AI Briefing#open-source
The latest AI and developer news about #open-source, with the original source and a short summary.
The latest AI and developer news about #open-source, with the original source and a short summary.
The open-source app uses a hidden WebView and local embeddings to retrieve web content without any backend or search API.
Konnect exposes 228 design tools via Model Context Protocol to AI assistants like Claude and Codex, replacing the legacy Python/TypeScript KiCAD-MCP-Server with a single static binary.
The open-source library supports non-commutative geometric spaces like hyperbolic and SO(n) rotations to preserve event order in sequence modeling.
AesCode achieves an overall score of 85.89 on infographic generation benchmarks, outperforming GPT-5.5 and Claude Opus 4.8 by using decoupled cross-modal rewards.
The new open-source Parsimony benchmark quantifies AI coding efficiency by counting net syntax units added, revealing that Gemini 3.1 Pro Preview adds the least code among top performers.
The update introduces a playground tab for testing decision models like Kev and OpenJev with probability bars.
Gavin Baker argues that lower model-layer margins will drive compute demand, citing a 37% IRR for B200 clusters even with steep price declines.
The framework now supports serving LLMs via OpenAI, Anthropic, and Ollama protocols simultaneously using a unified .flm file format.
The tool transcribes one hour of audio in approximately 13 seconds on a base M3 MacBook Air using the Neural Engine.
NVIDIA has open-sourced the full Run:ai platform, including KAI Scheduler and Model Streamer, under the NVIDIA DSX AI Factory GitHub organization.
The experimental C-based Janas-LLM achieves 24 tokens/s generation for the 48.4GB Qwen3-Next-80B-A3B on a 32GB RAM laptop, significantly outperforming llama.cpp.
The open-source Spotless game simulates a Bosch dishwasher to evaluate how well frontier models can stack dishes for optimal cleaning.
The open-source library implements the Graph of Agentic Thoughts framework to capture decision reasoning and rejected alternatives in LangGraph applications.
ManimGX offers a Rust and wgpu-based core that is 90 times faster than ManimCE while maintaining compatibility with the original Python API.
The method reduces pretraining costs by over 99% and enables controlled comparisons of hybrid architectures like Gated KalmaNet, Gated DeltaNet, and Mamba-2.
The new model achieves 13+ tokens/s decoding speed on iPhone 17 Pro Max using Lookahead Expert Scheduling.
The new model achieves 85.9% of the BF16 baseline score while being 13.9× smaller in file size.
The U-Lens method estimates model uncertainty in a single forward pass without resampling or labeled data, outperforming supervised baselines in initial evaluations.
The new tool ranks GPU kernels by execution time with readable names and supports remote monitoring via SSH or host flags.
The framework achieves up to 1.85x speedup over Colocate and lifts NL2RepoBench scores from 52.5% to 58.5% on Qwen3.8-2.4T-A95B.
The new Apache-2.0 model family, ranging from 0.6B to 27B parameters, achieves state-of-the-art scores on both text and vision boards of the Jev Decision Index.
Pensieve integrates with coding agents via MCP to provide semantic local search, cutting average tool calls per lookup from 3.4 to 2.0 and reducing costs by 18% in benchmarks.
Blind A/B testing on Claude Code showed the plugin raised requirement fulfillment from 34% to 53% across six agent-engineering tasks.
An open-source project bundling tools like Ghidra to enable coding agents to analyze executables and the web.