Manifestro
Research Alphav0.1.0-alpha

SANDAY

LISTEN

Pushing LNNs beyond benchmarks.
Toward living-brain-like behavior.
View Repository
Fluid
Temporal ResolutionBiological Dynamics
Dense
Synaptic ConnectionNeural Architecture
<50ms
State LatencyNeural Processing
CNN
LIQUID
DEC(O)DE
p_rob_ability
Architecture Stack

Acoustic Frontend

Refined ResNet blocks extract high-fidelity features from raw audio waves.

Liquid Backbone

Continuous-time neural networks adapt to the fluid tempo of human speech.

CTC Decoding

Alignment-free token prediction translates features into distinct characters.

A New Foundation.

Sanday is an active research project exploring the limits of biological mimicry. We are moving away from silicon-centric benchmarks toward a deeper understanding of neural fidelity.

Research Phase 2: Biological Logic