Open-Source AI

Today's Stories

1. 1

   Microsoft's open-source AI tools hacked to steal developer passwords

   Hackers targeted Microsoft's open-source development tools used by AI developers, successfully stealing login credentials. The attack specifically focused on tools that developers use to build AI applications, putting sensitive development work at risk.

   AI developers using these Microsoft tools need to change their passwords immediately and update to secure versions to protect their projects and accounts.

2. 2

   New open-source AI search tool Harness-1 outperforms GPT-5.4

   Researchers from University of Illinois, UC Berkeley, and Chroma released Harness-1, a 20-billion parameter AI search agent that scored 73% on information retrieval tasks, beating GPT-5.4's 70.9%. The tool is available for free under Apache 2.0 license, allowing anyone to use and modify it.

   Developers and companies can now access a powerful AI search tool for free that performs better than expensive commercial alternatives, potentially improving search features in apps and websites.

3. 3

   Student creates free legal AI workspace for lawyers

   An open-source legal AI workspace called AI-Blueprint was launched to help lawyers work with AI tools locally on their computers without sending data to external servers. The tool is designed to be privacy-focused and cost-effective for legal professionals.

   Law firms and legal professionals can now use AI assistance while keeping sensitive client information secure on their own computers instead of cloud services.

4. 4

   JetBrains releases Mellum 2 AI model with improved speed and performance

   JetBrains launched Mellum 2, a 12-billion parameter AI model that activates only 2.5 billion parameters at a time, achieving speeds of over 100 tokens per second even with large context windows. The model showed strong performance in coding and tool-calling tasks.

   Developers using JetBrains tools may soon get faster AI assistance for coding, debugging, and understanding large software projects.

5. 5

   Multiple technical improvements boost AI tool performance for developers

   Several updates were released for popular AI development tools including llama.cpp (a tool for running AI models locally), with improvements ranging from 2x to 3.7x speed increases. New tools were also created to help developers configure and optimize their AI setups more easily.

   Developers running AI models on their own computers will see significantly faster performance and have easier-to-use configuration tools.