APhone Launches Decentralized App Store AppNest

Yonsei University has embarked on a groundbreaking initiative by integrating AWS Trainium with Theta EdgeCloud to enhance its AI agent research. This collaboration signifies a pivotal moment as Theta Network becomes the first blockchain to deploy Amazon's advanced AI chips. The Data & Language Intelligence Lab, led by Professor Dongha Lee, aims to scale AI research while optimizing performance and reducing costs. The use of AWS Trainium allows the lab to leverage a decentralized infrastructure for high-performance deep learning, providing specialized hardware that enables training AI models at significantly lower costs, thus enhancing efficiency and reproducibility in large-scale experiments. The innovative research at Yonsei focuses on developing conversational recommendation agents that simulate human-like interactions. Instead of traditional human evaluators, the lab employs AI-simulated users with distinct memory and personality traits to assess models in real-time. This approach utilizes Direct Preference Optimization (DPO) for model training, which allows agents to refine their responses without manual labeling. Consequently, the research team can simulate millions of user interactions daily, expediting the evaluation and improvement of their AI models, leading to faster iterations and increased accuracy. The partnership between Theta Network and AWS not only provides a cost-effective solution for AI research but also enhances scalability. With AWS Trainium instances designed specifically for deep learning tasks, institutions like Yonsei can experiment with large models and extensive datasets without incurring significant financial burdens. The integration of Trainium with Theta EdgeCloud's extensive network of over 30,000 NVIDIA GPUs offers researchers the flexibility to select optimal computing resources for their workloads. This collaboration marks a new era in AI research, showcasing how decentralized infrastructure and advanced AI hardware can revolutionize academic research and development in the field of artificial intelligence.

As esports continues to explode globally, one issue keeps holding players and institutions back: outdated hardware. Most schools and gaming programs can’t afford frequent upgrades, and individual players often face steep costs just to stay competitive. In an industry where performance is everything, this creates a serious gap. DeepLink offers a game-changing solution. By delivering high-performance cloud gaming powered by decentralized GPU infrastructure, DeepLink makes AAA gameplay accessible from almost any device—no upgrades required. Whether you're a student, pro team, or solo gamer, you can now train and compete using the latest GPU power, directly from the cloud. No lag. No setup. No limits. Learn more: https://www.deeplink.cloud/blogInfo/cloud-gaming-esports-hardware-solution

Oblong Inc. has made a significant move by investing $8 million into Bittensor’s TAO token, fully staking it with a main validator to bolster network operations. This strategic pivot, executed by the company traded on Nasdaq under the ticker OBLG, marks a decisive shift in its treasury strategy towards decentralized AI assets. By aligning its financial resources with distributed AI ecosystems, Oblong aims to enhance its long-term asset growth and innovate within the blockchain space. This transition follows a successful $7.5 million capital raise, which was directed towards acquiring digital assets, positioning Oblong as a forward-thinking player in the crypto market. The complete staking of TAO signifies Oblong's departure from traditional cash management practices, as it embraces a blockchain-based asset portfolio. With no outstanding debt and $4.3 million in cash at the end of Q1 2025, the company is committed to balancing innovation with financial performance. Oblong plans to assess new TAO purchases on a weekly basis, indicating a sustained engagement with blockchain-driven AI infrastructure. Furthermore, the firm will disclose staking performance and token data in its Q2 2025 earnings report, which is expected to bolster shareholder confidence as it expands its digital asset footprint. TAO serves as the backbone of a decentralized AI marketplace, rewarding contributions through a permissionless incentive system. By staking TAO, Oblong not only increases its exposure to staking rewards but also actively supports decentralized computing systems. This strategic alignment with Bittensor’s architecture allows Oblong to facilitate open AI experimentation without centralized constraints. As the company explores potential partnerships within the Bittensor ecosystem, it aims to develop complementary technologies that reinforce its blockchain-first vision. With a market capitalization of approximately $6.6 million, Oblong is poised to redefine its operational outlook by integrating crypto-based assets into its core business strategy.

In Q2 2025, GEODNET reported a remarkable 19% quarter-over-quarter revenue increase, rising from $812,000 to over $963,000, driven by the global expansion of its Real-Time Kinematics (RTK) station network. With annualized revenue approaching $3 million, GEODNET stands out in the Decentralized Physical Infrastructure Network (DePIN) sector, surpassing many of its competitors. The network now boasts over 19,000 active Satellite Miners across 145 countries, solidifying its position as the largest RTK network globally. Notably, GEOD token burns reached an all-time high of 3.8 million in Q2 2025, reflecting a 67% increase from the previous quarter, which further supports the network's long-term sustainability and value capture initiatives. GEODNET operates on a decentralized model where individual operators set up RTK Base Stations, known as Satellite Miners, that capture satellite signals and provide real-time correction data with centimeter-level accuracy. The network's growth is attributed to strategic partnerships with companies like DroneDeploy and Quectel, enhancing its service offerings in enterprise drone mapping and IoT solutions. Additionally, GEODNET has implemented a 50% reduction in daily token commissions to mitigate inflation and support its contributors. The network's performance-based incentives introduced under the GIP-6 governance proposal are designed to improve data reliability while maintaining token supply discipline. Looking ahead, GEODNET is poised for further expansion and technological advancements. Plans include integrating Low Earth Orbit (LEO) satellite tracking to enhance GNSS reliability, fusing RTK data with Inertial Measurement Units (IMUs) and LiDAR for precise positioning in challenging environments, and developing a custom RISC-V-based robotics chip for real-time sensor fusion. These initiatives will not only strengthen GEODNET's infrastructure but also position it as a leader in precision geolocation and decentralized systems for autonomous applications, ensuring continued growth and innovation in the coming years.

July marked a significant milestone for Nubila, showcasing impressive growth and advancements in decentralized climate intelligence. The company successfully mined 73,734 B3TR across five mining rounds, with validator participation increasing and new partnerships being established. Notably, 415 Marco devices have been deployed globally, including 209 Cloud Nodes, 33 Rainy Nodes, and 109 Sunny Nodes. This expansion is crucial as Nubila builds the backbone of climate-aware AI, integrating real-time weather data into its ecosystem. The Nubila ecosystem has been active with various campaigns and initiatives. The ongoing Validation Campaign allows participants to earn rewards by reviewing real-time weather data, with the first 100 daily validators receiving B3TR. Additionally, the Nubila Global Tour has launched in Canada, offering opportunities to win Cloud Nodes and earn sNUBI. The Snap to Earn initiative, in collaboration with Nillion Network, encouraged users to share local weather images, resulting in hundreds of submissions and exciting prizes. Furthermore, the July rebate drop successfully redistributed 106,015 B3TR to Marco and Node owners, reinforcing the financial benefits of participating in climate infrastructure. Nubila has also formed strategic partnerships to enhance its offerings. Collaborations with Spheron, UX Link, NodeOps, and ChainAware.ai aim to provide cheaper decentralized GPU compute, unlock WeatherFi in Asia, and integrate predictive AI with weather data. As Nubila continues to innovate, it has gained recognition, winning the IoTeX Crypto’s Got Talent Season 1 and being featured in Cointelegraph. Looking ahead, the Validator Node Sale on August 6 and the launch of Snap-to-Earn Phase 2 are set to further propel Nubila's mission of connecting real-world environmental data to decentralized applications and AI, driving the climate economy forward.

In June 2025, IoTeX launched a beta interface for Quicksilver, a decentralized physical infrastructure network (DePIN) framework that allows AI agents to interact with data from connected devices. This interface enables developers to create and deploy AI agents, while users can query these agents through chat or API. On June 19, IoTeX implemented a significant upgrade (v2.2.0) that reduced block time from 5 seconds to 2.5 seconds, enhancing the network's theoretical throughput to 2,000 transactions per second (TPS). Additionally, IoTeX introduced a roadmap for Realms, which aims to create domain-specific ecosystems that group AI agents and contributors to facilitate structured intelligence across various sectors such as transportation and healthcare. Despite a decline in average daily active wallets by 26% quarter-over-quarter (QoQ) to 6,600, this figure still represents a remarkable 872% increase year-over-year (YoY). Similarly, average daily transactions dropped by 25% QoQ to 222,300 but were up 750% YoY. The protocol's revenue also saw a significant decrease of 53% QoQ, totaling $94,800, primarily due to a 93% drop in ioID activation fees. However, the overall revenue remains up 505% YoY, indicating a robust growth trajectory despite recent fluctuations in activity. The Quicksilver framework is designed to integrate AI agents with DePINs, enabling them to process real-world data and perform automated actions. The W3bstream layer verifies data from devices before it is utilized by AI agents. IoTeX's ioID framework provides a decentralized identity for devices, enhancing their interaction capabilities within the network. As IoTeX continues to evolve, it aims to solidify its position in the AI and DePIN sectors, with the Realms system expected to incentivize data contributors and enhance the quality of datasets for AI applications. This strategic direction highlights IoTeX's commitment to building a comprehensive ecosystem that leverages physical intelligence for various applications.