For Pablo · three company ideas
Not features for something that already exists. These are businesses worth starting, each with a working demo instead of a slide. Aval maps where the opening is; Torre and Tesoro are the same company taken to two different buyers.
The short version
There is a software company to build between regulated LatAm enterprises and the models they can't legally send data to.
One routing engine, one audit log, sold through whichever door the room opens first — cost or compliance. Owning in-region capacity is the unfair advantage, not the premise.
TesoroCFO sees recoverable AI spend.
TorreCISO sees governed, in-region routing.
AvalThe market map the other two came out of.
How they fit
Aval is the market map: settlement, company brains, sovereign inference. It explains why the opening exists.
What changed
Torre and Tesoro take sovereign inference and build it into an actual company. Torre sells governance to the CISO; Tesoro sells savings to the CFO. One engine, one audit log, two go-to-market motions — and both pull demand onto in-region compute, which is where an operator-backed founder starts ahead.
Recommended path
The AI treasury — the CFO door.
The AI control tower — the CISO door.
Three openings, three demos in one site.
The Rail · Aval — two AI agents transact through a non-custodial facilitator node; a bonded verifier attests identity and delivery, and issues a signed proof-of-settlement.
Company Brain · Bóveda — a sovereign, audit-logged company brain for a regulated bank, plus a marketplace of certified brains and a CLOUD Act comparison.
Inference · Núcleo — the cost-arbitrage calculator: frontier API versus sovereign open models, in-region.
That pair is the whole argument: the CFO sees money, the CISO sees jurisdiction, and it's one company serving both. Worth a conversation about whether it's the one to start.
All three sites are demonstrations. Every company, bill, rate, transaction and revenue figure is simulated client-side to show how the product would behave — none of it is live data, a quote, or a forecast. The pricing models are proposals to react to, not offers.