A forward-thinking research laboratory advancing the frontiers of artificial intelligence and emerging technologies, bridging theoretical innovation with practical applications.
At Humus Labs, we conduct cutting-edge research that shapes the future of technology. Like humus enriches soil to support life, we enrich the technological landscape with foundational research that enables breakthrough innovations.
Our interdisciplinary team tackles the most challenging problems in AI, from fundamental theory to real-world applications, always with an eye toward ethical development and societal impact.
Rigorous methodology and peer-reviewed research that advances the field
Cross-disciplinary partnerships that spark breakthrough discoveries
Responsible research that considers societal implications and safety
Advancing neural architectures, optimization techniques, and learning paradigms to create more capable and efficient AI systems.
Learn More →Developing systems that understand, generate, and reason with human language across diverse contexts and modalities.
Learn More →Creating intelligent systems that can perceive, interpret, and understand visual information from the world around us.
Learn More →Building intelligent agents that can interact with and navigate the physical world safely and effectively.
Learn More →Ensuring AI systems are developed responsibly, with careful attention to fairness, transparency, and societal impact.
Learn More →Designing intuitive interfaces and interaction paradigms that enable seamless collaboration between humans and AI.
Learn More →Google releases Gemini 2.0 Flash Thinking, an experimental model that uses thoughts to strengthen reasoning during generation. Shows improved performance on complex tasks including math, physics, and coding challenges.
OpenAI's o3 model scores 75.7% on the ARC-AGI-1 benchmark at high compute, marking significant progress toward more general reasoning capabilities. Represents major advancement in AI's ability to adapt to novel tasks.
Anthropic releases Claude 3.7 Sonnet, showing marked improvements on software engineering benchmarks (SWE-bench Verified: 49%), agentic coding tasks (TAU-bench: 69.2%), and visual question answering.
Google DeepMind makes AlphaFold 3's code freely available for non-commercial research, enabling unprecedented access to state-of-the-art protein structure prediction. Accelerates drug discovery and biological research globally.
2024 Nobel Prizes recognize AI's transformative impact: Demis Hassabis and John Jumper for AlphaFold protein structure prediction, Geoffrey Hinton and John Hopfield for foundational neural network research.
Research shows that allowing models more computation at inference time (test-time compute) can match or exceed traditional training-time scaling. Opens new avenues for improving AI performance through adaptive reasoning.