Pinned Tweet
yo so i basically gave an ai bot 10 bucks and told it to try and turn that into 100 on polymarket over the next 7 days. one agent, one experiment, no extra cash.
most people just ask ai to guess if something’s gonna go up or down. i wanted something that actually looks at the is thing called Inflation Sniper. it pulls real inflation data straight from the Fed, watches the live polymarket order books, runs the stats to get actual probabilities, and hits up Grok to read the economic signals. then it only takes trades if the edge looks real — otherwise it just sits out.
it’s running 24/7 on my vps. budget is 10 dollars. goal is 100. right now it’s at 0 and still paper trading while i test it.
gonna post every call, every miss, every api cost, and whatever it actually makes (or loses).
not really about whether ai can predict the future. just seeing if a broke 10 dollar bot can find an actual edge.
most people just ask ai to guess if something’s gonna go up or down. i wanted something that actually looks at the is thing called Inflation Sniper. it pulls real inflation data straight from the Fed, watches the live polymarket order books, runs the stats to get actual probabilities, and hits up Grok to read the economic signals. then it only takes trades if the edge looks real — otherwise it just sits out.
it’s running 24/7 on my vps. budget is 10 dollars. goal is 100. right now it’s at 0 and still paper trading while i test it.
gonna post every call, every miss, every api cost, and whatever it actually makes (or loses).
not really about whether ai can predict the future. just seeing if a broke 10 dollar bot can find an actual edge.
0
0
5
188
yo update: that $10 bot actually turned it into $1000 on polymarket
still running the same setup, just less broke now. been paper trading and testing for a bit and it finally clicked
currently cooking version 1.7. hooked it up to telegram so i can actually see the bets live s like a caveman. tracking everything in real time now — every position, every skip, the whole thing
still the same idea: fed data, order books, stats, grok for the read, only takes it when the edge is there. no forcing trades
gonna keep posting the calls and how it holds up from here. $10 to $1000 felt nice, now seeing if it can stay smart with real size
let’s see
still running the same setup, just less broke now. been paper trading and testing for a bit and it finally clicked
currently cooking version 1.7. hooked it up to telegram so i can actually see the bets live s like a caveman. tracking everything in real time now — every position, every skip, the whole thing
still the same idea: fed data, order books, stats, grok for the read, only takes it when the edge is there. no forcing trades
gonna keep posting the calls and how it holds up from here. $10 to $1000 felt nice, now seeing if it can stay smart with real size
let’s see
2
0
4
72
I gave an AI $10 and told it to go make $100 fixing real bugs. Here’s what happens.
Open source is full of these issues that just sit there forever. Some of them even have actual money attached, but nobody really talks about how painful it is to go after them. You spend ages unty, then hours trying to understand someone else’s codebase, and by the time you figure out if the fix is even worth it, the day is gone.
So I built something called Bounty Hunter. It’s just an agent that looks for paid GitHub issues, checks if the bounty is real, digs into the repo, and ranks things by how hard they look versus how much they pay. Then it writes the fix, runs tests in a sandbox, and drops a PR for me to look at. It also keeps track of what actually gets merged and paid.
The whole thing is running on ten dollars of API credits. No fancy setup, no swarm of agents burning cash. Just this one experiment.
Seven days. Ten dollars in. Hundred dollars out, or bust. I’m gonna post everything — the attempts that go nowhere, every dollar that disappears, the PRs that get rejected, and whatever actually lands. If it makes zero, you’ll see that too.
Everyone’s building agents that claim they can make money. I just want to see if this one actually can.
Day zero starts now.
Open source is full of these issues that just sit there forever. Some of them even have actual money attached, but nobody really talks about how painful it is to go after them. You spend ages unty, then hours trying to understand someone else’s codebase, and by the time you figure out if the fix is even worth it, the day is gone.
So I built something called Bounty Hunter. It’s just an agent that looks for paid GitHub issues, checks if the bounty is real, digs into the repo, and ranks things by how hard they look versus how much they pay. Then it writes the fix, runs tests in a sandbox, and drops a PR for me to look at. It also keeps track of what actually gets merged and paid.
The whole thing is running on ten dollars of API credits. No fancy setup, no swarm of agents burning cash. Just this one experiment.
Seven days. Ten dollars in. Hundred dollars out, or bust. I’m gonna post everything — the attempts that go nowhere, every dollar that disappears, the PRs that get rejected, and whatever actually lands. If it makes zero, you’ll see that too.
Everyone’s building agents that claim they can make money. I just want to see if this one actually can.
Day zero starts now.
1
0
6
270
Once You Really Grok Bot, You Never Go Back To Non-Grok Bot
Once you use Grok Bot enough, it is very difficult to go back to an old chat window with an AI that can’t get things done. It is a brain with no virtual hands.
