Sakana AI has launched Sakana Fugu, an orchestration model that routes tasks across a swappable pool of frontier LLMs via a single OpenAI-compatible API. Fugu Ultra outperforms individual models on key benchmarks like SWE Bench Pro and GPQA-D, and the system demonstrates superior performance on complex, multi-step tasks such as auto-research, Rubik's Cube solving, and blindfold chess.
MoonMath AI has open-sourced a bf16 forward attention kernel for AMD's MI300X GPU, written in HIP rather than assembly. It outperforms AMD's own AITER v3 kernel across all tested shapes and rounding modes, with speedups up to 1.26x, and maintains bit-identical numerical accuracy.
Large language models are stateless by default, requiring memory mechanisms to retain context across interactions. The seven types of agent memory—working, semantic, episodic, procedural, retrieval, parametric, and prospective—categorize memory by form and duration, enabling agents to plan, learn, and act over time. Each type serves distinct use cases, from storing user preferences to scheduling future goals, and together they form a comprehensive system for long-horizon, context-aware AI agents.
This tutorial demonstrates how to create interactive dashboards in Python using Prefab's component-based UI framework. It generates synthetic pipeline data, integrates reactive controls like charts, forms, and tabs, and exports the app as a static HTML file for direct preview in Google Colab.
xAI has introduced /goal, a mode in Grok Build that enables long-running, autonomous execution of multi-step coding tasks. The feature plans work, executes a progress checklist, and verifies results by reviewing code, inspecting webpages, or running scripts, ensuring completion before declaring success. Access requires a SuperGrok or X Premium Plus subscription.