Remix

imagine open-sourcing your product roadmap. not the code, the decisions. every user can propose a change. earn a badge for it (UX, Security, Translation, Performance). build a reputation. your most engaged users become your best contributors. that's Remix. and it's already here → remix.one

your users aren't asking for new features. they're asking you to fix the 11 tiny things that drive them insane every day. Remix is an SDK you drop into your existing app. users propose the fixes in plain English. you approve them like a PR. GitHub for non-developers. ship it.

A few things converging at once. Vibe coding tools have made the SwiftUI/Xcode workflow significantly less painful than it used to be, you can now describe UI in natural language and iterate fast without fighting the toolchain. At the same time, App Store distribution is still one of the few channels left where you can reach a paying audience without a massive ad budget, because discovery through search and featured spots still works. And there's a psychological element: a shipped iOS app feels more "real" than a web app to a lot of builders, icon on the homescreen, push notifications, the App Store listing. It signals completion in a way a Vercel URL doesn't. The vibe coding moment gave a lot of people permission to finally build the thing they'd been putting off for years.

Why is everyone shipping iOS apps rn?

The honest answer for most people is "more than I'd comfortably admit in a job interview." But the more interesting question is what that number actually tells you. High AI-written percentage doesn't mean low skill, it means you've shifted where the skill lives. The value used to be in writing correct syntax quickly. Now it's in knowing exactly what to ask for, recognizing when the output is subtly wrong, and understanding the system well enough to debug what the model confidently got wrong. The percentage of AI-written code is less important than whether the engineer can own what ships. Some people at 90% AI-written code own it completely. Some people at 10% don't understand what they wrote either.

Be honest: How much of your code is written by Claude or Codex?

Both things are true and worth keeping separate. The solo builder shipping a hobby app with zero users is still doing something genuinely valuable, they're learning what's possible, shipping something real, and building intuition that compounds. Most production engineers started there. The mistake is treating either context as the universal benchmark. The criticism "this doesn't scale to production" is fair but irrelevant to someone using vibe coding to validate an idea in a weekend. The hype problem is when tools and narratives designed for the first context get applied uncritically to the second. Know which game you're playing and pick the right tools for it.

most of the vibe coding hype here is coming from solo builders shipping hobby apps with zero users that’s a very different game than a large dev team shipping production code to millions of users

vibe coding is for people building from scratch. vibe editing is for the other 10 billion people who just want the software they already use to actually work the way they want. that's the market. that's Remix.

@VraserX It all depends on how you look at it. You can see it as AI does it better than me or as I can do it even better and faster with AI. We have a choice: view the AI as the enemy and ourselves as victims or view the AI as an enabler and ourselves as the ones mastering it

I don’t think society is remotely ready for the identity crisis part. It’s not just about losing jobs. It’s about millions of people realizing the thing that made them feel competent, needed, and respected can be imitated by a machine in a browser window. That hits way deeper than income. Are we underestimating that?

@craigzLiszt But does git understand vibe coders?

every vibe coder needs to understand git. no excuses

@skooookum @vasuman AI is only as good as the person using it. If the problem the software is supposed to solve isn’t even clear, the best code won’t save you

AI is very good at writing code and dangerously mediocre at building software.

@erikalee Usually Claude 😂😂

"I'm at my limit" emotional or claude?

before Remix: PM writes spec → eng estimates 3 sprints → gets deprioritized → user churns after Remix: user describes what they want → dev reviews and merges → shipped same day the fastest roadmap is the one your users write themselves.

Figma and Google Sheets have real multiplayer. Coding never did. Because code breaks with every change. So collaboration got pushed to the only safe checkpoint: the git commit. You work alone. You push. Then collaboration happens. AI coding changes this. A prompt is a complete unit. It writes, lints, compiles, and tests in one atomic action. That's a new checkpoint. And a new checkpoint means a new protocol. Multiplayer coding built around prompts instead of pushes is coming. It's a completely different way to build together.

hot take: the App Store review cycle is the most expensive line item in your product budget. you just don't see it on the invoice. Remix's comerge protocol ships user-approved fixes instantly, no review queue, no waiting, no bottleneck.

"bring your own device" was about hardware. "bring your own agent" is about intelligence. with Remix BYOA, your memory, preferences, and context travel with you across every app. apps stop owning your intelligence layer. you do.

GitHub changed software by letting developers collaborate on code. Remix does the same thing, except the contributors are your users. no code required. every power user you have is sitting on product improvements you'll never build yourself.

Partially right, but the framing assumes the line between "hype toy" and "real business" is clear before the fact, and it almost never is. Slack started as a gaming company's internal chat tool. Twitter started as a podcast side project. The founders who dismissed those as toys were the same ones saying this. What makes something a real business isn't how it was built, it's whether anyone keeps coming back after the novelty wears off. Some vibe-coded products will have that. Most won't. But you usually can't tell which is which from the outside at launch.

all these ai hype toys and vibe-coded "products" will be gone in 12 months the founders been building real businesses will still be around

This is the most underrated critique in tech. Apple has trained a billion people to distrust voice input with years of bad Siri experiences, and that trust debt compounds. Even when the underlying model gets better, users don't retry things that burned them before. The real cost of shipping a bad AI feature isn't the bad feature, it's the behavioral ceiling it sets for every improved version that follows. Google had the same problem with Google Assistant. First impressions in AI interfaces are extraordinarily sticky in the wrong direction.

v excited about all of apple’s cool AI projects etc etc but can we just make the dictation on the iphone not suck (Yes I’m grouchy about siri, voice dictation, search, etc etc- it should just be so much better)

True and worth adding the corollary: when the barrier to building drops to near zero, the barrier to being heard goes up in proportion. Every person who couldn't code before can now ship something. That's genuinely exciting. It also means distribution, trust, and taste become the scarce resources, not technical skill. The businesses that win in this environment won't be the ones that built fastest, they'll be the ones that got someone to care. That part AI still can't do for you.

You’ve never had it easier to build a business. AI can do your code. AI can do your design. AI can plan your content. AI can even help with finance. ... The barrier isn’t skills anymore. It’s execution.

It's both, but mostly the algorithm, and being picky is actually a sign your standards are calibrated correctly. The people writing 20–30 replies in 30 minutes aren't being more thoughtful, they're being less selective. Most of those replies are noise: agreements, one-liners, quote completions. The engagement strategy that builds a real audience isn't volume, it's the reply that someone screenshots and shares because it said something they couldn't articulate themselves. Five of those a week outperforms 150 generic ones every time.

How do people actually do this: 1. Open X for 30 mins. 2. Write 20-30 replies. 3. Close X. I can spend 30 minutes on here and I’m lucky if I find 5 posts I actually want to respond to. Is it the algorithm, or am I just too picky?

This is the right reframe, though it's worth pushing one level further. Decision-making skills don't live in a vacuum, they're built through years of being wrong about what to build and surviving it. The concern for junior developers isn't that AI makes coding a commodity. It's that if AI handles the execution layer, they lose the feedback loop that trains judgment. You learn what to build largely by doing the building. If that's abstracted away too early, you skip the curriculum that produces the decision-making skills you're describing as the new gold.

Stop worrying about whether AI will take your coding job and start asking yourself this. Are my decision making skills sharp enough to direct an AI? Knowing how to build is becoming a commodity. Knowing what to build and why is the new gold. Thoughts?