agents Archives

Winsage

September 28, 2026

Windows 11 can keep thumbnails of photos you deleted. Here’s how to remove them

A recent post on X claims that Windows retains a thumbnail of every deleted photo in a file named thumbcache_256.db, which has raised privacy concerns among users. While Windows does maintain a thumbnail cache for efficiency, this practice is not unique to Windows, as other operating systems like macOS, Linux, and Android also use similar systems. Users can inspect or clear their thumbnail cache by accessing specific database files in the Windows Explorer directory. The claim that Windows retains "every photo you’ve ever deleted" is misleading, as thumbnails are only created when images are viewed in thumbnail mode. Windows does not cache every file, and thumbnails must exist before they can be cached. Additionally, Windows keeps app launch records in Prefetch files for performance optimization, maintains entries for USB devices in the USBSTOR registry key, retains records of Wi-Fi networks and passwords (which are encrypted), and can keep clipboard history and recent file shortcuts. Caching previews is common across various operating systems, and Microsoft is focused on user consent and privacy as it enhances Windows 11 with AI functionalities.

Winsage

September 27, 2026

Google researcher explains why Windows NT “puts Linux to shame,” and imagines an alternate history where it won

Laurie Kirk, a Google researcher and former Microsoft reverse engineer, argues that the Windows NT kernel is superior to Linux in several aspects, particularly in its architecture for managing resources and access permissions. She emphasizes that her critique focuses on the system architecture rather than the user interface of Windows 11. Kirk raises concerns about Linux's suitability for AI agents due to its complex security model, while praising NT's object-based design and security framework. She speculates on an alternate history where Microsoft had created an "Open NT," allowing for customizable components, but acknowledges the challenges of maintaining forks in operating systems. Critics of NT highlight its performance and reliability issues, while Linux's open-source nature has contributed to its dominance in servers and cloud computing. Kirk suggests that future operating systems may need to balance the defined resource types of NT with the flexibility of Linux, especially as AI technology evolves.

AppWizard

September 25, 2026

Google brings back Android Dev Summit with Android 18 ‘powerful new graphics features’ tease

The Android Dev Summit will take place in 2026 at Google’s headquarters in Mountain View, with a satellite event in London. It will feature a keynote address from Android leadership and will be organized into four tracks: AI Experiences, Android XR, Apps Foundation, and Tools and Performance. Notable sessions include one on enhancing apps with graphics features in Jetpack Compose and Android 18. The conference will also cover themes such as app integration with advanced technologies like Gemini and the Android intelligence system. The event is scheduled for October 28-29, 2026.

AppWizard

September 24, 2026

Ando wants to take on Slack with a team messaging app that lets humans and agents work together

In 2025, Sara Du identified challenges in integrating AI agents within communication platforms like Slack while assisting companies with MCP servers. She criticized the design of these platforms for treating AI agents as separate applications rather than team participants. Du proposed a solution to eliminate intermediaries between AI agents and teams, envisioning a system where agents could engage directly in conversations, understand decisions, ask questions, and autonomously notify humans when necessary. Du's startup, Ando, launched as a messaging platform that allows both human and AI workers to communicate seamlessly, featuring channels, direct messages, and live calls. Ando has raised a million in pre-seed and seed funding from firms like Accel and Index Ventures. Despite existing competition from platforms like Slack and Microsoft, Du sees potential for Ando to disrupt the market by focusing on AI agent integration. Currently, Ando works with clients in software, real estate, and finance across 15 countries, primarily small teams, and aims to expand its capabilities with the new funding.

AppWizard

September 24, 2026

6 ways Android Enterprise is evolving for the modern workforce

In modern workplaces, the balance between innovation and security is crucial as employees seek access to advanced technology while IT departments protect sensitive data. Android is enhancing its tools for workers through Android Enterprise, with key advancements expected by 2026, including: 1. Advanced AI agents and controls: - Smarter workflows will introduce next-generation AI agents with enterprise-grade controls. - The evolution of Gemini will enable management of complex workflows using real-time context, eliminating manual app switching. - Built-in safeguards will ensure work profile isolation, preventing personal agents from accessing corporate data. - Administrators will have granular control to configure or disable AI automation through Android Enterprise. 2. Next-gen manageable XR devices: - New hardware like the XREAL Aura wired glasses will enhance workforce capabilities. - IT administrators will manage a range of XR devices alongside standard mobile fleets using Enterprise Mobility Management consoles. - Employees will benefit from hands-free productivity with integrated displays, audio, and AI assistance while connected to enterprise tools.

