Onenept Studios · applied AI
Our agents go through public data the way a good researcher would. They find the business sitting in it and go after it. Every step leaves a record you can click.
What we do
A restaurant in London whose website is broken. Two brands whose fans love the same films. It's all sitting in maps listings and taste data and job posts. Nobody has the hours to read it. So we built agents that do, and then act on what they find.
When an agent marks a job complete, that's a claim. We check it. That habit is where asheard came from, and it's why we've had fixes merged into DataHub and CALL-E.
Leads for freelancers, already qualified.
Pick a business type and a city. Warmleads hands back the ones that actually need your skill. Each comes with a score, a verified contact and the one thing that business needs most. You don't need a thousand leads. You need the right fifty.
From the warmleads demo
89
The Brass Fox Bistro
Restaurant · London · 4.6★ (320)
hold to reveal the contact + top need
syzygy (n.) when three bodies line up in a straight line
A sales floor with no people on it.
SYZYGY reads Qloo's taste data and looks for audiences that overlap when they really shouldn't. Bad Bunny fans and Lamborghini, say. Then a chain of agents takes it from there. Click any line on the floor and you get the evidence behind it.
Scout finds two audiences that line up when nobody expects it
Qualifier picks the side that would actually buy
Numbers builds the evidence from Qloo's data
Contact finder looks for the person who signs
Verifier checks the inbox is real
Strategist picks the angle
Writer drafts the message
Messages are drafted, not sent, until the mailbox goes live.
Did that call actually happen?
AI phone APIs can come back with the call finished, the job done and a confidence of 0.9, when an answering machine picked up. asheard reads the same payload and tells you what really happened. And it points at the exact field it read.
The API said
What happened
A recording picked up. Nobody spoke.
Built on CALL-E. The flat orange line, with the beep, is the voicemail.
Some of the checking happens in other people's code. These three are in, reviewed and merged by the teams who own the projects.
You can now raise an incident on a single column, not only on a whole dataset. The docs listed types the code didn't support, so we read the answer out of the source instead.
MERGED 20 AUG 2026
commit 8c823661e6The permission check for those column incidents. It now runs against the parent dataset, which is where the access rules actually live.
MERGED 24 AUG 2026
commit 1eeb440e65A call state reconciler. It reads what really happened to a phone call, because the status field alone can say done on a voicemail.
MERGED 11 SEP 2026
commit fc43128f2cThe company
An applied AI company. If it can't run on live data, we don't ship it. If an agent says it did something, there's a record of it.
Got public data you think has business hiding in it?
admin@onenept.com