What Mythex does:
Describe an app in chat. Mythex builds it with a live preview, attaches a database, and publishes it to a public URL in one click. Code and terminal are there when you want them.
Free plan, no card. 200 templates to start from:
https://t.co/Wbn0EkQyAf
Shipped: Mythex now works in 65 languages. It follows your browser, remembers your pick, and Arabic, Persian and Urdu run right-to-left.
The strings were the easy part. PDFs drawing pages backwards and Vite running out of memory in CI were not.
https://t.co/Wbn0EkQyAf
@mydevopsguy That's the line. Once it keeps data, it needs a backup you've actually tried restoring, and a way to export everything if you ever move. Most people learn the restore part on the day they need it.
@densancar The 38-0-0 lesson matches what we see: the version that shipped over a weekend taught more than any plan. For the game, which took longer: building it, or getting the first few hundred players before The Observer picked it up?
@nizamdesign Curious what the mess looked like from the design side. In the ones we've seen it's usually five slightly different button styles and spacing that changes on every screen, because each prompt started from scratch. Was it that, or the flow itself?
@piyush784066 I'd ask them to walk me through the last bug they hit: what broke, how they found it, what they changed. If they can explain that clearly, they can do the job. If they can only show the demo, not yet.
@BabaDev_ What saved my evenings: before starting, write down the one thing that counts as done tonight. When it works, commit and stop. The "while I'm here" fixes after that are where 2:30am comes from.
@itsalicesoul Vibe monitoring is the one that bites. The rest you notice while building; that one you notice when a user emails at midnight. A free uptime check and an error alert on your phone is enough to stop vibing it.
@SakshiSugandhi The cheapest review for a vibe-coded app isn't reading every line. Log in as two users and try to open each other's data by changing the ID in the URL. If that works, nothing else in the review matters yet.
@ChrisPirillo Agreed. The best ones are tiny tools for one person's odd problem: a chore wheel for one house, a tracker for one club. Nobody would fund them, and now nobody has to. The catch is keeping them alive once three friends start using them.
@asemota When an agent loop happens to me: stop, write two lines (what this screen should do, what it does now) and paste only that. Then ask it to explain the cause before it changes anything. Loops usually come from fixing symptoms one after another.
@pgpatel41 Solo: Expo (React Native), one codebase, with EAS building for both stores. Supabase or Firebase for auth and data. Store tip: get a TestFlight build out in week one. Review, privacy labels and in-app account deletion take longer than the code.
@jiggywastaken Video as code is underrated: pacing is a number, the fonts are your real brand tokens, and re-rendering after a UI change costs nothing. What did you render it with, Remotion or something else?
@stanlee0nX The easiest check anyone can run: sign in as user A, copy a record's URL, open it as user B. If B can see it, it's already hackable. That one test catches more than anything clever.
@LeoBuilds_ Manual coding still wins at one thing: knowing why the code is the way it is. Whoever can explain the last three bugs owns the codebase, whatever did the typing.
@bkarishma360 Engineering is deciding what the code must never do, then checking that it doesn't. You can skip reading every line if you test behaviour: two accounts, bad inputs, a reload mid-save. Skip that too and it's not engineering, it's generating.
@YashHustle_22 It's less about size than how much the agent has to hold at once. Big projects work when each task touches one module with a clear boundary and a test says what done means. Hand it the whole repo and yes, it falls apart.
@dev_maims 46 files for a login page is usually the agent never being told what already exists. One standing line, "reuse the existing auth, ask before adding files", plus reading the file list before the code, fixes most of it.
@shubhranshu2009 Depends what you're building. Cursor, Windsurf and Kiro are editors: best if you'll read the code. v0 shines at UI. If you want the whole app (database, login, hosting) from chat, a builder fits better. I'm building one: https://t.co/Wbn0EkQ0KH
@neerajjj6785 Enough to answer three questions about your own app: where the data lives, who's allowed to read each piece, and what runs when a user clicks save. Answer those and you'll spot most of the stupid stuff. Syntax is optional now; that model isn't.
@yevloop The 2 hours is real now, which is exactly why it stopped being interesting. The number worth posting is days from launch to the first stranger paying. Nobody posts that one because it's the hard part.