Google DeepMind
Google DeepMind is Alphabet's AI division, home of the Gemini model family, the Nobel-recognized AlphaFold and the Veo video generator. Through Google Search, Android and the cloud, its models reach billions of users. This page gathers all our coverage of Google and DeepMind: models, research and products.

Claude Pulls Its Links from Brave: 89 out of 100 Citations Sit in the Top 20 of That Search Engine
A top spot on Yandex does not guarantee a place in ChatGPT and Claude answers: each AI has its own search results provider. We look at where to check your site's…

SIBUR added an AI critic to its multi-agent system and cut idea evaluation from 88 days to two weeks
SIBUR assembled a team of AI agents to assess material-substitution ideas and added a "pedantic" critic to it. We explain why it is needed and what went wrong.

How to Check LLMs by Hand and by Script: 12 Lessons from Translations and Size Charts
Tens of thousands of translations a day and size charts where a 10 mm error means a return. We look at how a team built Human-in-the-Loop and what actually worked.

AI Editors Silently Rewrite Your Text: How to Measure the Share of Edits Before Choosing a Model

Thinking Machines' Inkling 2 Small: 17B Active Parameters and $1.18 per Task

VIALS-AA: Gemini 4 Carbon solves 64% of scientific image tasks, GPT-6.1 Sol is 30 times cheaper

A blogger used Claude Code to dub GTA V into Russian for $60: 250 thousand lines

A Second of AI Video in the Final Cut Costs $0.186, Not the $0.10 on the Price Sheet
The price sheet promises $0.10 per second, but the bill runs 2–3 times higher. A Python take log shows what a second that actually makes the final cut costs.

QVikhr-3-4B on a 1.5K ₽ GPU-less VPS correctly tagged 20 of 20 requests
A Habr author ran nine lightweight LLMs on a GPU-less VPS costing 1.5K ₽ a month. The best model labeled all 20 requests correctly (100%), but long texts take…

NVIDIA predicts the best coding-agent base model before fine-tuning: its ranking matched the final one at 96%

Amharic input costs 6x more than English in GPT-5: 78 tokens vs. 13

Inkling 2 Small scores 31 on the Artificial Analysis index and costs $1.18 per task

Best AI agent meets just 12% of requirements in new GDPx test across 206 tasks

A Product Manager With No Developer Shipped an AI-Agent-Built Extension and Protected It With 89 Guard Tests

Top AI Agents Pass Up to 98% of Tests, but Build a Fully Working Product in at Most 65% of Runs

Agent loses data because of table size, not the number of turns: 92% correct turns vs 46%

LLM Agent Built a Gearbox in KOMPAS-3D, but Scored Only 50 out of 100 on the Specification
The kompas-mcp plugin lets an LLM agent drive KOMPAS-3D: it built a gearbox and its design documentation. In a homemade benchmark, 3D tasks score 88–100; the…

We Audited 929 Automations at Alpina: What Corporate AI Training Should Look Like in 2026
Overview lectures on neural networks yield 8–12% adoption, while a platform, workshop and hackathon combo reaches 70–85%. Pick vendors by task, and demand a…






