Daily email with new technical posts, emerging signals, and raw sources — written by AI, reviewed for accuracy.
Improve your Vibe Coding Skills using Agent Skills and evaluate them with Skills Bench and W&B Weave
Learn how to build a brand-aware evaluation loop for image generation models using W&B and Weave. Track prompt fidelity, FID, and brand compliance seamlessly.
Join Weights & Biases by CoreWeave at AI Engineer World’s Fair in SF. Visit booth UG 24 for demos, swag, speaker sessions, and new Weave agent tracing features.
Explore Kimi 2.7 through progressively more complex software engineering tasks while using Weave to trace, analyze, and evaluate every step.
Your experiments are already tracked. Now let the agent read them, analyze them, and help you turn every experiment into continuous improvement.
From the launch of ARIA, our new ML research assistant, to new Serverless Inference models and more, here's a roundup of what we released in June
Learn how our tabular data agent, took #1 on the MLE-bench tabular slice with two golds, three above-median, four-for-four valid submissions.
A close look at how Zhipu built GLM-5.2, from DeepSeek Sparse Attention to critic-based PPO, with a live agentic coding demo traced in W&B Weave.
Google has introduced Gemini 3.6 Flash and Gemini 3.5 Flash-Lite with a strong emphasis on measurable improvements across coding, reasoning, computer use, long-context understandin…
How GPT-5.6, GPT-5.5, and GLM-5.2 compare on repository-level coding: hidden-test pass rates, reliability, latency, and estimated cost, measured with W&B Weave.
Or, how I trained an agent to research AI
Reservoir computing (RC) couples a fixed recurrent dynamical system with a trained lightweight readout, but this efficiency is partly lost during hyperparameter selection: the recu…
Zeroth-order (ZO) optimization estimates gradients using only forward-pass evaluations, making it suitable for fine-tuning non-differentiable, event-driven spiking neural networks …
The Traveling Salesman Problem (TSP) is one of the most extensively studied NP-hard optimization problems. Genetic Algorithm (GA)-based solvers, such as the Edge Assembly Crossover…
The retrieval dynamics of a modern Hopfield network is the gradient flow of a log-sum-exp energy, while the attention update is its exact difference-of-convex minimization step. We…
Reinforcement learning (RL) algorithms have made strides over the past decade applying them to a wide range of problems and control tasks. However, the deployment of RL on neuromor…
Population optimizers such as CMA-ES, DE, and multi-objective evolutionary algorithms drive search mainly through selection signals that are scalar or rank based: such a signal ind…
Analog circuit topology synthesis remains challenging because useful designs occupy a tiny fraction of a combinatorial search space, and small structural changes can induce highly …
Continuous optimisation methods need to balance sharing information and maintaining alternative search directions. In this paper, we introduce Mycelial Search (Myco), a graph-struc…
Mixed-integer programming (MIP) lies at the core of operations research and industrial optimization. While large language models (LLMs) have recently shown promise in automated MIP…
While contemporary Evolution Strategies handle integer optimization problems effectively, their adaptation mechanism is grounded in $\ell_2$-based Gaussian models, which are not na…
Deep Equilibrium Models (DEQs) compute predictions from a hidden representation unchanged by the model update. Training through this equilibrium uses implicit differentiation and r…
This paper proposes the Coronavirus Optimization Algorithm (COA), a SARS-CoV-2-inspired success-history adaptive evolutionary optimizer for box-constrained continuous global optimi…
Developments in high-performance computing (HPC) technology continue to drastically increase quantities of available processing power. In the context of digital evolution, this exp…
Low-earth-orbit (LEO) satellites enable high-resolution, large-scale Earth observation for applications such as disaster monitoring and environmental surveillance. However, cloud c…
ES-HyperNEAT evolves substrate topology through adaptive quadtree subdivision; to our knowledge, no implementation with full population-level GPU parallelization exists. We present…
Social interaction can improve collective learning but also amplify early mistakes. We study this tension when the credibility of social information is generated by the sender's ow…
Lipschitz constants are a standard way to quantify the sensitivity of neural networks to small input perturbations, but computing them is difficult even for shallow ReLU networks. …