Open Source News & Articles on Techminis

A naukri.com initiative

New

Home

Open Source News

TechCrunch

3.5k

Image Credit: TechCrunch

Open source devs are fighting AI crawlers with cleverness and vengeance

Open source developers are facing challenges with AI web-crawling bots, as they are disproportionately affected due to sharing more infrastructure publicly and having fewer resources.
An AI bot named AmazonBot caused DDoS outages by disobeying robot.txt rules on a Git server website hosting FOSS projects.
FOSS developer Xe Iaso created Anubis, a reverse proxy check inspired by Egyptian mythology, to block bots and allow human-operated browsers.
Anubis gained quick popularity in the FOSS community by preventing bot requests with a proof-of-work check and presenting a cute anime picture for successful human requests.

Read Full Article

24 Likes

Siliconangle

Image Credit: Siliconangle

OpenAI adds support for Anthropic’s MCP LLM connectivity protocol

OpenAI is adding support for Anthropic's MCP LLM connectivity protocol.
MCP is an open-source technology that allows large language models to perform tasks in external systems and access their data.
OpenAI will integrate MCP into ChatGPT's desktop client and Responses API for easier access to external systems.
Anthropic's MCP protocol provides software building blocks to quickly connect LLMs to external systems, making integrations easier and faster.

Read Full Article

3 Likes

Marktechpost

Meet Open Deep Search (ODS): A Plug-and-Play Framework Democratizing Search with Open-source Reasoning Agents

Researchers from University of Washington, Princeton University, and UC Berkeley have introduced Open Deep Search (ODS), an open-source search AI framework designed for seamless integration with any user-selected large language model (LLM).
ODS comprises two central components: the Open Search Tool and the Open Reasoning Agent, which enhance content retrieval and reasoning accuracy.
ODS demonstrates superior performance compared to proprietary alternatives, achieving higher accuracy on benchmark tests.
By democratizing search-enhanced AI, ODS encourages innovation, transparency, and broader participation in the development of sophisticated search and reasoning capabilities.

Read Full Article

Medium

394

Image Credit: Medium

Open Source Products — It’s Time

Applying open source methods can revolutionize various industries, not just software.
NodeCosmos offers a holistic view of product components and processes, organizing them into transparent chunks.
Using branches in NodeCosmos allows for collaborative contributions, ensuring continuous improvement.
Open models foster inclusivity, empowering unexpected sources of insights, and future-proof projects in a rapidly evolving world.

Read Full Article

23 Likes

Hackernoon

Image Credit: Hackernoon

Leveraging RAG With Reddit URLs on Diabetes: Open Source LLMs for Enhanced Knowledge Retrieval

Pre-trained LLMs, being open source, offer valuable assistance in knowledge retrieval, using RAG to leverage generative power while incorporating external knowledge.
Utilizing LLMs like Ollama for sourcing experiential data on diabetes is illustrated in a step-by-step experiment using Reddit URLs.
The process involves fetching and processing content from multiple URLs, splitting text into chunks, initializing Ollama embeddings, and creating a FAISS vector store.
Ollama LLM is then used to generate responses based on stored knowledge, allowing RAG to retrieve context and answer user queries effectively.
The ask_question_with_fallback function is designed to query and respond to user questions using stored knowledge, falling back on general knowledge if needed.
In the absence of relevant documents, the fallback function utilizes the LLM's general knowledge to address user inquiries, promoting efficient knowledge retrieval.
The methodology showcases the powerful application of RAG with URLs for improved knowledge access in diabetes inquiries, combining real-world insights from platforms like Reddit with LLMs.
By merging community insights with open-source LLM capabilities, this cost-effective approach enhances personalized information on diabetes management, understanding the importance of lived experiences.
The evolving potential of AI in healthcare, exemplified through knowledge-enhanced retrieval systems, signifies a collaborative and supportive future for individuals navigating diabetes.