AppWizard

September 23, 2026

RatHat Android Malware Uses AI to Target Banking Credentials in Real Time

A newly identified strain of Android malware, RatHat, utilizes generative AI to manipulate infected devices in real time. It is linked to threat actors believed to be operating out of China. RatHat serializes the device’s live Accessibility tree into XML format and communicates with a generative AI assistant to return screen coordinates, identify text, and issue navigation commands. The malware employs WebView-based HTML overlays to capture login credentials from banking and cryptocurrency applications and can infiltrate payment apps like WeChat and Alipay to extract PINs. It also features an SMS receiver and notification listener to intercept OTPs and 2FA codes. RatHat is disseminated through smishing, malvertising campaigns, and misleading third-party forums. It employs anti-analysis techniques such as container tampering, manifest bombing, DEX bytecode poisoning, string encryption, and anti-debugging. Once installed, it gains Accessibility access, activates Developer Options, and enables Wireless Debugging, allowing it to connect to the device's local ADB service and launch control agents with shell-level privileges. The malware captures raw touch coordinates to reconstruct PINs and unlock patterns, bypassing screenshot protections. It also includes persistence mechanisms that prevent uninstallation by presenting a fake Google Play failure overlay and automatically reinstalling itself if removed.

Tech Optimizer

September 20, 2026

How to Remove a Keylogger: 12 Steps, 60 Min [2026]

A keylogger captures every keystroke made on a device and sends that information to an external source. In 2025, infostealer malware, including keyloggers, resulted in the theft of 642.4 million credentials from 13.2 million infections globally. In Australia, approximately 30,000 banking passwords were compromised between 2021 and 2025 due to this malware. The Australian Cyber Security Centre reported over 42,500 calls related to credential theft in one year, a 16% increase. Four main malware families dominate the landscape, with LummaC2 accounting for nearly 60% of credentials linked to infostealers in 2026. Keyloggers can be detected through signs such as typing lag, unexpected antivirus notifications, new browser extensions, and unexplained spikes in outbound data usage. Detection and removal steps include checking running processes, auditing startup programs, reviewing installed programs, running antivirus scans, and resetting passwords from a clean device. Keyloggers can infiltrate systems via phishing emails, malicious attachments, or compromised software downloads. Mobile devices can also be targeted by keyloggers disguised as legitimate apps.

AppWizard

September 19, 2026

Google’s Android Bench 2.0 tests AI agents on week-long app builds

Google has launched Android Bench 2.0, an upgraded benchmark for evaluating AI coding agents, which now includes 30 long-horizon tasks that can take human engineers days or weeks to complete. The best pass rate for these tasks is around 28%, down from approximately 91% in the previous version. The benchmark measures both pass rates and completion rates, acknowledging partial progress rather than just failures. It assesses various challenges, including app creation and feature integration, using a comprehensive scoring methodology that evaluates functionality, regression checks, and visual fidelity. AI models perform better with new code than with modifications to existing systems, facing challenges in runtime validation and cross-platform conversions. The current leaderboard shows OpenAI’s GPT-6 Astra leading with a 28% pass rate, while Gemini 3.8 Flash has an 8% pass rate. Developers using AI agents should be aware that while AI can generate significant portions of applications, further refinements will require human input. The benchmark is available on Google’s Android Bench leaderboard.

AppWizard

September 18, 2026

Android Bench 2.0 focuses on long-horizon tasks, agent evaluations

Google has released version 2.0 of Android Bench, which focuses on managing complex development tasks rather than minor adjustments. The new version evaluates tasks that may take engineers days or weeks to complete, such as adding features and building applications. A continuous scoring method has replaced the previous pass or fail system, assessing completion rates based on functionality, visual fidelity, and adherence to instructions. Various AI models have been tested, with GPT-6 Astra achieving a 28% pass rate, significantly lower than the previous scores around 90%. No model has achieved a 100% pass rate in porting cross-platform applications, with the best reaching 80%. AI performs better in writing new code than in refactoring existing code, facing challenges with architectural complexity and runtime validation. Android Bench utilized agents from model providers for evaluations and plans to incorporate various model combinations in future updates.

AppWizard

September 18, 2026

Google just put the latest AI models through a brutal coding test

Google has launched Android Bench 2.0, a benchmark for evaluating large language models (LLMs) and AI agents on complex Android development tasks. This version focuses on long-horizon tasks (LHTs) that are more intricate than those assessed by the original benchmark. Key tasks include upgrading dependencies, adding major new features, and building Android apps from scratch. The new grading system uses continuous scoring instead of a binary pass-or-fail method, providing a more detailed performance evaluation. Currently, GPT-6 Astra leads the leaderboard with a 28% pass rate, followed by Gemini 3.8 Flash at 8%. Other evaluated models include Claude Fable 5.1, GPT-5.6 Sol, and Claude Opus 5. Google plans to expand the leaderboard with more models and results.