Open Source News & Articles on Techminis

A naukri.com initiative

New

Home

Open Source News

Amazon

274

Image Credit: Amazon

Unlocking Benefits with Bottlerocket: A Purpose-Built Container OS

Bottlerocket, an open-source, container-optimized Linux distribution - a Single-Purpose Operating System (SPOS) for hosting containers.
Bottlerocket images are specialized for a specific combination of container orchestrator and cloud environment.
Core Bottlerocket values of minimalism, operational reliability, and out of the box compliance offer increased uptime for container applications, improved security and resource utilization, lower management overhead and operational cost, and faster node ready times.
Security benefits from curated dependencies. Operational reliability through image-based updates. Compliance, out of the box as Bottlerocket offers AMIs hardened to CIS Bottlerocket Benchmark v1.0.0 optimized for use with Amazon ECS and Amazon EKS.
Bottlerocket does not have specific end-of-life policy other than a software lifecycle that falls under shared responsibility of users.
Bottlerocket images do not have an SSH server instead of being API-driven, offering an API to adjust system settings and trigger actions.
Moving to image-based deployments eliminates the problem of host-level diversity, simplifying fleet management, minimizing potential untested combinations, and eliminating the risk of unexpected local edits.
Bottlerocket Out of Tree Builds (OOTB) support is a mechanism for creating custom Bottlerocket variants that live within their own git trees.
Bottlerocket's development plan includes supporting Bottlerocket across Amazon EC2 instance types, doubling down on built-in compliance capabilities, making Bottlerocket ramp-up process easier and continuing to improve node ready times.
Bottlerocket is an open source project and welocomes feedback and contributions.

Read Full Article

16 Likes

TechCrunch

1.1k

Image Credit: TechCrunch

Anthropic proposes a new way to connect data to AI chatbots

Anthropic has open sourced the Model Context Protocol (MCP), which aims to connect AI assistants to data sources.
MCP allows AI models to draw data from various sources, improving their responses to queries.
Companies like Block and Apollo have already integrated MCP into their systems.
Anthropic invites developers to build with MCP and aims to provide toolkits for deploying production MCP servers.

Read Full Article

24 Likes

Analyticsindiamag

275

Image Credit: Analyticsindiamag

Why the Indian Open Source Community Loves Meesho Now More than Ever

Meesho has open-sourced its machine learning platform reacts in India.
The platform is used by more than 2 million sellers and 160 million plus consumers.
Meesho's ML platform handles cataloguing listing, pricing recommendations, ranking systems, fraud detection and user engagement.
The company plans to open-source more components in the future, starting with its feature store.
The feature store handles millions of requests per second and serves two trillion features per day.
Model Inference serves over 500k requests per second in prediction.
Meesho reduced the pod boot-up time from 6 to 7 minutes to just 10 seconds and reduced costs by 40%.
Mukherjee hopes startups will adopt their platforms, reducing costs and production time, while benefiting from community feedback.
Indian startups Zerodha, Flipkart, Swiggy, PhonePe, Zomato and CRED have public repositories on their GitHub page.
Meesho joined the group of Indian startups by open-sourcing a part of its tech stack.
Meesho benefits from open-source projects and now wants to give back.

Read Full Article

16 Likes

Bitcoinmagazine

333

You Can Now Invest In Bitcoin And Open-source Companies

A new platform called Timestamp has launched, allowing accredited and non-accredited investors to invest in Bitcoin and open-source companies.
The platform offers low investment minimums and provides users with the opportunity to review offerings and connect with company founders.
At launch, users can invest in companies such as CASCDR, Jippi, Lightning Bounties, Shopstr, and Sovereign.
Timestamp aims to make investing in Bitcoin and supporting open-source companies more accessible to a wider range of people.

Read Full Article

20 Likes

Marktechpost

342

Neural Magic Releases LLM Compressor: A Novel Library to Compress LLMs for Faster Inference with vLLM

Neural Magic has released the LLM Compressor, a state-of-the-art tool for large language model optimization that enables faster inference through advanced model compression.
LLM Compressor simplifies the process by incorporating various compression algorithms like GPTQ, SmoothQuant, and SparseGPT, reducing inference latency without sacrificing accuracy.
The tool supports activation and weight quantization, enabling utilization of INT8 and FP8 tensor cores for improved performance on NVIDIA GPU architectures.
In addition to compression, LLM Compressor also supports structured sparsity and weight pruning techniques to reduce model size while maintaining accuracy.

