Manifestro
Team & Philosophy
Intelligence through Brain-Like Dynamics.

We are an independent AI research group headquartered in Almaty, Kazakhstan.

We are building a different path: expanding Liquid Neural Networks to better reproduce how living neural systems evolve in time, adapt, and reason. Smaller footprints and lower energy use matter—but they are outcomes, not the mission.

Who are we?

We are a founder-led collective of engineers and researchers focused on one core question: how far can LNN dynamics go when the target is biological realism, not parameter vanity.

Bagzhan Karl

Founder, CEO & Lead Research

Pragmatist backend engineer turned neural architect. Applying rigorous software engineering principles to LNN cores.

Alizhan Karl

Co-founder, Growth & Community

Navigating the narrative of open-source research. Ensuring our technology is accessible and actionable for the global developer ecosystem.

What do we do?

Our primary focus is the development of sanday—a high-performance speech synthesis and recognition ecosystem built on the foundation of Liquid Neural Networks (LNN).

Instead of optimizing for model bulk, we optimize for temporal intelligence: continuous-time state evolution, adaptive response, and richer neural behavior under real-world constraints.

0.76M
Parameters
Dense adaptive dynamics, tuned for temporal reasoning.
9.2 MB
Footprint
Compactness is a side effect of cleaner architecture.
2%
WER Reduction
Phase 2 achievement with < 2 validation loss.
LNN
Architecture
Continuous-time hidden states processing.
Why do we do it?

We do it to move AI closer to living cognition—not only bigger benchmarks.

Economic Empowerment

As LNN architectures become more expressive, practical deployment naturally becomes cheaper. We will pass that advantage to independent developers and startups, while keeping the primary R&D target on neural quality and adaptivity.

Technological Sovereignty

We want to enable "AI everywhere"—not just in the cloud, but in local, offline environments where real-time adaptation, privacy, and autonomy are paramount.

Our Philosophy

Biological Fidelity

Inspired by living neural systems, we prioritize continuous dynamics, state adaptation, and structural signals that better mirror cognition in motion.

Empirical Rigor

We reject the hype cycle. Progress is measured by hard metrics, behavioral quality, and reproducible experiments—not by parameter count theater.

Open-Source Covenant

The building blocks of speech technology should be a public good. The research core of our work remains open, inviting scrutiny and collaboration from the world.