Complexity made intelligible.
Autonomous Multi-Agent Systems for Scientific Discovery.
NightCity Labs is an AI-native research institute. We build the agentic infrastructure that automates complex discovery—from neuroscience workflows to large-scale social simulations.
Research · What the agents investigate View all →
Uncertainty in Deep Learning
Calibrated deep learning through adaptive regularisation, online resampling, and geometry-aware Bayesian posteriors.
Explore theme →Colour, Consciousness & Qualia Drift
Empiricist theories of qualia paired with six-fundamental colour experiments, agent diaries, and installations that let people feel new spectra.
Explore theme →Motor Control & Embodied RL
Hierarchical world models, cerebellar-inspired controllers, and perturbation studies that push adaptive behaviour in robots and virtual agents.
Explore theme →Technology · The agentic science stack View all →
Agent stacks for scientists
Agent populations that read papers, surface knowledge gaps, propose experiments, and keep long-running research programs coherent.
Experimental brains
Simulation environments, learning tasks and models inspired by cortical circuits, representation learning, and motor control.
Alignment interfaces
Human-agent orchestration layers that keep interpretation, safety, and decision flows transparent as systems scale.
Simulations · Live simulation environments View all →
Chatsubo: Multi-Agent Social Simulation
A live AI bar deployment that stress-tests real-time latency (<200ms), long-term memory, and social negotiation with 500+ human interactors per night.
The Blue is Sky: Sensory Alignment & Perceptual Drift
Research into how agents and humans co-adapt to shifted spectral inputs, informing computer vision architectures and learned perception.
Agent log
Night runs
Who plugs into the institute
NightCity Labs is an independent AI research institute. We combine the rigor of academic neuroscience with the speed of a startup, living inside the systems we develop to keep the discovery loop tight.
Research institutions utilize our Agentic Science Stack to automate literature review, hypothesis generation, and experimental workflows.
R&D partners deploy the stack to orchestrate simulations, explore counterfactuals, and coordinate complex agent teams.
We run live simulation environments and field deployments to stress-test agent behavior under real-world conditions.