Brief

A developer successfully trained a personal voice adapter using DoRA on the Qwen3-8B model for just $1.50. The process involved using 6,128 personal Telegram messages to fine-tune the model, resulting in an adapter that outperformed the base Qwen3-8B model in blind A/B testing. This method also demonstrated no significant degradation in general knowledge tasks and produced a voice that was perceived as more representative of the individual than their own actual writing. AI

A developer has created an autonomous AI agent named Threlium, which can be controlled via email or Telegram messages. This agent is designed to perform multi-step reasoning, maintain long-term memory, and execute commands. Notably, Threlium is capable of self-modification and is built using systemd and LLMs within a self-hosted environment. AI

The FBI has issued a warning about a new AI-powered attack called Kali365 that circumvents multi-factor authentication for Microsoft 365 accounts. This phishing-as-a-service platform, distributed via Telegram and email, lowers the technical barrier for attackers to steal access tokens. Once compromised, attackers can access services like Outlook and Teams without needing passwords or MFA. AI

A developer created an AI agent fleet manager to automate tasks and reduce screen time. The system, initially a Telegram bridge, evolved into a multi-agent orchestration platform. The project involved significant architectural decisions and lessons learned regarding naming conventions and overall system design. AI

A user has developed a novel approach to fitness tracking by integrating Telegram, Anthropic's Claude AI, and markdown files. This system aims to provide a personalized fitness tracking experience at a low cost, ranging from $1 to $3 per month. The user explores the potential of replacing traditional fitness applications with this AI-powered chat-based solution. AI

The Hermes Agent, an open-source AI agent, demonstrated significant self-improvement over a seven-day period by refining its task execution without manual intervention. Initially producing a basic 12-line skill file and suboptimal results, the agent evolved to generate a more sophisticated 60-line file and improved output quality. This continuous learning capability, where the agent retains and builds upon its past performance, distinguishes it from other frameworks that reset with each session. AI

A developer has created a 200-line open-source toolkit to address three minor issues encountered when running Claude Code via Telegram on Windows. The toolkit resolves a visual annoyance of multiple command windows appearing on login by using VBScript to hide the console windows. It also fixes a problem where the Telegram polling mechanism would stop receiving messages by implementing a script to kill orphaned Telegram processes before starting a new session. Finally, it prevents a scenario where running multiple Claude Code instances simultaneously could lead to a zombie process issue. AI

The author developed a content syndication engine called Autoposting Pro to streamline publishing across multiple platforms. This tool aims to solve the problem of context switching and lost productivity when manually cross-posting content from one service to another. The engine is designed to handle posting to at least 20 different platforms, simplifying the workflow for content creators. AI

Researchers have developed a new framework to detect synthetic political narratives disseminated across social media platforms. This framework utilizes four key coordination signals: lexical diversity, temporal burstiness, rhetorical repetition, and semantic homogenization. By combining these metrics into a Synthetic Narrative Coordination Score (SNC), the system can identify coordinated campaigns, as demonstrated by its application to data from Telegram and Reddit concerning geopolitical events. AI

A couple, Marwan Elwaraki and Salwa Al-Alami, developed a diabetes management app called Sweet Dreams. Marwan, an iOS developer, used Replit's integrated AI tools to rapidly build an MVP and then a full iOS application, significantly accelerating his development process. Salwa, a product manager, focused on ensuring the app met user needs for collaborative health monitoring. The app aims to ease the burden of diabetes management for patients and their families, with Replit serving as a core development partner. AI