Research & innovation

Autonomous R&D programs, presented without code or promised availability dates, and not a commercial offer as such.

R01 Research

Menterias

Local execution of large language models.

Objective: Enable large language models to run on controlled local hardware.
Technical problem: Large models often exceed the memory (VRAM) available on a single machine.
Approach: Research on weight streaming across SSD, RAM and VRAM, use of Mixture-of-Experts architectures and hardware-resource optimisation.
Status: R&D / prototype program, not commercialised at this stage.

R02 Research

Persistent memory and knowledge representation for AI.

Objective: Explore a persistent external memory letting an AI system accumulate experience, relate concepts and progressively consolidate knowledge.
Technical problem: Go beyond a mere conversation history or an enlarged context window.
Approach: Explored capabilities around structured knowledge representation, progressive consolidation and context-based retrieval. Internal mechanisms are not disclosed.
Status: R&D project, distinct from Continuum (memory that could integrate into it).

R03 Research

Local intelligent-assistance architecture.

Objective: Study a local intelligent-assistance architecture integrating several complementary mechanisms.
Technical problem: Coordinate context, memory, rules and specialised components within a single system.
Approach: High-level work on context management, durable memory, persistent rules and coordination of specialised components.
Status: Research program, not announced as an available product. No claim of general intelligence or unlimited autonomy.

R04 Research

Persistent rules, control and alignment.

Objective: Explore a persistent-rules mechanism to govern an AI system’s behaviour in a traceable, adjustable way.
Technical problem: Ensure a system made of several agents and tools stays predictable, controllable and compliant with defined constraints.
Approach: Explored capabilities around adaptable profiles and constraints, safety limits, tool-access control and decision traceability. Internal mechanisms are not disclosed.
Status: Research component within the Continuum program, not commercialised at this stage.

R05 Research

Uncertainty and exploration management.

Objective: Study how an AI system can recognise the limits of its knowledge and respond to them appropriately.
Technical problem: Distinguish what is known, uncertain or unknown, rather than producing a confident answer in every circumstance.
Approach: Explored capabilities around uncertainty detection, confidence evaluation, seeking complementary information and learning from the unknown. Internal mechanisms are not disclosed.
Status: Research component within the Continuum program, not commercialised at this stage.

Information document — some statements must be completed and legally reviewed before final publication.

Describe your need or research direction: we will get back to you to discuss it.