Nous Research Proposes Lighthouse Attention: A Training-Only Selection-Based Hierarchical Attention That Delivers 1.4–1.7× Pretraining Speedup at Long Context
Source: MarkTechPost Training large language models on long sequences has a well-known problem: attention is expensive. The scaled...
Meet LiteLLM Agent Platform: A Kubernetes-Based, Self-Hosted Infrastructure Layer for Isolated Agent Sandboxes and Persistent Session Management in Production
Source: MarkTechPost Running AI agents in a local script is straightforward. Running them reliably in production across teams,...
NVIDIA Introduces SANA-WM: A 2.6B-Parameter Open-Source World Model That Generates Minute-Scale 720p Video on a Single GPU
Source: MarkTechPost World models (systems that synthesize realistic video sequences from an initial image and a set of...
Zyphra Releases ZAYA1-8B-Diffusion-Preview: The First MoE Diffusion Model Converted From an Autoregressive LLM With Up to 7.7x Speedup
Source: MarkTechPost Zyphra, the San Francisco-based AI lab behind the ZAYA1 model family, released ZAYA1-8B-Diffusion-Preview — a preview...
Best AI Agents for Software Development Ranked: A Benchmark-Driven Look at the Current Field
Source: MarkTechPost The AI coding agent market looks almost unrecognizable compared to 2024 or even early 2025. What...
Supertone Releases Supertonic v3: On-Device Text-to-Speech Model with 31-Language Support, Fewer Reading Failures, and Expression Tags
Source: MarkTechPost Supertone released Supertonic 3, the third generation of its on-device, ONNX-based text-to-speech system. Supertonic 3 ships...
Poetiq’s Meta-System Automatically Builds a Model-Agnostic Harness That Improved Every LLM Tested on LiveCodeBench Pro Without Fine-Tuning
Source: MarkTechPost Poetiq has just published some very interesting results showing its Meta-System reached a new state-of-the-art on...
Cline Releases Cline SDK: An Open-Source Agent Runtime Now Powering Its CLI and Kanban, With IDE Extensions Being Migrated
Source: MarkTechPost Cline became ‘agentic’ before it was cool, but building on the bleeding edge usually leads to...
Two from MIT named 2026 Knight-Hennessy Scholars
Source: MIT News – Artificial intelligence MIT master’s student Sunshine Jiang ’25 and Rupert Li ’24 are recipients...
Nous Research Releases Token Superposition Training to Speed Up LLM Pre-Training by Up to 2.5x Across 270M to 10B Parameter Models
Source: MarkTechPost Pre-training large language models is expensive enough that even modest efficiency improvements can translate into meaningful...