Joined Turing Post in April 2024. Studied control systems of aircrafts at BMSTU (Moscow, Russia), where conducted several researchers on helicopter models. Now is more into AI and writing.
How Microsoft's Rho-alpha, Google's Gemini Robotics, and six other breakthroughs built the bridge from VLA to VLA+ – the complete snapshot of where we are today
AI 101: DeepSeek mHC: Breaking the Architectural Limits of Deep Learning
your full guide to how residual connections started, how they evolved, and why DeepSeek’s Manifold-Constrained Hyper-Connections (mHC) are so important now
AI 101: Fusing Modalities: Basics + the New MoS Approach
We discuss how models fuse different data types, why it is important now, and what is special in the new Meta and KAUST's fusion method – Mixture of States
Can models add new knowledge without wiping out what they already know? We look at why continual learning is becoming important right now and explore the new methods emerging for it, including Google’s Nested Learning and Meta’s Sparse Memory Fine-tuning
AI 101: What is LeJEPA? The Theory Upgrade JEPA Has Been Missing
We discuss how Yann LeCun and Randall Balestriero's latest research drives JEPA foundations and puts it into solid, meaningful practice – and what it means for the community. Time to work on world models!
AI 101: What matters for RL? Precision! Switching BF16 → FP16
How this switch in precision influences the reinforcement learning accuracy and stability and why everyone, including Andrej Karpathy and Nathan Lambert, paid attention to this method
Everything you need to know about hybrid neuro-symbolic AI, how it blends strict logic and rules with neural networks, and why it shouldn't be overlooked
