Early feedback on Opus 4.8: it seems to be way less assuming overall, which is definitely a good thing

I'm finally using a task tracker for PixelWatcher (I was using a notebook until now 😅). I settled on GitHub Issues, since I'm already using GitHub anyway, and I don't feel the need for something like Linear at the moment. Claude can talk to GitHub via `gh`, which is good enough.

Anyone following the AI space has come across the term "evals", usually in the context of training models. But what do evals look like on the application side? This article covers why I ended up needing them for PixelWatcher, and how I implemented them:

Anyone following the AI space has come across the term "evals", usually in the context of training models. But what do evals look like on the application side? This article covers why I ended up needing them for PixelWatcher, and how I implemented them:

Pretty neat and straight-to-the point explainer of the difference between jobs, actions and services, with good examples. Only thing I would change is make the action invokable medium.com/@radenadr...

📢 Our next event is up! On Wednesday 10 June, Christopher Miller Christopher Miller will show us how to time travel through Event Sourcing 🔮 📅 Details and registration 👇 www.meetup.com/php-s... #php

Tout dernier jour pour prendre votre billet pour l'AFUP Day 2026, demain il sera trop tard ! N'hésitez plus, rejoignez-nous la semaine prochaine à Bordeaux, Lille, Lyon et Paris, ensemble profitons du partage des savoirs en PHP ! event.afup.org

Just ended a Claude Code session (Opus 4.7 xhigh) that lasted a couple of weeks. I can't say I noticed any performance degradation. Multiple compaction rounds kept the context relevant enough (my main reason for keeping it alive), and multi-step workflows ran as usual.

Some AI behaviour is funny to observe. Lately, Claude started leaving verbose comments. I ignored it for a while, then it seemingly picked up on its own habit and doubled down, leaving even longer comments in more places. A benign example, but also a reminder to stay vigilant 👀

Somehow missed the `mixin` PHPDoc tag until now. On a PHP trait, it tells static analysers (PHPStan, IDEs) to type-check it as if it lived inside a subclass, so they don't flag calls to methods that only exist in the subclass as invalid. phpstan.org/writing-...

There’s currently an uptick in people contacting me on LinkedIn offering to build my website. I’m inclined to feel offended.

Currently translating a presentation from English to French. It makes you realise how much more concise and efficient the English language is. French is completely messing with my slides' layout 😩

I've also been using the `/grill-me` skill instead of plan mode. So far I like it – the model asks questions one by one, which feels like more work, but in a good way: it forces you to focus on one aspect at a time, instead of glossing over a verbose plan

After using Opus 4.7 for a few days, I feel the same as everyone else: the model feels less smart overall (and has a slightly more annoying personality), but it's *much* better at keeping track of multi-step workflows, which seems to confirm it's geared towards agentic work.

Dans l'interview du jour, @osteel.bsky.social partage sa vision d’un code plus maintenable face à l’IA, sans perdre le contrôle du pipeline. On vous laisse découvrir cette interview enrichissante, avant d'en apprendre plus ensemble lors de son talk le 22 mai à l'AFUP Day 2026 Lille ! buff.ly/o7t3Y7v

Our Executive Director, Elizabeth Barron, has been listening to your feedback about the PHP ecosystem to help inform The PHP Foundation's 2026-2027 strategy. She shares what she's learned in our latest blog post. Exciting opportunities await! ✨ 🚀 thephp.foundation/blog/2026/04... #php #phpcommunity

Guess who’s speaking at Laravel Live UK in June 😌 https://laravellive.uk

I enjoyed this post about how engineering roles are being redefined in the age of AI. What are the traits that truly matter in employees and candidates? In short: understanding > implementation

Just used Peter Fox's `composer-upgrade` skill on PixelWatcher. Amazing work. It breezed through minor dependency upgrades, reported on the only major one, pulled the corresponding changelog, and checked whether my code would be impacted. Then it ran the test suite. Check it out!

