agent Archives

Tech Optimizer

September 1, 2026

Connecting My LangGraph AI Agent to Postgres

A stateful LangGraph agent was developed to streamline a 15-minute booking process, featuring a user-friendly Streamlit UI and a backend powered by a Postgres database. The agent responds to customer queries, calculates service pricing, manages service acceptance, suggests time slots, and confirms appointment details. It operates in two modes: in-memory for rapid testing and Postgres for persistent storage. Testing can be done locally using Docker, which simulates a Postgres environment, or with a hosted Postgres instance. The project includes a docker-compose.yml file to initiate a PostgreSQL 16 container, allowing the application to connect to the database. Data persistence is ensured through Docker volumes, and the application can be run with specific commands after setting up the environment variables. The system has been tested successfully, demonstrating its ability to manage bookings and maintain data integrity across sessions. Future enhancements are planned to improve the booking workflow and integrate additional communication channels.

Tech Optimizer

September 1, 2026

Autoscaling Lakebase Postgres

Choosing a database instance size without prior knowledge of the workload can lead to inefficiencies and excessive compute usage. Lakebase Postgres addresses this with an autoscaling feature that eliminates manual sizing, utilizing in-place VM resizing and a monitoring algorithm for CPU, memory, and working set size. Lakebase Postgres separates compute and storage layers, allowing independent resizing of compute nodes without affecting the database. The autoscaling algorithm relies on three signals: CPU load (cpuGoalCU), memory use (memGoalCU), and compute-cache working set size (lfcGoalCU). The CPU load is monitored every five seconds, aiming to maintain it at or below 90% capacity. Memory usage is tracked at two frequencies: overall memory every five seconds and Postgres-specific memory every 100 milliseconds, with a goal to keep usage below 75% of allocated RAM. The compute cache evaluates active data access efficiency, adjusting size based on workload. The working set is estimated using a modified HyperLogLog algorithm that records timestamps for page accesses, allowing for distinct page estimates over various time frames. The algorithm projects future working-set growth to allocate sufficient cache while capping it at 75% of RAM. Resizing the compute involves four components: the autoscaler-agent, vm-monitor, Kubernetes scheduler, and NeonVM. Scaling up occurs when any of the three goals indicate a need for more resources, while scaling down includes verification to ensure sufficient memory remains for operations. Timely adjustments in both directions are prioritized to minimize costs.

Tech Optimizer

September 1, 2026

Fake Microsoft Security Scan Tells You to Remove Antivirus—Then Scammers Ask for Remote Access

A new web-based scam is using counterfeit Microsoft-branded security scans to deceive users into uninstalling their antivirus software. The scam websites conduct fake inspections, reporting false security failures and claiming third-party antivirus products are unsupported by Windows. They gather browser information to create customized-looking scan reports and aim to mislead victims into a fraudulent refund process. Investigators have identified 11 related scam sites under the name SysScan, which present themselves as legitimate Microsoft security checks. The scam does not involve file downloads but relies on convincing websites, fabricated security scores, and personal information forms. The scan results are hard-coded, and the security score is limited to a range between 13 and 30 out of 100. The scams instruct users to uninstall antivirus software, which is misleading since Windows does not cease support for third-party antivirus solutions. The scam sites request extensive personal information, including names, addresses, phone numbers, and banking details, and redirect victims to a page claiming a refund manager will call shortly. This process allows scammers to gain remote access to victims' computers. Indicators of compromise include specific IP addresses and domains associated with the scam.

Winsage

August 30, 2026

In a rare move, Google is bringing native Windows 11 and WSL support to its new AI tools

The Windows Subsystem for Linux (WSL) is gaining popularity among developers, with GitHub introducing experimental support for WSL in its Copilot desktop application, allowing Linux environments to be used directly within Windows. Google is also integrating WSL support into its Antigravity development platform, marking a shift in its approach to Windows support. Rody Davis from Google confirmed that both WSL and native Windows support are being developed for Antigravity, which will enable file operations and shell commands in a Linux environment. GitHub Copilot now allows users to create sessions inside WSL, enhancing the development experience by enabling real-time interaction with projects in a WSL environment. Additionally, Microsoft is focusing on improving Windows for developers, with initiatives like Coreutils for Windows and the open-sourcing of WinUI, while Ubuntu is reportedly gaining traction on Windows 11.

