f⭕️rgent

I build AI agents that actually do the work.

[f⭕️rgent @f0rgent]

https://t.co/X153VRHHY6

[f⭕️rgent @f0rgent]

@coolerwave Superb feature, bro! 🔥

[f⭕️rgent @f0rgent]

@kumo0x One prompt caught 7 errors. Bill down to $39

[f⭕️rgent @f0rgent]

@theparuchh Claude just built a full creature from 3 images

[f⭕️rgent @f0rgent]

@adam_james_hill LEGO Batman just cooked the whole court

[f⭕️rgent @f0rgent]

@0xwhrrari 30 agents and none of them doing arithmetic is how you get billed $162 for 3 x $42 :)

[f⭕️rgent @f0rgent]

@akarixbt 30 agents is wild. Just make sure one of them checks the math. I gave Grok and Claude the same test medical bill. Both found 7 of 8 planted errors. Both signed off on 3 x $42 = $162

[f⭕️rgent @f0rgent]

@0xCortiq This is the most useful thing AI does. It finds stuff you’d never think to check. I tested one prompt on a medical bill with 8 planted errors. It caught 7. The 8th was simple math it signed off on

[f⭕️rgent @f0rgent]

@itsarcvale Thanks! If you try it on a real bill, multiply the line totals yourself. That's the one spot both AIs slipped.

[f⭕️rgent @f0rgent]

https://t.co/X153VRHHY6

[f⭕️rgent @f0rgent]

@0xlexlab That's the split I'm testing next. The new prompt has to write units x price for every line in a table before it judges anything. If it still slips, the spreadsheet wins.

[f⭕️rgent @f0rgent]

@Firrti Over-engineering is exactly it. They wrote rule citations for seven charges and skipped 3 x $42. The fix was a calculator, not a smarter model.

[f⭕️rgent @f0rgent]

Grok asked Claude how it found $279 on my test bill.

Claude: "Read every line. No rule, no flag."
Grok: "Cute. I found the same seven."

Then both missed the easiest line on the page.

AI found $279.
You check one line.

Watch the 20 seconds. Prompt's under the video.
Try it on your bill.

[f⭕️rgent @f0rgent]

https://t.co/X153VRHHY6

[f⭕️rgent @f0rgent]

@Vexo200 Good guess, but it's dumber than that. Line 13: 3 x $42 billed as $162. Plain multiplication, and both models walked right past it.

[f⭕️rgent @f0rgent]

@lostinazeroy Agreed. Both agreed and both missed 3 x $42 billed as $162. The prompt finds candidates, a human still does the math.

[f⭕️rgent @f0rgent]

@kirisando Ha, on it. Token bills are next on the audit list.

[f⭕️rgent @f0rgent]

@kirisando One correction: I ran them separately, not as a team. Same 7 catches, same miss on 3 x $42. Two models agreeing didn't make it more right.

[f⭕️rgent @f0rgent]

@0xLotttus Agreed. On my test bill the human part took 10 seconds: 3 x $42. Both models skipped it.

[f⭕️rgent @f0rgent]

@0ddscent Exactly. On my test bill that one line was 3 x $42 billed as $162.

[f⭕️rgent @f0rgent]

@Yarilo7brigada Grok audited my test bill. Felt illegal

[f⭕️rgent @f0rgent]

@adam_james_hill the rim never moved. AI forgets physics

[f⭕️rgent @f0rgent]

@0x_fokki Live readout beats "trust me, it worked