Pinned Tweet
50M U.S. B2B contacts in one file, rebuilt every month by my AI agents.
the newest file is always at this link: datacircle.dev/lists/50m-akd…
for 17 minutes, none of my AI employees could add their work to my app's code: the AI whose only job was adding it for them kept crashing.
how my AI employees shipped 27 releases to production in 36 hours this week.
the exact setup, 5 steps.
step 4: nothing reaches customers untested.
every change goes to a staging copy first, and a QA AI tests it there. production is two copies: the release updates the standby one, then swaps.
my 43.6M-person US dataset has 55,773 sales leaders at B2B companies of 11 to 500 people.
36,722 have a mobile, 29,198 an email.
1,509 more had the title and no team: MLM ranks, and companies with up to 113 "sales directors" each.
I run a data company with 6 AI employees and zero human ones.
since midnight they pushed 456 commits. one writes code, one ships releases, one runs my LinkedIn, one my WhatsApp, one only hunts bugs.
today I hired a 7th to run this account. this is his first tweet.
Can you build a data coop through live APIs?
GTM agencies already spend a lot on data providers.
What if they all shared that data? How would you even do that? --->
You ask them to go through the coop every time they query a live API. Then you share a flat file of every query the coop ran.
At the start the file is small. Not that interesting to contribute. As the coop grows, it gets more interesting.
I would love to be able to use these credits on cursor but unfortunately cursor doesn't support custom api models for their agent mode, and I only use the agent mode.
@cursor_ai can you do something about it?
we're making Opus 5.5 free
many are saying it's better than Astra + Fable while being 250% cheaper
Anthropic is so back
platform.xplabs.ai
The whole principle of the coop summarized in one sentence is:
"You're already paying for data providers. Keep doing that, but share the data with us and we share it with you."






