Zhemal Khamidun
Head of AI at Alpina Digital, CPO of AlpinaGPT (~30-person engineering & AI team). Founder of Vibe Academy & Hamidun Consulting, author of the AI channel «Готовим ИИшницу» (@jemal_hamidun).
BorisovAI tests MoE on an RTX 4090 and shows why perplexity breaks LLM evaluation
BorisovAI tested a pluggable-expert MoE on a single RTX 4090 and found that strong perplexity gains do not guarantee better performance on reasoning tasks.
LLM experiment shows how a model's "personality" takes shape in latent space
An experiment with a modular LLM suggests a separate latent vector can store not just text style but stable behavioral traits, resembling the scaffolding of a…
Ahead of GTC 2026, Nvidia teases an optical chip that could rework AI data centers
Before its March 16, 2026 keynote, Nvidia stoked the market with a promise of a "chip that will shake the world." Silicon photonics is the leading candidate for the…
Google pulls AI search feature that surfaced medical advice from random users
Unity tutorial shows how to build voice-driven NPCs with memory and world context
Harvard: AI is cutting junior hiring, and within three years the damage could spread across the whole industry
Moonshot AI unveils Attention Residuals, an alternative to residual connections in transformers
DeepSeek and GLM-5 beat Yandex in a test of 34 AI models for managers without a VPN
A large test of management scenarios found that DeepSeek V3.2 and GLM-5, both available in Russia without a VPN, clearly outperform Yandex's models and come close…
Google releases Gemini Embedding 2 for multimodal RAG with video, audio and PDF
Google has released Gemini Embedding 2, a model that embeds text, images, video, audio and PDFs in a single vector space, simplifying multimodal RAG.
Professor Dionne to present VINPix, nanophotonics and AI for multiomics on a chip
Why Yann LeCun's World Model Idea Doesn't Solve the Core Crisis of LLM Development
Alibaba folds its AI services into a new business unit to drive profit
Open-source Physical AI pipeline for the SO-101 arm built on ROS2 and LeRobot for 30,000 rubles
SimpleOne launches SimpleGen, an AI tool for building and deploying on its platform
SimpleOne has unveiled SimpleGen, an AI tool for generating solutions on its platform. A developer needs a repository, access to an instance and a workflow run in…
Google AI Ultra: Turning a Subscription into a Pool of Parallel Agents and Model Consensus
Google AI Ultra is pitched as the base for a multi-agent stack: run parallel Gemini workers, delegate routine tasks and cross-check decisions with a second model.