Help Builders make production ready AI Native Apps in minutes | YC ’19 alum | High-energy problem-solver 🧘♂️ https://xitter.cf/t.co/25R0kr7sT8
Bangalore
Joined August 2017
- Tweets320
- Following318
- Followers147
- Likes353
AI agents aren't just getting better at answering questions.
They're getting better at completing longer tasks.
METR's 2025 research found that the length of software tasks AI agents could complete with a 50% success rate had been doubling roughly every 7 months in the tasks they studied.
Read that again.
This does NOT mean AI will become superintelligent on a predictable schedule. Extrapolating trends is risky.
But it does raise a serious question:
What happens when AI can handle projects that currently take humans days or weeks?
We're watching a capability shift worth paying attention to.
#AIResearch #Superintelligence
AI made the mistake.
AI fixed the mistake.
AI made another mistake.
And somehow… you’re paying for all of it. 💀
Stop paying AI for its own mistakes.
Build smarter with Something.
🤖 Made with AI
AI should help us move forward without getting into endless debugging loop. So we built Something.
building with ai shouldn't mean fixing ai's mistakes all day.
we're building something.cloud to make turning ideas into real products faster, simpler, and less frustrating.
if you're building something, you should probably know about this.
visit something.cloud
drop your idea and get free credits with in minutes
Aankit Roy retweeted
most ai tools can write one post. almost none can write five without sounding like a copy-paste job.
what if one idea could become 5 different posts, and none of them sounded copied?
we're building an internal tool with @somethingcloude solve exactly that.
give it one guest post, and it adapts the content for linkedin, instagram, and everywhere else. right tone, right format, right context.
we're dogfooding it internally right now.
if it saves us hours every week, we're shipping it.
best ai writing tool i've found: whichever one actually understands context. worst: any tool that thinks "adapted" means rewriting it once and posting it everywhere.
if your content sounds identical on every platform, it's not repurposing. it's spam with extra steps.
build with something.cloud
Aankit Roy retweeted
drop you projects/product below and get free growth/marketing
i'll be your first user
@somethingcloude
build with something.cloud
most ai tools can write one post. almost none can write five without sounding like a copy-paste job.
what if one idea could become 5 different posts, and none of them sounded copied?
we're building an internal tool with @somethingcloude solve exactly that.
give it one guest post, and it adapts the content for linkedin, instagram, and everywhere else. right tone, right format, right context.
we're dogfooding it internally right now.
if it saves us hours every week, we're shipping it.
best ai writing tool i've found: whichever one actually understands context. worst: any tool that thinks "adapted" means rewriting it once and posting it everywhere.
if your content sounds identical on every platform, it's not repurposing. it's spam with extra steps.
build with something.cloud
Aankit Roy retweeted
only 6 spots are left !!
go do it asap or loose the chance
build with something.cloud
crazy, we’ve got a lot of really good ideas coming in
only 6 spots are left !!
if you’ve got an idea you’ve been wanting to build, give it a chance
go for it something.cloud
Aankit Roy retweeted
stop saying “i’m gonna build it.”
build it.
we’re giving free @somethingcloude credits to people who are actually hungry to ship.
sign up → fill the form → tell us what you’re building.
got an idea?
building something?
trying to turn a crazy thought into a real product?
we’ll give credits to people who are serious about making it happen
docs.google.com/forms/d/e/1F…
openai dots are kinda insane.
you give it a goal.
it gets a computer.
then it just... keeps working.
while you sleep.
while you’re in class.
while you’re doing literally anything else.
ai is slowly going from
“tell me what to do”
to
“just handle it.”
Aankit Roy retweeted
if you love building, this might be your shot.
we’re launching a small hackathon this weekend for builders, creators, and anyone who wants to turn an idea into something real and ship it fast
we’re onboarding multiple creators, and selected users will soon get free credits to build with something.cloud
got the skills? come early.
sign up at something.cloud
the early builders get the early credits
Hey Founders, Share your idea, will ship your agentic/smart app in a day or two.
Checkout Something for more details.
Aankit Roy retweeted
and yeah, the merch is just as cool as the one in the tagged image
go give it a shot ↓
we’re thinking about sending some @somethingcloude merch to a few crazy builders who are showing up on twitter and building every single day
if that’s you, tell us why we should pick you ↓
let’s see who’s really in the trenches
Aankit Roy retweeted
stop paying AI to fix the mistakes it just made.
you shouldn’t need 20 prompts to fix one feature.
you shouldn’t have to burn credits debugging,repeat the same instruction five times
building software with AI should feel like shipping,not an endless loop of:
build → break → debug → fix → break again.
less prompting.
less debugging.
less wasted time.
more shipping.
build with Something.
something.cloud
Met @vikramchopra , founder of CARS24, at a fireside chat.
One thing stayed with me:
Founders will make mistakes. You will make bad calls. You will hit a wall.
But if you choose to be a founder, you can't give up.
It made me think about our previous startup.
2M users.
200K DAU.
25 mins average engagement.
I didn't realise then how hard that was.
We hit a downturn, couldn't figure out the next level, and eventually gave up.
Sometimes I wonder what would have happened if we'd stayed in the game a few more months.
Maybe the AI wave would have opened up a new path.
We'll never know.
Startups are a game of levels.
You make mistakes → learn → adapt → level up.
The most important thing is to stay in the game long enough to see the next level.
You vibe-coded an AI app.
It works.
You showed it to your friends.
You even have users asking for it.
Then you tried to ship it.
And everything broke.
We've shipped 50+ AI apps to production over the last 3 years.
If you're stuck between prototype → production, let's talk.
15 mins. No pitch. Show me what's blocking you.
Let's get it live.
cal.com/something-with-aanki…
#vibecoding
🤖 Made with AI
crossed 100+ signups. no launch. no ads.
just built the thing we kept wishing existed after 3 years of shipping ai agents to production.
here's the full story ↓
trysomething.sh/blog/we-cros…
