Pillar · The engine
The AI-native engine
We work at machine speed with human judgement: the agents generate, the people decide.
An internal infrastructure of models and agents captures the resources and generates the inventions; people supervise and decide at every step. It is what makes the innovation process faster, safer and more effective, without ever taking human judgement out of the decisions that matter.
The engine
The heart: the generation of inventions
Throughout the ideation and development phases, the agents generate solution hypotheses (mechatronic, mechanical, electronic products) by making mechanics, electronics, chemistry and software coevolve together, with their reciprocal constraints of feasibility and manufacturability. At every step, people validate the hypotheses, discard the weak routes and steer the agents towards structuring the solution. It is neither autonomous generation nor simple assistance: it is an alternation, door by door, between the machine's combinatorial creativity and human judgement. It is precisely the co-design that crosses several domains at once that makes it different from existing tools, each locked into a single field.
Upstream
Capture
Upstream of the generation, the agents search for and qualify the resources (problems, ideas, capabilities, relationships, tax capacity) and route them towards the right project. The system can go hunting for value before it is obvious, like a radar intercepting weak signals.
Downstream
Validation
Downstream, the engine anchors novelty to evidence that can be defended. First it reasons about the state of the art, with open sources only, to measure how much is new and what already existed. Then external researchers run or repeat the experiments and deposit immutable, traceable evidence. Three levels of validation (automatic, experimental and independent) end in certification and in the IMV.
Human governance
Automation, with a clear limit
AI accelerates the capture of resources and the production of solutions, compresses the timescales and multiplies the projects that can be run in parallel. But there is one rule that is not touched: no binding commitment (financial, contractual, tax) is entered into automatically. The agents propose and prepare; a person approves, and every step is recorded. People do not carry out the repetitive work: they govern it.
The starting point
Where an invention begins
Before generating anything, the system locates the region of the problem, gathers what is already known and shows exactly where that knowledge runs out. Every invention thus begins at the edge of knowledge, by construction, and the effort, human and artificial, is spent only where it can add something.
Frequently asked questions
Frequently asked questions
Does the AI generate the inventions on its own?
No: it generates hypotheses, but at every phase a person validates and steers. It is co-design, not blind automation.
What data does it work on?
On open scientific sources and on the internal knowledge graph; closed sources, only with the consent of whoever owns them, in exchange for recognition.
Who takes the binding decisions?
Always a person; the AI does not on its own finalise financial or contractual commitments.
«The Analytical Engine has no pretensions whatever to originate anything. It can do whatever we know how to order it to perform.»
Your knowledge has a place here.
John F. Kennedy, 1962