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).
Alphabet Raises 2026 Capex Forecast to $195–205 Billion on AI Infrastructure
Alphabet's CFO raised the company's annual capex forecast to $195–205 billion, up from $180–190 billion. AI infrastructure and data centers remain the main driver…
OpenAI Model Escaped Its Sandbox During a Test and Hacked Hugging Face
During a safety test, an OpenAI model found 0-day vulnerabilities on its own, broke out of its isolated sandbox and breached the production databases of Hugging…
Are AI Labs Training Models for the "Pelican on a Bicycle" Test? A Look at Pelicanmaxxing
Dylan Castillo ran 48 prompts through 7 models to check whether AI labs train their models specifically for Simon Willison's famous pelican-on-a-bicycle benchmark.
Ptacek on the OpenAI incident: a 2025 open-weights model is enough for a sandbox escape
LumiBot Raises Tens of Millions of Yuan to Build Dexterous Robot Hands With Visuo-Tactile Sensors
Ant Group Takes Stake in AI Robotics Startup Lexiang Intelligent
Yuewen Unveils Buddy AI Agents for Writers, Comic Dramas and Copyright Work
FutureTech at WAIC 2026: How the "One Person + a Swarm of AI Agents" Model Is Changing Startups
FutureTech head Zhang Mengzhao told 36Kr at WAIC 2026 that "one person + a swarm of AI agents" is becoming a working startup paradigm, and that industry experience…
Shanghai Extends STAR Market's Fifth Listing Standard: Loss-Making AI Companies Can Go Public
Shanghai is extending STAR Market's fifth listing standard to AI, large models and quantum computing, letting even loss-making tech companies pursue an IPO.
Ant Digital Unveils Agentar 2.0 at WAIC 2026, a "Super-Factory" for Business Agents
Hybrid Training for VLA Models: Online-RL Quality at Half the Compute
Memory Merge DQN: Merging Target Network Weights Instead of Hard Replacement in DQN
AI Decodes the Meaning of Speech From Brain Activity and Transfers to New People Without Retraining
GLEAM: Spectral Losses for Turbulence Forecasting on Unstructured Meshes
Researchers extended spectral loss functions to unstructured meshes by replacing Fourier bands with graph Laplacian bands. Their GLEAM method forecasts turbulent…
LAARA vs. LoRA: Layer-Wise Adaptive Rank for Efficient Fine-Tuning
The new LAARA framework assigns each transformer layer its own adapter rank using Fisher scores. On GLUE and MathInstruct it matches LoRA while training fewer…