Lille 😌

I have to say, having Claude use Nightwatch's MCP server to pull the latest issue and suggest a solution within minutes is freaking cool

How to cache static pages in a Laravel application (full post at the end of the thread): 1️⃣ Create a separate route file for static pages 2️⃣ Create a dedicated middleware adding cache headers to responses

ChatGPT is far too verbose, and no amount of customisation seems to fix it. I end up losing my train of thought in most conversations. We're expected to manage these tools' context windows, but what about ours 😩

Amazing post by Linear acknowledging their original product is doomed. I just got comfortable keeping all useful context inside PixelWatcher's repo, but as I get involved with more early-stage products, a context-management tool starts to make sense

Did my first attempt at automating an entire workflow with Claude today (plan > implement > simplify > review > test > PR), the idea being to get to a decent first draft for a task, which I can review and refactor. But it keeps stopping at random steps. Not sure how to fix this

Finally taking the time to use my faster.dev subscription and look at the video series. I almost skipped the "Claude Code from Scratch" one entirely, but decided to have a quick look anyway. Guess what – I didn't know you could customise the status line 🤦‍♂️

Giving Codex a go while Anthropic gets its shit together. Managed to import my Claude setup fairly easily and it's working well overall, but my project runs via Docker and it seems completely unable to interact with that. Is that a massive limitation or am I missing a trick?

First the reduced session limits, now an Opus 4.6 outage. Time to give Codex a shot I guess!

Floorboards are great until you have a kid

I just love refactoring with AI. It's like pair programming, really. I'm the navigator, keeping the big picture in mind, and implementation is so fast that I never lose my train of thoughts. No more "I know I should do this, but it would take too long to do it now"

📢 Our next event is up! On Weds 8 April, Gina Banyard and Oliver Davies will join us for two great talks about backwards compatibility and Nix for PHP developers. 📅 Details and registration 👇 www.meetup.com/php-s... Big thanks to Jump24 for sponsoring this event! https://jump24.co.uk

The recording of my talk Escaping the Code Maze at PHP UK is available! Big thanks to the organisers. I just need to learn how to speak into a microphone on a stand 😬

I sat down with Claude Code to build a test analyser. Our DX survey keeps flagging test efficiency, and I had a hunch our test pyramid was off. Wrote it up with some interactive bits. acairns.co.uk/posts/evalua...

Just spent some time reviewing my agent skills using Anthropic's `skill-creator` skill (install via `/plugins` in Claude Code). Not just to improve the skills themselves, but also to tighten integration between them. Will do this regularly from now on

Claude Code Remote Control – start a session in your terminal, monitor it and update it AFK (e.g. on your phone)

Been using Claude Code's `/model opusplan` all day (Opus 4.6 for planning, Sonnet 4.6 for implementing). Working a treat so far, and feels like a good balance in terms of token consumption 👌

I'm two episodes into The Night Manager, and I have to say, so far it's incredibly silly. Is the whole series like that? I'm having a hard time understanding the rave reviews at the moment 😅

Oh merde Bruno Salomone est mort 🙁

Really enjoyed speaking at PHPSW yesterday. Friendly vibe, great conversations… and good beer! Now checking out Bristol for the next couple of days 🌆

I have some thoughts on that PR review tool that Anthropic released yesterday. artisan.build/posts/the-tr...

On my way to Bristol! There’s still time to sign up for tonight’s PHPSW meetup

On the one hand, the SaaS model is dead. On the other, everyone's building one. Whichever narrative you subscribe to, this might seem like a strange time to build PixelWatcher. So why am I doing it? A few reasons in this short thread ⬇️

I understand that the Cursor team is moving fast and shipping a lot, but having to constantly restart the IDE to apply updates is really annoying

📢 Bristol! I'm speaking at PHPSW this Wednesday (11 March). I’ll be giving an extended version of my talk Escaping the Code Maze, introducing the Closed-by-Default Principle. Are you based in Bristol or in the area? Come say hi!

Goddammit Claude! What am I gonna do, open my own PR like a peasant? 🙄

Just tested Claude Code's recently-added `/simplify` skill on PixelWatcher's code (Laravel codebase). Pretty impressed with the result – it found some non-trivial things to improve. Definitely adding it to my workflow

Just wrote my first custom PHPStan rule... except I didn't. Claude did. All I asked was "How can we enforce this rule?" Claude was aware of PHPStan in the project, so went ahead and wrote the rule, and tested it by introducing a temporary violation. What a time to be a developer

As I'm working on PixelWatcher, my Claude setup is slowly taking shape. Relevant skills are listed later in the thread if you're in a rush, but for context: I've just completed a full implementation plan, based on existing documents (PRD, technical spec, stack, HTML wireframes...).

Man, I really welcome FrankenPHP, if only so I never have to look at a PHP-FPM config again 🙏

My SaaS officially has a logo and a name! Can you guess what it does 🤪 Teaser page with newsletter form to be informed about launch and get early access in the thread Also in the thread a quick tutorial on how to use Remotion and Claude to animate your own logo