Meta AI has launched Muse Code (in beta), a terminal coding agent in beta, powered by its new Muse Spark 1.2 model. Meta positions the pair as its subsequent step towards the frontier, with bigger fashions on the method. Muse Code targets complicated software program engineering throughout giant repositories: it plans adjustments, writes code, and…
NVIDIA has launched Alpamayo 2 Super, a 34B-parameter vision-language-action (VLA) model for autonomous driving, underneath an open business license. The said design goal is the long-tail occasions: uncommon, multi-agent conditions that typical detection-and-prediction stacks deal with poorly. The mannequin pairs a 32B VLM spine, constructed on NVIDIA Cosmos 3 Super Reasoner and post-trained with reinforcement…
CopilotKit has simply launched the Channels SDK. It is an open supply library that places an present agent inside a messaging platform. The core assumption is concentrated and verifiable. You have already got an agent. It already has a mannequin, instruments and enterprise logic. Channels offers it a spot to work with folks, with out…
In this tutorial, we construct an entire pixel-native retrieval-augmented era pipeline from scratch and study how doc retrieval works with out counting on standard HTML parsing, textual content extraction, or mounted chunking methods. We render net pages and PDF paperwork as photographs, divide them into overlapping tiles, generate multimodal embeddings with SigLIP, CLIP, or an…
Cursor Research has open-sourced Mixture-of-Kittens (MoK), the mixture-of-experts coaching megakernel behind its Composer fashions. MoK fuses each MoE communication and computation step right into a single deterministic kernel. Cursor group studies as much as 2.37x greater throughput than the strongest public baseline. It already powers Composer coaching throughout tens of hundreds of GPUs. Is it…
Reflex AI has launched XY, an Apache-2.0 Python charting library for interactive 2D visualization. Most Python charting stacks create one drawable object per row, so previous just a few hundred thousand factors, render, hover, and zoom degrade. XY strikes the work right into a native Rust core, sends the browser typed binary buffers as an…
In this tutorial, we construct a workflow for evaluating the safety posture of AI expertise with NVIDIA SkillSpector. We create an artificial ability market containing clear, dangerous, malicious, and MCP-based examples, then scan every ability by means of SkillSpector’s LangGraph inspection pipeline. We look at danger scores, categorized findings, confidence ranges, analyzer completeness, and executable-script…
Y Combinator crew has open-sourced QM (quartermaster), the multi-agent harness it makes use of internally. QM is described as a multiplayer agent harness for work, working in Slack and on the net. YC crew runs QM throughout accounting, authorized, occasions, and engineering, together with constructing QM itself. The entire venture ships underneath the MIT license….
Genspark has launched GenOffice as open supply beneath the Apache License 2.0. The repository describes it as an AI-native workplace suite for macOS and Windows: a phrase processor, a spreadsheet, a presentation editor and a PDF device. The implementation is 5 Electron apps sharing one engine layer, constructed round AI modifying as a first-class workflow…
In this tutorial, we design an end-to-end analysis workflow for PerceptionBench. This multimodal benchmark measures fine-grained visible notion capabilities throughout duties resembling OCR, counting, localization, contextual reasoning, comparability, depth understanding, and hallucination detection. We start by configuring a Colab-compatible atmosphere, putting in the required libraries, and loading a balanced subset of the dataset by a…