Read Full Article

1 Like

TechCrunch

371

Image Credit: TechCrunch

eBay backs WunderGraph to build an open source GraphQL federation

eBay-backed startup WunderGraph raises $7.5 million in Series A funding to scale its open source GraphQL federation, with investors including eBay Ventures.
WunderGraph aims to address API sprawl in the GraphQL ecosystem by offering an open source alternative to GraphQL products like Apollo.
Founded in 2020, WunderGraph's core team includes CTO Dustin Deus, CEO Jens Neuse, COO Björn Schwenzer, and CCO Stefan Avram, with operations based in the U.S.
GraphQL optimizes API data retrieval, reducing bandwidth usage and costs while supporting the API economy's shift towards microservices.
WunderGraph's platform unifies multiple APIs and databases, enabling collaboration and exploration, with a recent focus on open source federation.
In response to Apollo's license change, WunderGraph introduced Cosmo as an open source alternative for GraphQL federation in late 2023.
Partnering with eBay, WunderGraph gains design input for building a scalable product, while eBay benefits from an open source GraphQL federation tailored to its needs.
WunderGraph emphasizes open source as essential for attracting big-name customers and plans to expand its team and offerings with the new funding.
With a focus on smart collaboration and governance tools, WunderGraph sees open source as the future of API management, catering to enterprises' demands for transparency and control.

Read Full Article

22 Likes

Mjtsai

259

Apple’s Dormant CUPS

CUPS (Common UNIX Printing System) was originally developed by Michael Sweet in 1999 and later adopted by Apple in 2002.
Since the lead developer's departure from Apple in 2019, the CUPS git repository has become dormant, with minimal activity observed.
The Linux Foundation's OpenPrinting has created a temporary fork of CUPS to continue bug fixes and distribution patches.
If Apple ceases CUPS development, OpenPrinting plans to continue the project with the help of Michael Sweet.

Read Full Article

15 Likes

Silicon

296

Image Credit: Silicon

Signal App In Spotlight Amid Secret Chat Controversy Of US Officials

The Signal messaging app is under scrutiny after a journalist was inadvertently added to a secret chat of senior White House officials.
The US National Security Council is investigating how the journalist's number was added to the chat.
The use of a commercial messaging app for discussing secret military plans has raised security concerns.
Signal is a messaging app with end-to-end encryption, owned by a US-based non-profit organization.

Read Full Article

17 Likes

Siliconangle

208

Image Credit: Siliconangle

Google’s Gradient Ventures leads $1.2M round for Silverstream AI to scale autonomous web agents

Silverstream AI has raised $1.2 million in pre-seed funding from Gradient Ventures to scale autonomous web agents.
The company focuses on developing infrastructure and tools to build, scale, and monitor autonomous web browsing agents.
Silverstream AI provides a Privacy & Safety Layer to manage sensitive data and a Decision Making Engine to control browser sessions.
Silverstream AI also announced the upcoming release of Pasta-1T, an open dataset of web trajectories for training web agents.

Read Full Article

12 Likes

VentureBeat

300

Image Credit: VentureBeat

DeepSeek-V3 now runs at 20 tokens per second on Mac Studio, and that’s a nightmare for OpenAI

Chinese AI startup DeepSeek has released a new large language model, DeepSeek-V3-0324, under an MIT license allowing commercial use.
The model can run on Apple's Mac Studio with M3 Ultra chip, achieving over 20 tokens per second.
DeepSeek's launch lacked typical fanfare, with no whitepaper or marketing, but the model has shown improvements over its predecessor.
DeepSeek-V3-0324 operates with a MoE architecture, activating only 37 billion out of its 685 billion parameters for specific tasks, enhancing efficiency.
The new model incorporates MLA and MTP technologies, boosting output speed by nearly 80%.
With a 4-bit quantized version offering reduced storage footprint, it can run on high-end consumer hardware, challenging traditional AI infrastructure.
Chinese AI companies like DeepSeek opt for open-source licensing, contrasting with Western companies keeping models behind paywalls.
This strategy enables rapid transformation and AI innovation in China, with tech giants like Baidu, Alibaba, and Tencent also embracing open-source models.
DeepSeek-R2, an advanced reasoning model, is anticipated to build upon DeepSeek-V3-0324, potentially competing with models like GPT-5 from OpenAI.
By democratizing access to AI technology through open-source models, DeepSeek is reshaping the future of AI development and adoption globally.
DeepSeek's approach reflects a broader trend towards making AI more accessible and empowering a wider range of developers and researchers in the field.

