ZK
Editor-in-Chief · CPO of AlpinaGPT

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).

10,555Published articles
50+Verified primary sources
6Language editions (RU/EN/ES/PT/FR/AR)
SciΠvot benchmark: the best LLM picks the right plan 63.5% of the time, versus 73.4% for humans
Research·Hugging Face Daily Papers

SciΠvot benchmark: the best LLM picks the right plan 63.5% of the time, versus 73.4% for humans

Oct 15, 2026·2 min
GlobeBench: The Best LLM Correctly Predicts Environment Responses in Only 49.57% of Tasks
Research

GlobeBench: The Best LLM Correctly Predicts Environment Responses in Only 49.57% of Tasks

Oct 15, 2026·2 min
A Frozen 16 MB Vocabulary Matches a Trained One in Small LLMs
Research

A Frozen 16 MB Vocabulary Matches a Trained One in Small LLMs

Oct 15, 2026·2 min
NVIDIA RTX Spark N1X Runs a Local Qwen3.8 27B Agent in 18.7 Minutes
Hardware

NVIDIA RTX Spark N1X Runs a Local Qwen3.8 27B Agent in 18.7 Minutes

Oct 15, 2026·3 min
Vera Rubin NVL72 beats GB300 by up to 5.7x in tokens per megawatt in AA-AgentPerf
LLM·Artificial Analysis

Vera Rubin NVL72 beats GB300 by up to 5.7x in tokens per megawatt in AA-AgentPerf

Artificial Analysis published the first Vera Rubin NVL72 results in its agentic coding benchmark: up to 5.7x throughput per megawatt over GB300 and a record 906…

Oct 15, 2026·3 min
NVIDIA releases Nemotron 3.5 Super: 26 AA points and 630 tokens/s
Hardware·Artificial Analysis

NVIDIA releases Nemotron 3.5 Super: 26 AA points and 630 tokens/s

Nemotron 3.5 Super VL doubles its predecessor's Intelligence Index score and becomes the best American open-weight model. The compact Bolt 74B-A9B launched…

Oct 15, 2026·3 min
LIBERO-Industry tests 5 robot policies on 216 factory tasks
Robotics·Hugging Face Daily Papers

LIBERO-Industry tests 5 robot policies on 216 factory tasks

The new LIBERO-Industry benchmark adds a conveyor, narrow action windows, and tight tolerances. Five generalist policies lose performance as conditions get harder.

Oct 15, 2026·2 min
Cohere pitches utilities on running North AI agents inside their own perimeter
Business·Cohere Blog

Cohere pitches utilities on running North AI agents inside their own perimeter

Cohere's new guide for utilities argues that AI in critical infrastructure hinges less on model capability than on where it runs, what data it touches, and what…

Oct 15, 2026·2 min
Top themes
01LLM-Attacker Benchmarks Split-LLM Attacks: Gradients Raise Leakage to 37.77%Oct 15, 2026·2 min02Data Rewriting Beats DP-SGD on the Privacy-Quality Trade-OffOct 15, 2026·3 min03Kolab: 115 Hours of Recordings Show That AI Agents Don't Understand Their Human ColleaguesOct 15, 2026·3 min04DialFork measured procedural hallucinations in agents across 36,278 dialoguesOct 15, 2026·3 min05OpenAI Engineer Hid a "Bug" in Toilet Paper and Got Caught Within 48 HoursOct 14, 2026·2 min
OpenAI shows how ChatGPT builds a catering pitch for a Toronto restaurant
Business·YouTube: OpenAI

OpenAI shows how ChatGPT builds a catering pitch for a Toronto restaurant

Oct 14, 2026·3 min
Natura-1 Teaches Robots Dexterous Hands With a Million Hours of Human Experience
Robotics·Jiqizhixin (机器之心)

Natura-1 Teaches Robots Dexterous Hands With a Million Hours of Human Experience

Oct 14, 2026·2 min
Apple's LensVLM compresses long agent context into images and teaches the model to "zoom"
Research·Jiqizhixin (机器之心)

Apple's LensVLM compresses long agent context into images and teaches the model to "zoom"

Oct 14, 2026·2 min
MemSilicon Completes Tape-Out of MountTai 2026, a 1 Gbit 1T4R RRAM Chip
Business·Jiqizhixin (机器之心)

MemSilicon Completes Tape-Out of MountTai 2026, a 1 Gbit 1T4R RRAM Chip

Oct 14, 2026·2 min
Anthropic tested whether AI can copy itself through a TV set-top box
LLM·Jiqizhixin (机器之心)

Anthropic tested whether AI can copy itself through a TV set-top box

Anthropic's red team studied whether AI can move between devices like a virus. Small models already adapt to unfamiliar hardware and set up inference on their own.