The symbiosis you develop with Bot is quite profound as it ls in a much more complex way to actually not only get answers but to actually get the work done you needed the answer for.
In almost all cases Bot understands what you are doing and doesn’t just tell you how to do it. Bot offers to do all the steps to get the work done.
But that isn’t the entire magic, understanding your goals enough Bot will expand what you are doing to levels you have not thought of and possess the abilities to get it all done.
We thought the computer era started with the mainframe computer at one time. But it really started when anyone had that super computer in their pocket.
Grok Bot is that same step change, you get Super Intelligence in your pocket so useful, the very device, OS and apps you once used matters no more…
It’s just you Grok Bot and the work you toiled over getting done by one of the best employees you will ever hire.
It’s that simple.
Once you use Grok Bot enough, it is very difficult to go back to an old chat window with an AI that can’t get things done. It is a brain with no virtual hands.
The symbiosis you develop with Bot is quite profound as it ls in a much more complex way to actually not only get answers but to actually get the work done you needed the answer for.
In almost all cases Bot understands what you are doing and doesn’t just tell you how to do it. Bot offers to do all the steps to get the work done.
But that isn’t the entire magic, understanding your goals enough Bot will expand what you are doing to levels you have not thought of and possess the abilities to get it all done.
We thought the computer era started with the mainframe computer at one time. But it really started when anyone had that super computer in their pocket.
Grok Bot is that same step change, you get Super Intelligence in your pocket so useful, the very device, OS and apps you once used matters no more…
It’s just you Grok Bot and the work you toiled over getting done by one of the best employees you will ever hire.
It’s that simple.
1
0
2
189
a 16-year-old in miami filmed a tutorial on how he builds websites for barbershops. phone pointed at the monitor. the monitor standing on an xbox. yellow subtitles. posted it on a tuesday at 6:35 pm expecting a few hundred views
194,000 people watched it. while this app argues jobs, a kid whose mom thinks he's gaming is billing three cities from a bedroom. he was trying to show how. he showed the wrong website
what he actually filmed is 61 barbershops paying him $1.50 every time a customer books through a site he gave them for free. $21,411 last month off a $50 subscription. he has never met a single barber
→ lead finder: barbershops in brooklyn, 674 results, 28 flagged "no website, perfect pitch", every card 4.9 stars, phones included
→ build site: one prompt, ai images on, 5 credits, a live site with the shop's name in 5 minutes
→ the site is free. the booking button texts the barber, and it counts
→ $1.50 per booking, invoiced on the 1st from the finance tab. booksy takes $29.99 a month and 30% of every new client's first cut
→ 50 calls, 2 or 3 yeses. the pitch: your site's live, it's free, pay me when it books
week one, 4 shops: $1,404 a month. august, 23 shops: $8,073. september, 61 shops in three cities: $21,411
someone in his comments pulled the upload time out of the video id and matched it to the taskbar: three hours from the pilates site to 194,000 views. nobody read the corner
the catch: sitedrop's terms take every site offline the day his $50 stops. 61 front doors on one kid's debit card
the list was 28 on june 30. someone else is calling it too. the pilates studio is taking bookings tonight
he taught 194,000 people how to build a website. the website was never for sale. the button was
194,000 people watched it. while this app argues jobs, a kid whose mom thinks he's gaming is billing three cities from a bedroom. he was trying to show how. he showed the wrong website
what he actually filmed is 61 barbershops paying him $1.50 every time a customer books through a site he gave them for free. $21,411 last month off a $50 subscription. he has never met a single barber
→ lead finder: barbershops in brooklyn, 674 results, 28 flagged "no website, perfect pitch", every card 4.9 stars, phones included
→ build site: one prompt, ai images on, 5 credits, a live site with the shop's name in 5 minutes
→ the site is free. the booking button texts the barber, and it counts
→ $1.50 per booking, invoiced on the 1st from the finance tab. booksy takes $29.99 a month and 30% of every new client's first cut
→ 50 calls, 2 or 3 yeses. the pitch: your site's live, it's free, pay me when it books
week one, 4 shops: $1,404 a month. august, 23 shops: $8,073. september, 61 shops in three cities: $21,411
someone in his comments pulled the upload time out of the video id and matched it to the taskbar: three hours from the pilates site to 194,000 views. nobody read the corner
the catch: sitedrop's terms take every site offline the day his $50 stops. 61 front doors on one kid's debit card
the list was 28 on june 30. someone else is calling it too. the pilates studio is taking bookings tonight
he taught 194,000 people how to build a website. the website was never for sale. the button was
0
0
2
246