Afaz Khatri

instead of vibe coding, do vibe designing of the spec. then let ai work on it. you will get surprisingly better output than vibe coding directly. and it is almost similar effort.

@ashleyhindle that is just perfect. 👌

Fuel's moving home 🫶 I've been checking on this domain for weeks! Snagged it the moment it dropped, pretty snazzy! 🙌

when working with spec driven development, what is the breakeven point when you should break the spec into mutliple items? how much big of a spec can ai agents handle nowadays?

@ashleyhindle Do you have any clear pricing right now for fuel?

Agentic coding is still hit and miss, it needs constant babysitting So I'm building Fuel to fix it and the early access waitlist is now open 🚀 An opinionated pipeline that ships production-ready code. Works out of the box with your favourite agents and models. Intelligent task routing so you don't go bankrupt with tokens 🫰 Automated quality gates, full mobile parity - free!

@intellectronica are you using any specific agent harness for doing the non-coding stuff? like openclaw or something?

The OpenAI $100 pro plan is frustrating. I got it with the intention of mostly using it for agentic non-coding stuff (I use my GitHub Copilot Pro+ for coding), but because the quota is so generous I don't manage to utilise much of it. And that makes me feel like a loser. 😢

@mattpocockuk indeed it is important to know when to use the framework and not. for features definitely yes, for small bugfixes, it can be just a one shot.

@mattpocockuk i have been happily using open-spec. helps me spec out features nicely and then ai takes over once we have clarity. i can also review the specs with the team to get on agreement, it has been so useful for me.

I just ran an AI coding course for ~2,000 people One massive piece of feedback was how dissatisfied people are with frameworks like BMAD, GSD, Spec-Kit Turns out that giving away control of context to a framework makes things a lot harder to debug My advice: own the process

haha, almost there!

seems this could be an inference/provider issue, it eventually completed.

i am a fan of opencode, but looks like pi will get me. @opencode gets stuck during the sessions, pi just keeps on going. opencode is literally here since last few minutes.

small fix but effective and predictable deployment pipelines. will catch issues earlier.

using worktree for the first time with ai agent. as the feature i am working on is quite large and we need to apply a bug fix, which can easily be applied on separate session on worktree.

x wants to sell itself, is it worth using the subscription? good to know i am getting significant impressions on my posts.

@freekmurze That's pretty awesome. Did you guys use ai to generate these? This practice seems really important on internal team structures as well.

🚀 We released a new thing: `spatie/guidelines-skills`. It turns our internal coding guidelines into AI skills, so coding assistants can follow the same conventions we use at Spatie. That means more useful AI-generated code, with less cleanup afterwards.

@nativephp @shrutibalasa Haven't checked the docs, but does it have transition support for screen navigation transitions?

We've got exciting news! 🤩 The wonderful @shrutibalasa is going to be your teacher on the NativePHP Masterclass 🎉 It's the most in-depth and practical course for building native apps with NativePHP First lessons available to early bird customers very soon! Don't miss out

@marcelpociot how did it turn out?

As it was so unnerving to watch I had to share it with you. You're welcome

it has been handed to ai now. 🤞

i did a few commits and push to update the specs only today. code will follow once the specs are clear and team is in agreement.

what are you using as you ide for laravel projects? or are you still using them or moved to use harnesses like codex, cursor, claude.

@kr0der i like my agent to have opinion and having to feel that with codex/gpt models. that helps drive useful/productive outcome of the sessions i am working with ai.

which one's gonna happen first: Opus becomes less lazy/becomes more thorough GPT becomes good at frontend and nicer to talk to whoever achieves this first will win a lot of users as it becomes the first "all-in-one" model that can do everything

@wanderingshade1 you can even explore that with ai and come up with strong plan/specs before starting to generate code.

@khatriafaz I always do this. You need an implementation plan

instead of vibe coding, do vibe designing of the spec. then let ai work on it. you will get surprisingly better output than vibe coding directly. and it is almost similar effort.