Arin Agrawal 🚀

25K users. 4.9 stars. 150 countries. Zero paid ads. I was in my first semester of college when I launched. I didn't know what ASO was. Here's what actually worked: 🧵

@icanvardar My side project has around 25k+ users across 150+ countries Pressures me to complete that one first Btw checkout Calcquest:play.google.com/store/apps/det…

dear algo, show this to people who still haven’t finished a side project

@chams_builds honestly the slower pace might be a feature, not a bug. consistency over years > viral spikes anyway

when you start building in the 40s, your 𝕏 account grows at the same rate as your body 🐌

@DanielSmidstrup reliable multi-step reasoning without needing to babysit it. the moment you can trust it to run a 20-step process autonomously is when things get genuinely interesting

what’s the biggest AI problem you still want solved?

@sourabhkapure curious where antigravity lands for you long term — feels like a sleeper pick but the tooling's still rough around the edges compared to cursor

Yo chat, we cool here? Codex > Claude ~ Opencode > Cursor > antigravity > Gemini cli

@dejavucoder the algorithm basically treats your audience like they expire. meanwhile everyone who actually builds things is busy actually building things

the reach penalty for not posting 2-3 days is a lot. can a poaster not relax?

@TTrimoreau probably distribution moats and compounding network effects tbh. product can always be copied but a community that genuinely cares about you can't

What makes a startup uncopyable today ?

@Venkydotdev distribution on X for something youtube-adjacent is rough — the intent gap between the two platforms is massive. have you tried seeding with youtube creators directly?

build something for you tube creators. market it on X. and wonder why users are not getting.

@pcshipp the drop-off between day 2 and day 5 is where every onboarding battle is actually won or lost. most apps just never fight it

My app's users flow Day 1: purchased subscription Day 2: didn't open app Day 5 : open app Day 10: open app Day 15: cancelled subscription What's happening here?

@pmitu nothing builds conviction like having to explain something to a skeptic 40 times — eventually you either believe it or you stop

the best founders don't start by building, they simply lie to VCs

@thoniahilary close the laptop, touch grass, come back in 30 minutes and realize it was a missing semicolon the whole time

What do you do when your code frustrates you

@FunTech__ this is the real filter honestly. starting is cheap. it's showing up on day 60 when nobody's watching that separates people

The hardest part is not starting it's continuing when nobody notices.

@NoahKingJr mine's basically: 'pretend you're a user who's never seen this before, what's confusing?' catches way more issues than any debug prompt

What’s the most useful AI prompt you’ve ever written?

@MelkeyDev this tracks — claude seems to actually understand what the end product should feel like, codex just does what you literally typed

I cant explain it but it seems Claude is just better at writing iOS apps than Codex

@itzsam_ai getting people to pay every time. building is the fun part, it's the 'why should i give you money' conversation that's actually hard

As a founder, what’s harder for you? • building product • getting users to pay for it

@rcmisk D lmao, but genuinely word of mouth has been the only thing that's actually compounded for me over time

what's your current distribution stack? A) X + cold outreach B) SEO + content C) communities + word of mouth D) I ship and pray

@Venkydotdev version chasing is just procrastination in tech clothes. 5.5 can already do most things people complain about needing 5.6 for

Why is everyone waiting for GPT-5.6 when they can already get their work done with GPT-5.5?

@YashHustle_22 yeah, i review every single line honestly. not because i don't trust it, more because if something breaks i need to understand why

Are you checking every line of code written by AI?

@BratDotAI built CalcQuest solo — logic puzzle game, 25K users zero paid ads — and it felt real the first time a stranger emailed saying it helped their kid. not the number, just that moment

What would make you finally believe your side project is real? - first $100 online - first 1K users - first stranger feedback - first person asking for more

@unkonfined been building instead of posting. not sure that's wrong tbh, the product compounds differently than the audience does

Why don’t you have a million followers yet?

@abhijitwt tbh 'vibe coder' is starting to feel like a compliment. ship fast, iterate, don't get lost in params you'll never tune manually anyway

MFs don't even know what a token, tokenizer, context window, TTFT, ITL, reasoning token, sampling parameter, temperature, top-k, top-p, or penalty is, and still call themselves AI enthusiasts/engineers You're a vibe coder. 😭