Read Full Article

20 Likes

Medium

338

Image Credit: Medium

The Case for an Open Search Foundation: Just an Idea.

Google's search engine monopoly has led to stagnation in innovation and made it increasingly difficult for smaller companies to offer meaningful competition as the ad-based model prioritizes profit over people, leading to reinforcing echo chambers that polarize public opinion.
Despite Google facing significant scrutiny from regulators for anti-competitive behavior and privacy concerns with fines imposed, the overall market dominance of Google remains largely unshaken and compounded by the immense barriers put in place that any potential competitor must overcome.
Emerging technologies such as decentralized web initiatives and AI-driven contextual search have shown promise, but are limited by heavy market dominance and the lack of funding required to achieve broad adoption.
To solve these problems, an Open Search Foundation is proposed, providing a non-profit, open collaboration that provides a neutral, trustworthy, and community-driven backbone for internet search together with big tech companies, governments, academic institutions, and independent developers to create a search infrastructure that belongs to everyone.
The Open Search Foundation would develop and maintain the core components of a search engine, ensuring that search remains fair, transparent, and focused on providing accurate and relevant information without the influence of advertisers.
Companies could use this core infrastructure to build their own customized search experiences, such as privacy-focused search by Apple, deep productivity integrations by Microsoft, or optimized product discovery by Amazon, while still contributing to the shared backbone.
The open foundation model brings several key advantages such as neutrality, transparency and innovation by fostering a true ecosystem of competition and innovation.
The timing for an initiative like this has never been better. The scrutiny on big tech is mounting, and governments are eager to find alternatives to the current monopolies.
An Open Search Foundation — funded by big players seeking a competitive edge against Google, supported by governments, and trusted by the public — could redefine the very fabric of the internet.
An open, community-driven backbone would not only provide a trusted search experience for users but would also foster a true ecosystem of competition and innovation unto building a fairer, more democratic system for information retrieval.

Read Full Article

20 Likes

Marktechpost

125

NVIDIA Introduces Hymba 1.5B: A Hybrid Small Language Model Outperforming Llama 3.2 and SmolLM v2

NVIDIA has introduced Hymba, a new family of small language models featuring a hybrid architecture that combines Mamba and Attention heads running in parallel.
Hymba models integrate transformer attention mechanisms with SSMs to enhance efficiency, allowing attention heads and SSM heads to process input data in parallel.
The Hymba-1.5B model combines Mamba and attention heads running in parallel with meta tokens to reduce the computational load of transformers without compromising memory recall.
Hymba outperforms other models in terms of efficiency and performance, making it suitable for deployment on smaller, less capable hardware.

Read Full Article

7 Likes

VentureBeat

160

AI2 closes the gap between closed-source and open-source post-training

The Allen Institute for AI (Ai2) has released Tülu 3, a model training family that brings open-source models closer in quality to closed-source models.
Tülu 3 allows researchers, developers, and enterprises to fine-tune open-source models without losing data and core skills, making them comparable to models such as OpenAI's GPT and Google's Gemini.
Ai2 claims that improving fine-tuning with open-source models like Tülu 3 will increase the adoption of these models in enterprises and research.
Tülu 3 offers enterprises the ability to mix and match datasets during fine-tuning and provides infrastructure code and evaluation frameworks for easy implementation.

Read Full Article

9 Likes

VentureBeat

106

Image Credit: VentureBeat

Chinese researchers unveil LLaVA-o1 to challenge OpenAI’s o1 model

Chinese researchers have developed LLaVA-o1, an open-source vision language model (VLM)
LLaVA-o1 introduces a structured reasoning process with four distinct stages
The model incorporates a novel technique called stage-level beam search for inference-time scaling
LLaVA-o1 demonstrates improved performance and outperforms other models in multimodal reasoning

Read Full Article

6 Likes

Marktechpost

392

Apple Releases AIMv2: A Family of State-of-the-Art Open-Set Vision Encoders

Apple has released AIMv2, a family of state-of-the-art open-set vision encoders.
AIMv2 improves upon existing models in multimodal understanding and object recognition tasks.
It incorporates a multimodal autoregressive pre-training framework with a Vision Transformer (ViT) encoder and a causal multimodal decoder.
AIMv2 achieves strong performance, scalability, and versatility in various applications, setting a new standard for open-set visual encoders.

