Research at Cogitave
Our research moves between models, runtimes, operating systems, browsers, and physical machines—where reliable autonomy depends on the layers beneath the interface.
Research principle
We separate architecture from evidence, prototypes from validated systems, and intended properties from demonstrated ones. The result is research that can become accountable engineering without losing its intellectual honesty.
Active programmes
We develop and study model architectures, adaptation, inference, and task-level evaluation as one system. The aim is not an isolated benchmark result, but useful behaviour within an explicit quality, latency, cost, and deployment envelope.
We study the runtime responsibilities beneath capable agents: process isolation, scheduling, identity, budgets, signals, memory, checkpointing, evaluation, and an auditable event stream.
Yuva explores a no_std Rust unikernel and micro-VMM with machine-checked leaves, fail-closed evidence, and a strict distinction between what is proved, measured, and still unknown.
Nimbus investigates a reproducible Chromium baseline, Nimbus-owned identity and shell, native content filtering, and a Rust-native, client-neutral MCP companion for local-first human and agent work.
Diyar connects perception, real-time control, simulation, embedded execution, and edge policy so autonomous behaviour can remain bounded when networks, compute, and sensor confidence change.
Massar investigates outbound-only remote access, signed authority, session evidence, and operational recovery for devices that must remain governable across unreliable or untrusted networks.
Cogitave Learn
Documentation, guided learning, hands-on practice, and the standards behind Cogitave products are published through the same corpus—with an MCP interface for agent access.
Open Cogitave Learn