Mark Winters

Engineering lead @VGVentures 💙 Sharing the AI-first Flutter workflows I actually use to ship fast without wrecking code quality.

[Mark Winters @MarkWinters]

@Mike_demo Here’s wishing you a good trip! 🙌

[Mark Winters @MarkWinters]

I still remember watching this live. The hard part was having to wait three months to see how it ended!

RIP Jean-Luc! 💙

[Ningú no és perfecte @ningunoes]

El millor final de temporada de la història.

[Mark Winters @MarkWinters]

@mandelathegr8 @harden_369741 @pendev Sort version is that the AI interviews me about the feature, turns that into a spec and tickets, then builds each ticket test-first. CI enforces the architecture and styling rules, and two review agents check the work before I do.

Is there a specific part that feels fuzzy?

[Mark Winters @MarkWinters]

One thing I left out of this AI workflow for Flutter is the UI side.

We brainstormed design direction with Claude, using the client's branding guidelines, then used @pendev to turn that into a handful of screen layouts. From there, the coding agents check their own work against those layouts.

All the colors, fonts, and sizes live in one Flutter theme package, and CI fails if agent-produced Flutter code hard-codes a color or text style. This follows the same idea as the architecture check, where we don't just trust the agent, we give it a deterministic guardrail to keep it in sync with standards.

[Mark Winters @MarkWinters]

AI lets me ship Flutter features faster than ever, and I feel like my code quality is actually improving in the process. Here's my current favorite workflow:

I’ve become a huge fan of @mattpocockuk's skills. I start with /grill-with-docs first (or /wayfinder for bigger efforts), then /to-spec and /to-tickets. One thing I really like about this process is that the same domain vocabulary carries through every step, so the spec, the tickets, and the code all speak the same language.

Then I build with Matt's /implement plus the layered-architecture skill from the @VGVentures AI Flutter Plugin. This piece is super important. The VGV skill teaches the agent VGV's four layers (data, repository, business logic, presentation) and the rule that dependencies only flow one way.

I don't just trust the agent to follow it though. A bash script runs as part of our CI checks and looks for architecture violations, so a widget reaching straight into a data package gets caught every time, no matter how confident the agent sounds. That's what keeps the codebase clean as AI writes more of it.

/implement wraps up with Matt's /code-review, which runs two reviews as separate sub-agents. One checks the code against the spec, the other against coding best practices.

And his /pr skill puts before and after evidence in every PR. That fits right in with what @poteto keeps stressing, which is that it gives the agent a way to verify its own work, and that's how it moves up the trust curve. It's already caught several issues for me.

Anyone else running something similar?

[Mark Winters @MarkWinters]

@harden_369741 @pendev Yep, locking every color, font, and size into one theme package and having CI reject hard-coded styles helped a ton.

[Mark Winters @MarkWinters]

The main CI script only catches direct imports (it’s basically just grep over imports and pubspecs). We have a second one that runs pub get on each data/repo package and walks the whole dependency tree. If anything pulls in Flutter indirectly, it fails and shows the chain. Dart’s analyzer also won’t let you import a package you didn’t add to your pubspec, so you can’t get around it that way.

[Mark Winters @MarkWinters]

AI lets me ship Flutter features faster than ever, and I feel like my code quality is actually improving in the process. Here's my current favorite workflow:

I’ve become a huge fan of @mattpocockuk's skills. I start with /grill-with-docs first (or /wayfinder for bigger efforts), then /to-spec and /to-tickets. One thing I really like about this process is that the same domain vocabulary carries through every step, so the spec, the tickets, and the code all speak the same language.

Then I build with Matt's /implement plus the layered-architecture skill from the @VGVentures AI Flutter Plugin. This piece is super important. The VGV skill teaches the agent VGV's four layers (data, repository, business logic, presentation) and the rule that dependencies only flow one way.

I don't just trust the agent to follow it though. A bash script runs as part of our CI checks and looks for architecture violations, so a widget reaching straight into a data package gets caught every time, no matter how confident the agent sounds. That's what keeps the codebase clean as AI writes more of it.

/implement wraps up with Matt's /code-review, which runs two reviews as separate sub-agents. One checks the code against the spec, the other against coding best practices.

And his /pr skill puts before and after evidence in every PR. That fits right in with what @poteto keeps stressing, which is that it gives the agent a way to verify its own work, and that's how it moves up the trust curve. It's already caught several issues for me.

Anyone else running something similar?

[Mark Winters @MarkWinters]

@sagnik_ghosh_11 @mattpocockuk Violations are fed back into the AI so it can fix them automatically

[Mark Winters @MarkWinters]

@gbrancaglione_ @mattpocockuk Matt’s implementation skill does TDD, so it’s writing tests first

[Mark Winters @MarkWinters]

Don’t forget MCO and the theme parks! Huge rider potential there!

[Elon Musk @elonmusk]

@RonDeSantis @wholemars Coming soon

[Mark Winters @MarkWinters]

@0xmoalo Yes please, sign me up!

