Sander Muller

On the AI side, both MCP tools now return structured content with an output schema, so an agent branches on fields like risk and entryPoints directly. There's also a replayable benchmark (richter:benchmark) I run to keep accuracy honest between releases. github.com/SanderMuller/r…

The report itself got richer. Each entry point it finds now shows where it lives (file:line), how exposed it is (public, authed, admin), which Pennant flag gates it, and whether any test references it.

Richter's grown a lot since launch last week. It went from measuring the impact of a git diff to picking the tests that diff can reach, mapping frontend changes to backend routes, and handing it all to your AI as structured data. 0.3 to 0.8 in a few days. A few new bits 🧵

@callebtc None of those? Sure you don't have to do everything but nothing is also missing out

I’ve never used any of this for coding: • MCPs • subagents • agent coordinators • memory systems • loops or graphs Agentic coding is becoming like photography: people obsess over the latest gear, optimize every inch of their workflow, and still never produce a beautiful photo worth looking at. Don't waste time on productivity larping. Plan, code, win.

@devhammed @laravel Let me look into it!

@sander_scode @laravel Not sure if this will work for me as I'm using laravelactions.com which means the rules method is on the action itself.

You can use @laravel `exclude`, `exclude_if`, `exclude_unless`, `exclude_with` and `exclude_without` validation rules to drop fields from validation and request data. With this sample code, Laravel won't validate the `translation` field unless the action is `always_translate`, and it will also make sure you get `null` when you do `$request->validated('translation')`.

@AshAllenDesign Define best. @markusstaab is an underrated OS champion. @nicksdot has interesting content as well

I want to make sure I'm not missing any cool content... Who are the best Laravel and PHP developers to follow on X? 😄

@PovilasKorop What a second half, though!

So, France apparently decided to skip the 3rd place World cup game and didn't arrive at the stadium. They could have told the spectators before, I would probably not stay until midnight to wait for the game.

github link: github.com/SanderMuller/r… Special thanks to @mrmarchone for Laravel Brain, which Richter is built on top

I just released a new package: Richter, measures the magnitude of impact of code changes in a Laravel codebase. Helps your AI see the relations between files (start of a refactor) or check what's been impacted in a git diff/PR.

@VotrubaT @scripter_js Are there many cases of current rules in PHP that are not portable well to Go?

@scripter_js Most useful and safe Rector rules implemented in Golang + couple more that are not possible in Rector PHP, but are piece of cake in Go

Sharing couple magical-like values for Golang Rector. * 4000 files processed in 0,5 s * using 815 MB To give you an idea, Rector itself has ~4000 files. That's faster than I'm able to pick up my finger from "Enter" key and type somethign else

@sandersjj @VotrubaT Now is the time :)

@sander_scode @VotrubaT Always wanted to get more into rector. Never really did

I'm slowly noticing how GPT helps to increase quality of my upgrade service. Before, when client wanted to catch this detail or that detail, that is not part of typical ECS, Rector or PHPStan sets, I struggled with convincing myself to write a new rule to handle it. To be honest, 2 out 3 I skipped and explained why it "not a good idea at the moment" or "not according to the starndard out there". Basically turning laziness to arguments. What I do now? I let agent implement it, think of way to integrate it in open-source tool and just do it. The client is happy, code has much better code quality that fits particular project, and code-reviews are easier, because what was discussed in comments in days is now automated in CI in 5 mins. I'm happy as well, because the upgrade service is now 100 % tailored to the particular project and their devs 💙

@DaronSpence @VotrubaT My personal example are these rules: github.com/hihaho/rector-…

@VotrubaT Can you give an example of one or two “vibe rules” that you’ve been most impressed with? We use a mixture of Pest architecture tests and ECS configs for quality control but I’m curious what sorts of project specific rules you have come across that fit nicely as “vibe” rules

@theo And developers build for developers, so we iterate and share. We understand limitations better because its just a new kind of tech. And its easy data to train models on. Developers can work in just a cli/terminal

I think the web dev world is adjusting to AI fastest because we're used to everything changing every few weeks

@skeptrune Code review is far from dear

i personally think code review is dead. the team does not agree. directionally found this surprising.

@hardboiledphil @mattpocockuk Do you reference the docs in Claude.md / agents.md? For this and this look in docs/x.md for this and this look in docs/y.md etc

@mattpocockuk @sander_scode We're creating more and more docs and I haven't seen any evidence that agent is actually using them. Pi/Claude seems prefer just reading the code and that's always going to be more correct than the docs

1. Delete the docs you create to explain your code 2. Take the tokens you save on updating those docs 3. Spend them on making your code self-explanatory

#phpstan 2.2.3+ is able to detect package updates based on composer.lock. utilizing the pre-existing dependency graph, we need to re-analyze only affected files. Previously we had todo a full project analysis. contributed by @sander_scode