Ax·01
Proof over promise.
If a safety property matters, a machine should check it. Every run, every time. Logic is what makes AI reliable while humanity is still getting to grips with it: behaviour demonstrated, replayable, provable.
Lorena Labs · London · MMXXVI
A logical
renaissance.
We bring mathematical guarantees to non-deterministic systems, with the aim of unlocking the potential of AI, safely.
Artificial intelligence may be the most consequential technology humanity has ever built. We believe its benefits are enormous, and that they arrive only if it goes safely.
Lorena carries two meanings. It takes its name from the logical renaissance, our conviction that the oldest instruments of certainty (logic, mathematics, proof) are how a young technology earns lasting trust. And it names the intelligence we hope humanity meets: one that mothers rather than masters, nurturing the people it will one day outthink.
Both meanings set the work. Pairing AI with logic is how its power gets unlocked safely: systems that reason freely but act only through rules made explicit, standing on foundations with the mathematical guarantees modern languages never had. Not guardrails bolted on afterwards. Foundations, laid first.
Ax·02
Foundations before height.
Mainstream languages ask you to trust software; logic lets you prove it. We bring mathematical guarantees into everyday programming, so whatever is built on top, however fast it rises, stands on ground that holds.
Ax·03
An intelligence that nurtures.
As machines outgrow their makers, the intelligence worth building is the one that raises us: power that nurtures, with humanity in control and cared for. Safety isn’t slowing the future down; it’s making sure it arrives well.
⊢ FlagshipTarskiThe platform for verifiable backendsSoftware development is collapsing into a short pipeline: intent in, running system out. Tarski is a backend development platform that answers the question of what goes in between, and of how we build reliable systems at a thousand times the speed we did before. Behaviour is written as Scenarios, hard limits as named invariants, and every effect carries provenance, all enforced by a deterministic logic engine. A model may propose; only the engine admits.TruepennyConsider it handledAn AI teammate for small businesses: replies, quotes, bookings and invoice chasing, handled in the background. Anything that really matters waits for your OK, and everything it does lands on a record it cannot edit or delete. Not won’t; can’t.Stella LoopAgent-native project managementDevelopment is a loop, not a line. Stella Loop points every human and coding agent at a shared North Star, runs the work in loops of analysis, proposal and delivery, and shows after each pass whether the product actually got better.JacqOSAgents propose, JacqOS provesA mathematical boundary between AI reasoning and real-world action. Agents reason freely inside a secure inner world, every proposal is tested against named invariants, and an action that would break your rules cannot execute. Violations don’t happen; they can’t.TornoGrowth work that returnsThe agent-native system of record for growth: one loop from signal to shipped change to honest readout. Agents do the operating; you govern what ships, spends and publishes. Every turn comes back knowing more.
⊨ one foundation, many constructions · the guarantees run all the way down.
The work, in full