Independent research project
Frida
Interpretive artificial intelligence — at the crossroads of hermeneutic philosophy, software architecture and language models.
The project
What can a machine truly understand?
Frida is an independent research and development project in artificial intelligence. At the crossroads of hermeneutic philosophy, software architecture and language models, it takes as its subject a decisive question: what can a machine truly understand?
The project does not merely seek to produce answers. It interrogates the conditions of possibility for an artificial intelligence capable of continuity, structured memory, hierarchisation of evidence and coherence over time. An artificial intelligence organised not around plausibility alone, but around interpretive rigour: making its uncertainty explicit, acknowledging incomprehension, suspending response when the conditions of validity are not met.
Frida begins with a diagnosis: current systems often excel at generation, but still struggle to sustain genuine interpretive stability. Between plausible response and situated understanding, the gap remains immense — and constitutes one of the major aporias of contemporary language models. It is this gap that Frida takes as its object.
Designed by a philosopher, philosophy teacher and independent developer, the project articulates technical experimentation, transmission and theoretical rigour. The aim is less to make a machine speak than to build an architecture capable of sustaining a more rigorous relationship to language, memory, context and dialogue.
Frida is less a product than a construction site: a living inquiry into the possibility of interpretive artificial intelligence.
Research areas
01
Understanding and interpretation
Distinguishing plausible response from situated understanding. Articulating the conditions under which a machine can be said to understand, and not merely generate.
02
Structured memory and continuity
Building a memory architecture capable of maintaining coherence over time, distinguishing traces by their probative value and managing their obsolescence.
03
Rigour and suspension of response
Developing the capacity not to respond when the conditions of validity are not met. Explicit uncertainty as an epistemic stance, not a system defect.
04
Hermeneutic architecture
Designing arbitration and inference layers that operate on contextualised, hierarchised, revisable traces — and not on simple surface probabilities.
Architecture — overview
A structure organised around interpretation
Frida's architecture distinguishes three fundamental layers: structured memory, contextual interpretation and an arbitration module. The latter is responsible for evaluating the coherence of inferences before any response, making residual uncertainty explicit, and suspending processing when the available evidence is insufficient.
The separation between generation and validation is one of the project's central invariants.
Participate
A construction site open to serious contributions
This site presents part of the ongoing work: hypotheses, structures, invariants, architectural directions, summary tables and evolving documentation. It does not present a closed system, but a set of problems currently under development.
Contributions and exchanges
The project is open to exchanges and contributions. Frida is not an open project in the vague sense — it is a demanding construction site, open to discussion and useful contribution. Rigour takes precedence over quantity.
Profiles from complementary backgrounds are welcome: