Dusty Mag

Poco, Xiaomi’s sub-brand, is officially launching its first smartphone with the ‘Ultra’ branding, the Poco F7 Ultra. The release comes a week after the company announced the rollout date.

The Chinese brand is widely known for producing affordable yet powerful devices. However, its F lineup is where the overall performance and price tag are near the flagship territory. Accordingly, the new F7 Ultra still honors this essential quality. More specifically, the latest offering from Poco is technically considered a mid-range smartphone, but it is equipped with flagship-level features.

Premium Specs Meets Affordability

The all-new Poco F7 Ultra includes the powerful Qualcomm Snapdragon 8 Elite chipset, which is similar to the high-end Samsung Galaxy S25. Additionally, it is bundled with up to 16 GB of RAM and 512 GB of storage, ensuring top-tier performance for every user. What makes it more exciting is the inclusion of the VisionBoost D7 Chipset, which enables ultra-clear graphics at high frame rates, while maintaining high energy efficiency.

In line with this, the handset features a whopping 6.67-inch OLED display with a 1440p resolution and 120Hz max refresh rate. On top of this, the F7 Ultra pioneers the ultrasonic fingerprint sensor below the screen for better security. Speaking of protection, the phone boasts an IP68 rating for its native Shield Glass.

Users can also snap like a pro with Poco’s “best imaging system ever.” It has a 50-megapixel main camera, 50 MP 2.5x telephoto camera, and 32 MP ultrawide camera. On the other hand, it includes a 32 MP front-facing camera for selfies.

Moreover, 120W wired and 50W wireless charging supports the 5,300mAh battery. However, wireless chargers are sold separately.

As always, the Poco device comes in a popping colorway, this time in a vibrant yellow finish. However, there is also a black version for those users who prefer a low-key or more subtle style.

The Poco F7 Ultra is now available at £649 (around $835) in the United Kingdom and Europe.

Google officially introduced Gemini 2.5, an update for its Artificial Intelligence (AI) models. It follows Gemini 2.0 Flash Thinking, which was the tech giant’s first thinking model.

Yesterday, the company presented its first 2.5 release, which it considers an experimental version of 2.5 Pro. In line with this, it claims that the new thinking model is the most intelligent among its AI models.

More concretely, the Gemini 2.5 models can reason through their thoughts before sending out a response. Subsequently, it enables improved performance and better accuracy.

Prior to the announcement, Google has been exploring ways to create smarter AI. Similarly, the company wanted to make the models more capable of reasoning using reinforcement learning and chain-of-thought prompting techniques.

“In the field of AI, a system’s capacity for “reasoning” refers to more than just classification and prediction,” states Google Deepmind CTO Koray Kavukcuoglu.

Accordingly, the corporation’s newest model reached a higher performance level by integrating improved post-training with an enhanced base model.

The Gemini 2.5 Pro

Based on a Google article, the Gemini 2.5 Pro Experimental is the most advanced model designed to accomplish complex tasks. Specifically, it has enhanced reasoning and advanced coding.

The company also claims that its latest thinking model beats its competitions in the field, including OpenAI, Anthropic, xAI, and DeepSeek. It highlighted that it ranks first by a notable margin in the LMArena leaderboard across common AI capabilities.

“Gemini 2.5 Pro is an awesome state-of-the-art model, no.1 on LMArena by a whopping +39 ELO points, with significant improvements across the board in multimodal reasoning, coding & STEM,” Google DeepMind CEO Demis Hassabis said in a post on X.

Google says it will integrate these thinking capabilities into all models to boost overall performance as it moves forward.  

As of now, Gemini 2.5 Pro is only available in Google AI Studio, while Gemini Advanced users can access it through the Gemini app.

Microsoft is unveiling several new agentic solutions powered by Artificial Intelligence (AI) to expand Security Copilot. The latest additions elevate the safety tool’s capability of detecting, investigating, and responding to security incidents to empower defenders.

More specifically, the AI agents can autonomously assist with growing cyber concerns, including data breaches, phishing, and more. It is in response to the company’s findings of damaging cyber threat cases from last year. Accordingly, the Windows maker asserted that these AI agents are crucial for modern security to keep up with complex cyberattacks.

“Scaling cyber defenses through AI agents is now an imperative to keep pace with this threat landscape,” Microsoft Security Corporate Vice President Vasu Jakkal says in an official article.

Six AI Agents from Microsoft Security

According to Microsoft, the six agentic solutions can independently manage multiple security and IT tasks and seamlessly integrate with Microsoft Security solutions. The agents learn from feedback, operates safely, and adjusts to workflows.

Furthermore, the Security Copilot agents are designed for the Phishing Triage Agent in Microsoft Defender, Alert Triage Agents in Microsoft Purview, Conditional Access Optimization Agent in Microsoft Entra, Vulnerability Remediation Agent in Microsoft Intune, and Threat Intelligence Briefing Agent in Security Copilot.

Five AI Agents from Microsoft Security Partners

The software corporation worked with several partners, including OneTrust, Aviatrix, BlueVoyant, Tanium, and Fletch. Together, they aim to empower a security ecosystem with an open platform wherein they can build to better serve customers.

Among the agentic solutions from Microsoft’s partners are the Privacy Breach Response Agent, Network Supervisor Agent, SecOps Tooling Agent, Alert Triage Agent, and Task Optimizer Agent. All of these will be available in Security Copilot.  

Based on an article by Microsoft, the tech giant plans to continue an agile innovation in the Security portfolio to build a safer world for all.

The AI agentic capabilities will be available for preview in April 2025.