally great, globally drifting. Here's how I did it, how I used AI to help me out, and the ruleset I came back with.

Most posts about AI coding are about generating code. This one's about pointing Claude at an existing codebase to read it. The tools, the prompts, and where AI got it right and wrong.

Not every problem or package is equal. As serial package builder, how do we make the decision between relying on an external library versus making the robots dance to build us exactly what we need?

Our postcard page felt a bit neglected and hadn't been updated for a while, so we decided to create Kaartje (/ˈkaːr.tʲə/), a digital version of our wall, during a company-wide hackathon.

We built a way to all listen to a shared queue on the speakers in our office, to once and for all answer the question: 'What the hell is this and who put it on'?

We turned our guidelines into AI skills that work with any coding assistant.

Introducing ourray.app, a shared cloud-based Ray instance where everyone's debugging output streams live on one dashboard. Install spatie/our-ray, prepend our() before your ray() calls, and your dumps are visible to the world.

Manually running seeders on a production server is error-prone and can cause downtime if it goes wrong. Seed data from migrations to stay in the deployment pipeline without manual intervention.

Native HTML/CSS features like dialog, the Popover API, CSS Anchor Positioning, and customizable selects are catching up to what libraries like shadcn and Radix provide. Here's what changed, what's possible now, and how it might reshape how we build frontend components at Spatie.

With multiple large, long-running Inertia apps in production we know what we like, don't like, and what works. It’s time to formalize our thoughts and decisions as we’re starting new projects and onboarding a few developers this year.

Where possible, we try to stick with the popular choice for our tools. There's a lot of value in sticking to defaults. We only stray away from them when we have a really good reason. In the case of Laravel (and Inertia), the popular choice would be Vue, but we've landed on React. Why? Story time!

We've converted our PHP and Laravel coding guidelines into AI-friendly instructions so Claude Code and other AI agents can generate code that matches our established programming style.

Laravel's toArray collection method might not return what you might expect.

Need to handle image uploads in your Laravel app? In this post, we start with the basics and level up step-by-step—from a simple form upload to full-blown image handling with resizing, optimization, and a slick drag & drop UI. Whether you're building something quick or scaling up, there’s a Spatie package (or two) that can help you out.

Passkeys are very convenient for users to log in: they don’t have to type in an email/password, you aren’t redirected to a third party for authentication, and it’s fast.

We've made a new package that makes it easy to let your users authenticate to your app with passkeys.

Today we're launching php-operators.com, a little web page to browse, search, and learn more about all operators in PHP. Sure, you could just ask AI to explain them the you, but where's the fun in that?

We've been building Laravel applications with Livewire and Inertia for over 5 years. If you held an individual Spatie employee at gunpoint (please don't) and forced them to pick a favorite, you'll get a 50/50 split. It's like choosing between chocolate or vanilla ice cream: there's no objective winner (except vanilla). But when we look at the choice at hand as a team, it becomes easier to determine the right fit.