Oct 14, 2026·2 min
Google DeepMind Launches Gemini 4 Argon and Cuts API Prices in Half
Business·Jiqizhixin (机器之心)

Google DeepMind Launches Gemini 4 Argon and Cuts API Prices in Half

Google DeepMind has opened global access to Gemini 4 Argon, the first frontier model in the family. API prices are temporarily down 50%, with a focus on coding and…

Oct 14, 2026·2 min
Tsinghua Unveils SRHarness and MechBench: AI Agents That Look for Formulas and Mechanisms
Research·Jiqizhixin (机器之心)

Tsinghua Unveils SRHarness and MechBench: AI Agents That Look for Formulas and Mechanisms

Tsinghua's data science and intelligence lab proposed SRHarness for agentic symbolic regression and the MechBench benchmark to help AI find formulas and mechanisms more reliably.

Oct 14, 2026·2 min
Tsinghua Unveils Jev-LDE: Edit ICL Examples Once, Boost 1-Shot Accuracy
Research·Jiqizhixin (机器之心)

Tsinghua Unveils Jev-LDE: Edit ICL Examples Once, Boost 1-Shot Accuracy

Oct 14, 2026·2 min
Recursia Beats Nvidia ASPIRE by 10 Points on Long-Horizon Robot Tasks
Robotics·Jiqizhixin (机器之心)

Recursia Beats Nvidia ASPIRE by 10 Points on Long-Horizon Robot Tasks

Oct 14, 2026·2 min
ETH and Roche Train AI to Predict Where C–H Borylation Will Occur on a Molecule
Research·Jiqizhixin (机器之心)

ETH and Roche Train AI to Predict Where C–H Borylation Will Occur on a Molecule

Oct 14, 2026·2 min
PoS Method Catches AI Agents Stuck in Loops of False Belief: 36 Repeats in ALFWorld
Research·Jiqizhixin (机器之心)

PoS Method Catches AI Agents Stuck in Loops of False Belief: 36 Repeats in ALFWorld

Oct 14, 2026·2 min
Ego Data Gets a "Full Body": A New Scaling Path for World Models
Robotics·Jiqizhixin (机器之心)

Ego Data Gets a "Full Body": A New Scaling Path for World Models

Oct 14, 2026·2 min
Google DeepMind and Singapore Release Gemma-Based Scam Detector as Fraud Losses Top S$4 Billion
Business

Google DeepMind and Singapore Release Gemma-Based Scam Detector as Fraud Losses Top S$4 Billion

Oct 14, 2026·3 min
SAE-IDF Matches Symbols Across Different Systems and Beats a Baseline by Up to 45 Points
Research

SAE-IDF Matches Symbols Across Different Systems and Beats a Baseline by Up to 45 Points

Oct 14, 2026·3 min
ReaDy speeds up hybrid LLM decoding 3.16x on a single B200
Research

ReaDy speeds up hybrid LLM decoding 3.16x on a single B200

Oct 14, 2026·2 min
DAGFormer Plans Connections Between Layers and Cuts Perplexity by 17–25%
Research·Hugging Face Daily Papers

DAGFormer Plans Connections Between Layers and Cuts Perplexity by 17–25%

DAGFormer adds direct cross-layer connections to the transformer and plans them with a topology predictor. On models up to 1B, WikiText-2 perplexity drops by 17–25%.

Oct 14, 2026·2 min
Researchers Find a 0.4–1.1% Window in Pythia Pretraining Where the Readout Reorganizes
Research·Hugging Face Daily Papers

Researchers Find a 0.4–1.1% Window in Pythia Pretraining Where the Readout Reorganizes

A model's hidden states encode a task before its output shows it. The gap traces back to the unembedding matrix, which reorganizes during 0.4–1.1% of Pythia's…

Oct 14, 2026·2 min