13 August 2025
Topic: Application
The end of perimeter defense: When your own AI tools become the threat actor
Russia's APT28 tested LLM-powered malware on Ukraine. The same tech that breaches enterprises is now selling for $250/month on the dark web....
TL;DR
Russia's APT28 group is leveraging LLM-powered malware, and underground platforms offer the same capabilities for $250 per month, posing a significant threat to enterprises.
Key Takeaways:
Key Takeaways:
- Approximately 270 stolen Hugging Face API tokens are used to power nation-state attacks via LLM-powered malware.
- Underground platforms offer unrestricted AI capabilities for $250 per month, creating a malware-as-a-service economy.
- Enterprises deploying AI tools at unprecedented speed are facing attacks that use novel tradecraft, highlighting a troubling gap in security readiness among AI companies.
Google adds limited chat personalization to Gemini, trails Anthropic and OpenAI in memory features
Google updated the Gemini app running of Gemini 2.5 Pro to reference all historical chats and offer new temporary chats....
TL;DR
Google upgrades Gemini app with personalization and data control features, including Temporary Chat and the ability to restrict data usage in future model training.
Key Takeaways:
Key Takeaways:
- Google's Gemini app now features Personal Context, which enables the model to learn from past conversations and provide tailored responses.
- The app also includes Temporary Chat, allowing users to have one-off conversations that won't influence future ones.
- Additional data controls allow users to restrict their data usage in future Google model training.
Pocket FM gives its writers an AI tool to transform narratives, write cliffhangers, and more
Pocket FM is building LLM to help users write faster and in turn, produce more audio shows....
TL;DR
Pocket FM, an India-based audio series platform, is leveraging AI to rapidly generate content, expand globally, and explore new formats like video and comics, with plans to release its own large language model next year.
Key Takeaways:
Key Takeaways:
- AI has helped Pocket FM increase writer productivity by up to 50% in the German market and reduce content production costs by 2-3 times in the U.S.
- The company plans to release its own singular large language model (LLM) by next year, which will be based on data collected from its shows and incorporate various tools like writing assistance and adoption.
- While AI has had its side effects, such as layoffs and concerns around quality and user adoption, Pocket FM argues that its AI tools can enhance creativity and enable faster content creation without diluting quality or relevance.
Community talk
AI Tools
UI-TARS-desktop
The Open-sourced Multimodal AI Agent Stack connecting Cuttin..
Opensourcetdi/awesome-private-ai
Curated list of tools, frameworks, and resources for running..
Opensource
HR have been using AI against you for years
Faceseek security tool or privacy threat?
now it can turn your PDFs and docs into clean fine tuning datasets
[UPDATE] DocStrange - Structured data extraction from images/pdfs/docs
The SERVE-AI-VAL Box - I built a portable AI-in-a-box that runs off solar, hand crank, and battery power for about $300
Dear Anthropic, Please give us 1M Context in max plan
I used AI tools for coding, and here is why AI is not going to replace soft engineers anytime soon! feel free to prove me wrong
How we reduced LLM spend by 60x (and Get 20 % Faster Responses)
MCPs that are part of my day-to-day Claude Code workflow
The 4-letter framework that fixed my AI prompts
I used ChatGPT to help me build a tool for studio-quality product photos because I was sick of paying so much money.
ChatGPT → Claude Memory Converter
Are the skeptic setting themselves behind?
I’m not saying I couldn’t live without 4o, but I will say, the difference is night and day.