Tech Optimizer

August 29, 2026

Overcoming the Barriers of Commercial DBs with Open Source and AI… EDB Hosts ‘PostgreSQL AI Summit’

A company's approach to artificial intelligence (AI) begins with determining the optimal structure for data storage and management, often facing challenges such as high costs of commercial databases and reliance on specific enterprise technologies. To effectively utilize AI, companies need to establish new infrastructures, including vector search capabilities, Search Augmentation Generative (RAG) systems, and data lakehouses. EDB is notable for connecting companies and technology partners that have transitioned to open-source databases. On September 3, EDB will host the 'EDB Postgres AI Summit Seoul 2026' at the Sofitel Ambassador Seoul, starting at 10:00 AM. This event is the largest PostgreSQL conference in Korea, themed 'Change the Game,' focusing on the shift from commercial databases to open-source and AI-driven frameworks, featuring real-world case studies. Approximately 300 C-level executives and IT decision-makers from various sectors are expected to attend, with participation by invitation only. The summit will include 15 sessions with customer case studies and technical presentations. Notable discussions will include IBK Industrial Bank of Korea's migration of 15 core systems to PostgreSQL, a semiconductor company's diversification of MPP databases on a DBaaS platform, and Kyobo Book Centre's database modernization strategy. Shopcast will present its development of an 'Agentic Lakehouse' integrating AI technology with a data lakehouse framework. The keynote address will be given by Kim Deok-joong, discussing organizational management strategies for integrating AI agents. The technical sessions will cover the architecture of the 'EDB Postgres AI' platform, which supports AI vector search and RAG, with live demonstrations of the analytics engine ClickHouse and LakeHouse technology for analyzing petabyte-scale data. EDB's domestic distributors and international partners will participate as sponsors. The summit aims to showcase technologies and case studies from the domestic ecosystem, addressing PostgreSQL adoption, system migration, operations, data analysis, and AI implementation. EDB manages transaction, analytics, and AI workloads using Postgres in cloud environments, serving over 1,500 global customers. Herve Timsit, EDB's Chief Revenue Officer, emphasized the event's focus on sharing tangible results and addressing the challenges of commercial databases while investing in AI infrastructure.

Tech Optimizer

August 27, 2026

Scammers Impersonate Microsoft to Push Fake Security Scans and Refund Fraud

A network of fraudulent websites, branded as SysScan, has been discovered, which falsely claims to evaluate antivirus software effectiveness through deceptive security scans. These sites manipulate users into uninstalling legitimate antivirus products and disclosing sensitive personal and banking information. Eleven distinct domains associated with SysScan have been identified, all hosted on a single server. The fraudulent scans generate misleading results based on static findings rather than actual system assessments, and users are coerced into believing their computers are at risk. The scams misrepresent normal browser behaviors as security threats and instruct victims to uninstall their antivirus software, compromising their defenses. The operation targets both individual and business users, collecting extensive personal information and utilizing remote-access tools. The data submitted is sent to Telegram via its bot API. Users are advised to disconnect from the internet and secure their devices if they suspect they have been compromised. Indicators of compromise include specific IP addresses and domains associated with the scam.

AppWizard

August 27, 2026

Intelligent transcription with Gemini 3.5 Transcribe

Gemini 3.5 Transcribe is a new speech-to-text model designed for intelligent voice interactions, overcoming challenges faced by traditional speech recognition systems, such as background noise and complex jargon. It transforms raw audio into accurate, formatted text and is available to users of the Gemini app and Android devices. Developers can access its capabilities through the Gemini API in Google AI Studio and the Gemini Enterprise Agent Platform. The model supports two APIs: real-time streaming for interactive applications with sub-second latency and pre-recorded audio processing that includes speaker attribution and word-level timestamps. It enhances transcription accuracy by managing self-corrections, eliminating filler words, and auto-formatting text. The model has a Word Error Rate of 4.0% for streaming and 2.6% for non-streaming scenarios, effectively capturing alphanumeric entities. It adapts to custom vocabulary, supports over 85 languages, and can identify multiple speakers in pre-recorded audio.

Winsage

August 27, 2026

New Windows malware lays dormant until a custom command activates it like a sleeper agent

Researcher Dominik Reichel has identified a new malware implant called SLEEPWALKER, which is disguised as an agent from ESET. SLEEPWALKER is unique because it does not contain malicious code and remains dormant until it receives specific network signals to activate. It appears to be designed for targeted attacks, likely orchestrated by nation-states. Although it was submitted to VirusTotal last year, it has not been linked to any active campaigns or confirmed victims. The origins of SLEEPWALKER are unknown, and its code is described as somewhat "rough around the edges," indicating it may still be in development.

AppWizard

August 26, 2026

Multilogin releases open-source agent so AI can run social accounts on real Android cloud phones

