Yandex Drops Earbuds Under X-Ray Microtomography: Detailed Internal Analysis of First AI-Powered TWS Headphones
Yandex marked its birthday by allowing engineers to run its first AI earbuds, Yandex Drops, through a high-resolution X-ray microtomograph. The non-destructive scan produced two voxel datasets—one of the charging case with an earbud inside and one of a standalone earbud—capturing internal structures at 24.44 µm and 14.12 µm resolution respectively.
The earbuds contain an eight-layer HDI PCB with blind vias, a QFN-packaged central SoC approximately 2.8 × 2.0 mm that is presumed to house the NPU for always-on voice detection, and a smaller companion QFN die likely serving as memory or PMIC. A 4 × 4 BGA flash-memory candidate sits adjacent to the main processor.
Each earbud integrates three microphones: two external feedforward sensors positioned at the stem ports and one internal feedback microphone located roughly 5 mm from the driver diaphragm. Signal routing was traced from the MEMS elements through spring connectors and flex cables to the main board without any visible hardware switch or disconnect in the accessible layers.
The 11 mm dynamic driver and 46 mAh pouch cell were measured directly from the calibrated slices, matching the manufacturer’s specifications. The charging case houses a 520 mAh cell, USB-C receptacle, and a three-pin Hall-effect sensor on a flex tail that detects lid position; no shielded radio module was observed inside the case.
Capacitive touch electrodes are formed by stamped metalization on the inner surface of the stem, while a miniature SMD tactile switch on a daughter board provides physical button input. Charging contacts on the earbuds are spring-loaded claws that mate with corresponding pads inside the case.
All imaging was performed at the PRODIS facility using a 2 µm-resolution microtomograph; reconstruction and multi-volume analysis were carried out in the Russian-developed VolÅn software. The authors note that while the physical microphone traces appear continuous, any determination of what audio is actually transmitted requires separate RF traffic analysis.
Related articles
Russia Sets Tech Collection Fees: Smartphones from 250 Rubles, Laptops from 500 Rubles Starting December 2026
The Russian Ministry of Industry and Trade has proposed specific rates for a new technological collection fee that will apply to imported smartphones, laptops, and landline telephones. Payments begin on December 1, 2026, with the final amount depending on whether the device model is registered in the national marking system. Base rates are reduced by 33 percent for registered models and increased by the same margin for unregistered ones. The fees were calculated using 2025 median customs values and a 4 percent inflation target. Manufacturers and importers must pay the collection, though costs are expected to reach consumers through higher retail prices. Separate lower rates apply to corded phones with wireless handsets and basic mobile phones.
How Social Media Recommendation Algorithms Shape Content, Behavior, and Regulation Worldwide
Social media platforms such as TikTok, Instagram, and YouTube rely on sophisticated recommendation systems that analyze user behavior to predict engagement rather than simply ordering posts chronologically. These algorithms form candidate sets and rank content using signals including watch time, likes, comments, and device data, creating personalized feeds that can diverge significantly between users. Research shows mixed effects on political views and mental health, with personalization capable of both amplifying polarization and reducing it depending on optimization goals. Regulatory responses vary: China mandates transparency and youth mode limits on Douyin, the EU's Digital Services Act requires non-profiled alternatives for large platforms, and Russia enforces disclosure rules under Federal Law No. 408-FZ since 2023. Practical steps allow users to reset recommendations and provide negative feedback to retrain feeds. The article debunks myths of dopamine addiction and inevitable radicalization while highlighting both risks and benefits of algorithmic curation.
Russia Proposes 22% VAT and 100-Ruble Customs Fee on All Foreign Online Purchases
The Russian Ministry of Finance has submitted draft amendments to the Tax Code and the federal budget for 2027-2029 that would impose a standard 22% VAT on goods purchased from foreign online stores and marketplaces. In addition, every postal shipment valued at up to 200 euros would incur a separate customs fee of 100 rubles. Electronic trading platforms are designated as tax agents and are expected to pass the new costs directly to buyers through higher final prices. The proposals eliminate the current duty-free threshold for low-value cross-border orders under EAEU rules and replace an earlier plan for gradual VAT increases with an immediate full-rate application. Officials state the measures are intended to create equal competitive conditions for Russian sellers and to further formalize the economy. The changes have not yet been adopted and must still undergo government and legislative review.
VK Tech Rolls Out Unified AI Assistant Across VK WorkSpace Corporate Tools
VK Tech is preparing a unified AI Assistant for its VK WorkSpace platform, along with semantic search and an MCP server to connect autonomous agents. The features target small and medium businesses as well as large corporations by automating routine tasks in messaging, email, and calendar services. Semantic search will allow users to locate documents, emails, and messages by meaning rather than exact titles or phrasing. Meeting recordings will be automatically transcribed with extraction of key topics, decisions, and action items. The AI Assistant launches first in the Messenger module of the On-Premise version, with cloud SaaS availability scheduled for October and later expansion to Mail, Disk, and Calendar. Access remains strictly limited by user permissions and company policies. An MCP server will also be introduced, enabling AI agents to query data, prepare meeting materials, and perform actions inside the platform services, initially supporting Messenger and extending to other modules by year-end.