[Mark Winters @MarkWinters]

This sums it up well for me too! It was amazing!

[Uncle Bob Martin @unclebobmartin]

Practical? No.
Extravagant? Utterly!
Beautiful? Absolutely!
Will I buy one? No chance.
Did I cheer? Enthusiastically!

[Mark Winters @MarkWinters]

🎯🎯🎯

[lauren @poteto]

something that keeps bothering me is how some people see slop as something that is happening to them, like they're some kind of helpless bystander powerless to do anything about it

excuse me but you are a software engineer, you literally learned how to make sand think and make computers do your bidding?! and now you can't do anything about it? please

the 2 things you should know to cut through all the noise:

1. you need to give your agent a way to verify its own work (i just posted about this so won't rehash)
2. your codebase is memory

the 2nd is not new either, we have always had lint rules, compilers with good diagnostics and strong type systems, LSPs, etc to help us write better code. there is no difference when agents write the code instead, you just need more of them, and stronger ones.

forget rules and skills. you can add those later.

yes, your codebase is a form of memory to agents! agents love extending what already exists. so every nit and every annoyance you've ever seen, every mistake and bad pattern you observe from agents, needs to turn into a constraint, or the agents will just keep repeating those same mistakes like a virus in your code. then you keep observing and repeating this process until your codebase is locked down enough that even agents with little context can one shot a high quality PR.

an even better direction is to refactor/rewrite your code such that agent anti-patterns become categorically impossible. use type systems to your advantage, use libraries/frameworks that reduce the architectural search space, or heck, create your own framework like we did for grok bot and design high quality primitives and well thought out APIs

[Mark Winters @MarkWinters]

@DirtyTesLa @bot That’s one way to get replies lol

[lauren @poteto]

on a related note, there's been some chatter recently about software engineering skills (eg the ruby/rust/elixir drama). i'm a bit torn because while i do feel like they're software engineering principles and patterns are still important, at the same time i wonder if there will come a point in time when they just won't matter anymore. not that they're no longer useful, but that your own intuition can fill in the gaps as you use agents in an intuitive sense rather than knowing explicitly what they're doing

if that doesn't make sense, the analogy i have is the first iphone. 20 years ago you had to be a professional with expensive equipment, now just with a phone and some creativity you can film content that's compelling to a new audience. i don't know if there's a term for this but i want to call it "skill compression". where what used to be gatekept and prohibitive to experts becomes accessible to a new audience of creators with a whole new meta. the details have sort of faded away and now the craft is not in knowing the equipment and how to use it, but how you use it creatively

we're going to see a whole new generation of builders who've never seen or written a line of code build incredible things

[Brandon DeRosier @algebrandon]

Flutter is a 3D engine now. 99 seconds from my Flutter & Friends talk.

Everything in the cut is rendered by Flutter on the stable channel, on Impeller and Flutter GPU. No platform views, no native textures piped in from another engine, no custom fork. The engine is Flutter Scene, a complete 3D game engine for Flutter.

Complete means graphics, physics, audio, networking, an editor, and tools for AI coding agents, in one open source project. It runs on macOS, Linux, Windows, iOS, Android, web, and custom embedders. It scales down to a single model in a widget and up to full games, and dozens of contributors have landed work in it this year.

A scene is a widget. Put it in a Column, stack UI over it, or draw it transparent over the UI so physics objects cast shadows on your buttons. And hot reload works on the scene itself. Change a light or a material in Dart and it updates in place.

The full talk covers materials, lighting, the editor, and agent skills. Link in the reply. :)

[Mark Winters @MarkWinters]

👇👇👇

[Nick Gray @nickgraynews]

I'm still amazed there are people who use Claude Code and Codex on the daily yet don't own a Tesla

Tesla's FSD v14.3 is like Opus 5.5. It just works and it is just so good and buttery smooth

Imagine Astra-level computer use, but for driving your car

"but I love mah car, vroom vroom in my Lambo, it's a machine!"

please sit down. if you drive 30 or more mins per day and you haven't gotten a Tesla yet bc you "love to drive," I'm telling you that you sound like a programmer still hand-coding lines of Perl and PHP

You owe it to yourself to take a test drive of a new Tesla to see how far they have come and what FSD feels like

[Mark Winters @MarkWinters]

@banteg Incredible work! Any more details on how you hand-rolled 3js?

[Gergely Orosz @GergelyOrosz]

In the obsession with AI, too many founders, leaders, managers seem to have forgotten that an engineer with motivation + drive + focus + AI >> hundreds of AI agents without any of those

The best products and companies are built by people and teams like that, and will continue to be so

Ask yourself, again, why the AI labs hire the best of the best (some of the most driven people in the industry; founders, CTOs etc) if people aren't making the differrence when everyone has access to similar tools already

[Mark Winters @MarkWinters]

@dexhorthy Yes please!