Language interface
Translate human intent into structured quantitative objectives.
QGI vs. AGI
QGI is not “AGI for quants.”
Frontier general-purpose AI is increasingly capable of mathematics, code execution and tool use. The QGI thesis is not that general-purpose AI cannot calculate. It is that a quantitative-native architecture starts from a different cognitive substrate.
GENERAL-PURPOSE AIQUANTITATIVE GENERAL INTELLIGENCE
Native representationTokens / learned multimodal representationsQuantitative state / equations / distributions / constraints
Primary inputPrompt or observationObjective + state + data + constraints
ReasoningRepresentation-space inference and planningModel selection, computation, simulation and inference
Quantitative toolsInvoked when neededNative model / solver / simulator federation
OutputResponse, plan or actionComputed state, prediction, optimisation or experiment
UncertaintyMay be inferred or estimatedFirst-class calibrated object
MemoryConversation / agent memoryScientific memory and hypothesis lineage
VerificationTool checks, critics, testsNumerical, formal and evidence-chain verification
The distinction in one sentence
General-purpose AI is becoming quantitatively capable. QGI proposes intelligence whose native cognition is quantitative.
LANGUAGE-NATIVEQuantitative capabilities are added through reasoning, code, calculators, solvers and specialised tools.
QUANTITATIVE-NATIVEModels, state, constraints, simulation, uncertainty, experiments and evidence are the architecture itself.
Where they converge
QGI should be compositional, not anti-LLM.
Scientific literature
Read papers, extract methods and propose candidate hypotheses.
Code generation
Construct adapters, workflows and candidate implementations.
Explanation
Explain quantitative results without becoming the numerical source of truth.
Public claim boundary
LargeQuant does not claim that QGI replaces AGI or that QGI has been achieved.
The proposal is architectural: some classes of material intelligence may require quantitative cognition to be native, especially when systems must operate on finance, science, engineering, biology, energy, infrastructure or other measurable state spaces.