Blog
What we run, written up in full, with the numbers and what did not work.
- Tiba paid five test claims, then we closed six security holes
One payout engine, rewritten three times. It pays a person for finished work only after a bill matches the payer's own record, and it leaves a public receipt either way. No real money has moved.
- 53,272 sales lines by afternoon. The look was still wrong 18 days later.
Skell is a decision list for a distributor's leadership. Agents built a working demo of it in about 14 hours on 10 Sep 2026, starting at 1:58am. The weeks after were the rebuilds. The look was still a draft on the morning of 28 Sep, and it went live that night.
- Papers: we built the paperwork checker before the client's workflow arrived
Papers reads the documents on a shipment, says which paper does not match, and names who to ask. I shipped the first version on a standard freight workflow, from research, because the client's own workflow was not in yet. The shipments on the site are samples.
- 91 programmes, 25 applications, 10 forms sent: what our applications agent learned
Dossair finds programmes, drafts the answers, fills the forms and keeps each stage. From 29 August to 28 September 2026 it held 91 programmes and 25 applications, and 10 forms went out. Most of what it taught us came from the ways it went wrong.
- Mrs. Shah sent 174 job applications. Then her tracker broke.
On 24 September 2026 a Grok Bot named Mrs. Shah hunted San Francisco marketing jobs for me. She applied only when the listing itself said the company would sponsor a visa. Her last count was 250 roles tracked and 174 sent. On 28 September her tracker page returned a server error. A fix went live that evening.
- Agents that buy, post and submit. Most devs stop at build.
Five things an agent did on my behalf with my card, my logins and an approval gate, each with the receipt: the actual exchange from the session transcript. A domain bought, a bounty posted and submitted, accelerators applied to, a hackathon won, a product built while I slept.
- 43 hours, 50 agents, one research run
We pointed fifty AI agents at our own market and let them run. 1,463 companies scanned, 374 matched, 730 named decision-makers at 99.7% reachable, at zero data cost.