Read Full Article

18 Likes

Medium

267

Image Credit: Medium

Building a HyperStartup Alone: Tackling Hard Problems in 2025

Most startups focus on speed, funding, and rapid iteration, while avoiding hard problems that require specialized knowledge and take too long.
HyperStartups are deep, solo-driven, open startups focused on solving hard problems like AI, abductive reasoning, language models, etc.
The primary focus is NeuralLang, a new DSL designed to make neural networks as expressive and modular as programming languages.
The creator of HyperStartups wants to document every step, involve others in the progress, and create a culture of deep problem-solving.

Read Full Article

16 Likes

Medium

234

Image Credit: Medium

11 Open-Source Projects That are Legit Game-Changers

APITable is an open-source tool that enables building apps with a spreadsheet vibe.
It integrates with over 6,000 other apps via Zapier, allowing for easy automation of tasks.
APITable is cost-effective and team-friendly, making it a great choice for quick app development.
GitHub link: https://github.com/apitable/apitable

Read Full Article

14 Likes

Hackernoon

255

Image Credit: Hackernoon

C++ Metaprogramming: Compilation of Calculations, from Basic Techniques to Advanced Methods

Compile-time calculations in C++ provide advantages such as data and logic validation, reduced execution time, and enhanced code safety.
Templates and constexpr are key compilation mechanisms for computations in C++.
Historically, metaprogramming in C++ emerged through templates, allowing complex computations using recursive templates.
Examples like factorial and Fibonacci calculations showcase template metaprogramming but can lead to code complexity and debugging challenges.
Debugging complexity, increased compilation time, and template limitations are drawbacks of template metaprogramming.
The introduction of constexpr in C++11 provides a modern approach for compile-time computation, offering simplicity and ease of use.
constexpr functions and variables allow evaluation at compile time, simplifying code and improving readability.
Modern constexpr features include conditional operators, loops, and constexpr classes for more advanced compile-time computations.
Useful features like static_assert and if constexpr further enhance compile-time checks and conditional operations.
Pitfalls of extensive metaprogramming include increased compilation time, expression limitations, potential hidden copies, and debugging complexity.

Read Full Article

15 Likes

Bigdataanalyticsnews

251

Image Credit: Bigdataanalyticsnews

Top 20 Open-Source LLMs to Use in 2025

In 2025, open-source large language models (LLMs) are becoming more powerful, democratizing access to advanced AI capabilities.
Key models in the open-source ecosystem offer unique strengths for various applications in AI.
LLMs are transformer-based AI systems with hundreds of millions to billions of pre-trained parameters.
Models like Llama 3.3, Mistral-Large-Instruct-2407, and Gemma-2-9b-it excel in different tasks such as chatbots, NLP, and coding assistance.
LLMs like DeepSeek R1, Claude 3.5 Sonnet, and GPT-4 Turbo offer enhanced reasoning and efficiency for AI tasks.
Alibaba's Qwen2.5-72B-Instruct and Elon Musk's xAI's Grok 3 are notable models for enterprise and conversation AI.
Choosing the right LLM depends on factors like use case, model size, open-source licensing, and compute requirements.
Open-source LLMs provide cost-effective solutions, customization, transparency, and independence from major AI providers.
They also offer faster innovation, community support, multi-language capabilities, and ethical AI practices.
With open-source LLMs shaping the future of AI, developers have powerful tools for coding, research, and automation.

Read Full Article

15 Likes

Pymnts

213

Image Credit: Pymnts

Perplexity Proposes Open-Source Version of TikTok Algorithm

Artificial intelligence (AI) search engine firm Perplexity is proposing an open-source version of the TikTok algorithm.
Perplexity states that it is singularly positioned to rebuild the TikTok algorithm without creating a monopoly, combining technical capabilities with Little Tech independence.
They aim to maintain the personalized 'For You' feed while building a real-time recommendation system, ensuring a seamless user experience.
Perplexity plans to make the TikTok algorithm transparent and the For You feed open source in their vision for a neutral and trusted TikTok.

Read Full Article

12 Likes

For uninterrupted reading, download the app

Open Source News

Discover more