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).
GitHub Now Counts Draft Pull Requests Toward Open PR Limits
GitHub now lets repositories count draft pull requests toward PR limits, giving maintainers another tool against a flood of low-quality contributions.
How Distributed AI Training Is Reshaping Inter-Datacenter Networks
Google, Microsoft, AWS, and Meta are spreading large-model training across multiple datacenters — but that requires rebuilding infrastructure: bandwidth 14 times…
Moscow to Get 200 Robotaxis by the End of 2026, Half of Them Open to All Passengers
Deputy Transport Minister Vladimir Poteshkin set a target of about 200 robotaxis in Moscow by the end of 2026, with 100 of them open to ordinary passengers.
SaveLink updates SaveTest: requirements tracing, branch analysis, and run planning
Developer Used Claude to Write Rust Alternatives to Four Adobe Programs
Arenadata adapts its data platform for AI use cases: what we know as of October 8
MSI Prestige N16 Flip AI+ Heads to Russian Pre-Order: The First RTX Spark Convertible
YADRO and SberServis to deploy Russian infrastructure for banks and corporations
YADRO and SberServis have signed a memorandum on developing infrastructure solutions based on Russian hardware. YADRO's servers, storage systems, and network…
Waymo: Sober Drivers Face Nearly 4x the Risk at Night
Waymo cites a peer-reviewed study: alcohol is behind 30% of US road deaths, but nights are nearly four times riskier for sober drivers too. It calls for a Safe…
Wasserstein BaryBooster Turns Diffusion Realism into Event Forecasting Accuracy
A $6.3B Open-Weight Model Lost to the French: The Week of Mistral, Reflection and Moonshot
DeepLearning.AI and Qdrant Release a Free Course on AI Assistants with On-Device Memory
DeepMind Releases AlphaGenome Atlas: A Map of the Effect of Every Single-Letter DNA Substitution
Consistent Distribution Matching distills diffusion to 1 step with FID 2.04
A new distillation method needs no dataset or teacher runs: on ImageNet 256×256, one step gives FID 2.04 and four steps give 1.37 after 40 epochs of training.
PAIR Links Vision and Action in VLAs: OpenVLA-OFT Success Rate Rises From 51.4% to 65%
The PAIR framework teaches VLA models a shared intermediate representation between perception and action. On real robots, OpenVLA-OFT's success rate rose from 51.4%…