PRE-NEED
Detect what is likely to be needed next from bounded signals. Prediction can create a proposal, never an implicit mandate.
Machine Action Infrastructure
Intelligence is becoming abundant.
Action is the scarce right.
VALO sits between machine intent and real-world consequence — governing what may happen, proving what happened, and improving the economics of verified outcomes.
ModelIntentVALOActionProof
Models think. Agents intend. VALO governs action.
Systems execute. Evidence proves.
VALO is not a catalogue of unrelated AI projects. We develop bounded capabilities for the operational chain around machine work. A workload uses only the capabilities it needs.
Synapse
Let new work benefit from verified prior experience without carrying the full history forward.
FRAMLEIS
Research whether long-running machine work can preserve goal continuity under controlled tests.
HEIMEL / reht
Determine whether this exact consequence-bearing action may execute now under current authority, purpose and constraints.
TraXin
Bind risk and underwriting-relevant evidence to consequential work where the domain requires it.
EXUP
Improve existing machine workloads and establish whether the gain in cost, quality, latency, throughput or resources survives verification and production.
Veritas
Preserve evidence of what was authorized, executed and observed so outcomes can be inspected and replayed.
Governed state
Only governed, admitted evidence may affect future operative state. Persistence alone is not authority.
No machine-caused consequence without a fresh, verifiable right to act.
Intelligence → intent → risk & admissibility → fresh Handlingsrett → governed execution → consequence → economics & evidence → governed learning.
The stack around this boundary may change. Models, agent runtimes, clouds, hardware and external effect providers may be replaced. The Handlingsrett boundary remains. Prediction is not mandate. Risk and admissibility do not grant authority. Economic value does not grant authority.
Start with the problem. Not the stack.
Machine Action Infrastructure is the category. The Execution Layer is its technical placement. The commercial entry remains narrow: one workload, one consequence or one measurable economic problem.
AI workload economicsEXUP: improve one costly workload and prove retained economic value.
Consequential agent actionHEIMEL/reht: establish whether the exact machine action may happen now.
Repeated knowledge workSynapse: test whether verified prior experience improves first-pass outcomes while reducing context and resource use.
Future demandPRE-NEED: detect likely needs before the explicit request.
Risk-bearing executionTraXin: bind risk and evidence around the consequence.
Claims earn their place through tests.
VALO publishes bounded results and material limitations where appropriate. Product communication describes capability, measured outcome and evidence — proprietary mechanisms remain internal.