4+ hour, 40+ min ago   (262+ words) Edge0 releases an 8B sparse MoE that runs in under 1 GiB of active memory on Apple Silicon, streaming experts from SSD on demand. Edge0 has released the 8B checkpoint and an accompanying streaming runtime that load mixture-of-experts weights from SSD as each token needs…...

3+ day, 14+ hour ago   (238+ words) AlphaSignal - Google open-sourced ARTEMIS, an Android automation agent from the Pixel Test Engineering team. - Claims 99%+ success on Google Research's AndroidWorld benchmark of 100+ multi-step tasks. - Flash mode hits 3-5s per step via an optimistic async pipeline with snapshot rollback. - Native MCP server…...

3+ day, 21+ hour ago   (230+ words) AlphaSignal - CozyClay is an open source browser-based previs studio: npx cozyclay and you're in. - Built with Three.js and React Three Fiber, licensed AGPL-3.0, requires Node 22+. - Block scenes, pose characters, author camera moves, then hand the shot to an AI…...

3+ day, 22+ hour ago   (291+ words) Perplexity released Q2D-Web, a large-scale benchmark with 190M documents and 70K agent-reformulated queries for evaluating retrieval in agentic RAG systems. Q2D-Web also targets a mismatch specific to agentic systems. Benchmarks typically use human-written queries, whereas RAG agents rewrite user requests into their…...

3+ day, 20+ hour ago   (313+ words) Zed's weekly release adds call hierarchy navigation, multi-select Git staging, auto language detection for scratch buffers, and live search-as-you-type by default. Zed 1.19 is out, and it's a polish release rather than a headline-feature swing. The update adds call hierarchy navigation,…...

4+ day, 4+ hour ago   (740+ words) Breakpoints, TTL math, Batch stacking, and a two-session split. Measured on OpenRouter, $0.003 vs $0.022 per turn Automatic caching billed us $0.022 per turn for the same 8,400 tokens that cost $0.003 once we placed the cache marker by hand. And as far as we…...

6+ day, 2+ hour ago   (400+ words) Fine-tuned 0.8B Qwen beat GPT-5.6 Sol xhigh, 2 million buyer profiles a day became 72 million, and GraphQL serving fell from $27 million to $1 million Shopify just got a 0.8B model to outperform GPT-5.6 Sol on a specialized task, while increasing throughput from 2 million to…...

1+ week, 1+ day ago   (395+ words) Perplexity open-sources details of Ivy, Tulip, and ROSE, its Rust plus Python embedding stack that beats vLLM on latency and throughput. Every Perplexity answer begins with a quiet but expensive step: turning a query and billions of candidate documents into…...

2+ week, 6+ day ago   (290+ words) Liquid AI and Artificial Analysis release an open-source suite that measures model quality, speed, latency, and memory across real phones, laptops, and embedded hardware. Model cards keep telling you what a language model can do on an H100. They rarely tell…...

4+ week, 1+ day ago   (139+ words) Open-source LLM inference and serving engine, originated at UC Berkeley's Sky Computing Lab. Built around PagedAttention for efficient KV cache memory management, with continuous batching, tensor and pipeline parallelism, and quantization support (FP8, GPTQ, AWQ). Supports 200+ Hugging Face model architectures with…...