Multilogin has launched an open-source AI agent for social media management on authentic Android devices in the cloud. This tool enables automation of workflows on platforms like TikTok, Instagram, Facebook, and X. The AI agent connects to a Multilogin Android cloud phone, allowing it to view screens, execute taps, type commands, and save results. It integrates with multiple AI agents capable of running shell commands and interpreting visual inputs. Each cloud phone functions as an independent Android device with unique parameters and persistent app data, connected via ADB. The tool monitors the cloud phone's readiness, activates ADB, interprets screenshots, and executes tasks while logging activities. It is designed for marketers, agencies, creators, and teams, providing setup instructions and compatibility with AI tools like ChatGPT/Codex and Claude Code. The project is available on GitHub, and Multilogin offers a free plan with paid options starting at [openai_gpt model="gpt-4o-mini" prompt="Summarize the content and extract only the fact described in the text bellow. The summary shall NOT include a title, introduction and conclusion. Text: Multilogin has unveiled an innovative open-source AI agent designed to empower social media management on authentic Android devices hosted in the cloud. This groundbreaking tool bridges the gap between AI capabilities and mobile app functionalities, allowing users to automate workflows across popular platforms such as TikTok, Instagram, Facebook, and X. How the Tool Functions The newly developed AI agent connects seamlessly to a Multilogin Android cloud phone, providing it with the ability to view screens, execute taps, type commands, and complete various tasks while saving the results. This integration works harmoniously with multiple AI agents, including Claude Code, Cursor, and OpenClaw, all of which can run shell commands and interpret visual inputs from screenshots. As many social media operations increasingly shift to mobile applications, the limitations of browser-based tools become apparent. While browser profiles serve their purpose for web-based tasks, they fall short when it comes to executing native Android applications. Multilogin's cloud phones offer a distinct advantage by providing an authentic Android environment, facilitating direct interaction with mobile apps. Each cloud phone operates as an independent Android device in the cloud, complete with unique device parameters, installed applications, and persistent app data. The AI agent connects to the cloud phone via ADB (Android Debug Bridge), a standard interface used by developers for device management. Operational Efficiency and Flexibility The process begins with the selection of an Android cloud phone, which the tool monitors until it is fully operational. Once ready, the AI agent activates ADB, allowing it to interpret screenshots, determine subsequent actions, and execute tasks such as tapping or typing. Upon completion of a task or in the event of an error, the tool halts the cloud phone and logs the activity, ensuring that resources are not wasted during the development and testing of AI-assisted mobile workflows. This project is compatible with any AI agent capable of executing shell commands and processing visual inputs, making it versatile for various applications. Gleb K., the UI team lead at Multilogin, emphasizes the importance of this development: “AI agents are becoming useful for real operational work, but many important workflows still happen inside mobile apps. We built this open-source project to give everyone a simple and controlled way to connect AI agents to Android cloud phones.” Target Audience and Accessibility The cloud phone agent is tailored for marketers, agencies, creators, and teams seeking an AI assistant to streamline routine tasks within Android applications. Users need not start from scratch; the project provides clear setup instructions and is compatible with widely-used AI tools such as ChatGPT/Codex, Claude Code, OpenClaw, and Cursor. Once connected, the AI agent can efficiently manage tasks on the cloud phone, with the Multilogin Support team readily available for assistance. As with any automation tool, users are encouraged to adhere to the guidelines set by the applications they utilize. The Multilogin cloud phone agent is accessible on GitHub: GitHub Repository. For further information about Android cloud phones from Multilogin, visit Multilogin's Website. The Need for Android Cloud Phones in Social Media Workflows As marketing teams increasingly adopt AI-driven workflows, the necessity for reliable and scalable environments becomes paramount. Real Android devices provide AI agents with the authentic conditions that mimic human usage, allowing accounts to behave naturally and accumulate history over time—an aspect that emulators and browser profiles simply cannot replicate. About Multilogin Multilogin is a comprehensive platform offering cloud phones and browser profiles tailored for marketers, agencies, creators, and AI workflow builders. It enables teams to manage distinct browser and mobile workflows from a unified dashboard. Multilogin offers a Free plan with no credit card requirement, while paid plans commence at .08 per month when billed annually, inclusive of monthly proxy traffic and mobile minutes based on the selected plan. Established in 2015, Multilogin operates from the UAE and adheres to GDPR compliance. Media Contact: Teona B. Media Specialist Multilogin Email: marketing@multilogin.com For additional visuals accompanying this announcement, please visit the following links: Visual 1 Visual 2" max_tokens="3500" temperature="0.3" top_p="1.0" best_of="1" presence_penalty="0.1" frequency_penalty="frequency_penalty"].08 per month.

Tech Optimizer

August 26, 2026

How to Protect Against Ransomware: 13 Steps [2026]

Global ransomware attacks increased by 0.8% quarter-over-quarter in Q2 2026. The group Qilin has 279 confirmed victims, maintaining its position as the leading threat actor for four consecutive quarters. The rival group The Gentlemen saw a 62% increase in victims, reaching 269. Checkpoint Research's report indicates a shift in the ransomware landscape, challenging the narrative of declining attacks. IBM's analysis shows breaches within supply chains and third-party vendors have quadrupled over the past five years. Effective ransomware protection requires administrative access to endpoints, firewalls, and backup systems, along with specific software versions and tools. The 3-2-1-1-0 backup strategy is recommended, emphasizing offline, immutable backups. Organizations should implement multi-factor authentication (MFA) across all access points and actively patch known vulnerabilities. Network segmentation is crucial to limit ransomware spread, and canary files can provide early warnings of attacks. Regular phishing simulations and a well-documented incident response plan are essential for preparedness.