Read Full Article

23 Likes

VentureBeat

Lightricks bets on open-source AI video to challenge Big Tech

Lightricks has released LTX Video, an open-source artificial intelligence (AI) model capable of generating high-quality video in real-time that could challenge the dominance of proprietary systems such as OpenAI, Adobe and Google.
Lightricks produced and launched the model, which has two billion parameters, hoping to spur innovation and accessibility in the burgeoning generative AI sector.
With LTXV, Lightricks produces high-quality video up to 121 frames at 768 x 512 resolution in just four seconds on just Nvidia’s H100 GPUs.
Competing head-on with tech heavyweights such as Adobe and Autodesk, Lightricks has placed its LTXV model on both GitHub and Hugging Face to allow initial previews and testing before the eventual full release under the OpenRAIL license.
Lightricks’s model will be freely available for academic and commercial use.
Lightricks’s goal is to make professional-grade generative video technology open and accessible for all industries such as gaming and e-commerce, and for creators and researchers.
As the future of AI video is up for grabs, Lightricks is hoping that collaboration and openness may prove the ultimate competitive advantage.
The model has a Diffusion Transformer architecture that ensures smooth motion and structural consistency between frames, reacts quickly, and generates high-quality results on affordable hardware.
Lightrick’s LTXV stands out due to its speed, which is achieved without sacrificing quality, giving studios and creators greater flexibility in generating high-quality results with rapid design cycles.
There is growing disagreement in the AI industry between open-source and proprietary voices. Lightricks believes that open innovation is crucial for smaller participants aiming to compete in an increasingly closed, proprietary world.

Read Full Article

5 Likes

Marktechpost

Meet Arch 0.1.3: Open-Source Intelligent Proxy for AI Agents

Arch 0.1.3 is an open-source, intelligent proxy for AI agents built on Envoy.
It functions as an intermediary for secure and efficient communication between AI agents.
Arch offers features such as contextual routing, protocol translation, and efficient data management.
It reduces latency by approximately 30% and encourages community-driven development.

Read Full Article

1 Like

Marktechpost

312

SmolTalk Released: The Dataset Recipe Behind the Best-in-Class Performance of SmolLM2

SmolTalk is a synthetic dataset designed to address challenges in the NLP landscape.
It combines synthetic and publicly available datasets to optimize learning and model training.
SmolTalk consists of datasets for instruction tuning, output generation, rewriting, and summarization tasks.
The SmolLM2 model trained on SmolTalk outperforms comparable models and improves performance in NLP tasks.

Read Full Article

18 Likes

Kaspersky

254

Image Credit: Kaspersky

Packages with infostealer found in PyPI repository | Kaspersky official blog

Malicious packages imitating popular LLMs found in PyPI repository.
Packages installed JarkaStealer malware on victims' machines.
Packages downloaded over 1700 times from 30 countries.
PyPI administrators removed the malicious packages, but caution is advised.

Read Full Article

15 Likes

VentureBeat

268

Image Credit: VentureBeat

OpenScholar: The open-source A.I. that’s outperforming GPT-4o in scientific research

A new AI system called OpenScholar, built by the Allen Institute for AI (Ai2) and the University of Washington, is helping researchers access, evaluate and synthesise scientific literature fast and with greater confidence.
OpenScholar has a retrieval-augmented language model that uses a datastore of open-access academic papers to deliver citation-backed, comprehensive answers to complex research questions.
It differentiates itself from models like OpenAI's GPT-4o by retrieving real papers and synthesising their findings to generate a grounded answer, outperforming GPT-4o with citation accuracy and factuality.
OpenScholar could democratise access to AI tools for smaller institutions, underfunded labs, and researchers in developing countries, being 100 times cheaper to operate than concurrent proprietary systems.
The system's datastore is restricted to open-access papers, leaving out paywalled research that dominates some fields, although the researchers hope future iterations can responsibly incorporate closed-access content.
OpenScholar represents a watershed moment in scientific computing, demonstrating the capacity to process, understand, and synthesise scientific literature with near-human accuracy, ushering in a new era of AI-assisted research.
The researchers have released everything—code, models, data, and tools—betting that openness will accelerate progress more than keeping their breakthroughs behind closed doors.
They've answered one of the most pressing questions in AI development: Can open-source solutions compete with Big Tech’s black boxes?
The answer, it seems, is hiding in plain sight among 45 million papers.

Read Full Article

16 Likes

For uninterrupted reading, download the app

Open Source News

Discover more