#Search

AILeapfrog, a small team doing high-volume AI visual and production work for fashion and brand clients, is building a "one brain" system that makes company knowledge and client context searchable through natural-language agents. The starter stack described is OpenClaw in a sandbox, a GitHub repository, Obsidian on the local machine, and Telegram as the access point. The system's research structure had roughly 1,200 files at the time of the talk.

AINVIDIA Dynamo uses a unified session-level identifier to make its inference stack aware of agent sessions, subagents, and their KV cache across turns and tool calls. On SWE-bench, two TP4 MiniMax-M2 replicas on one 8xH100 node gained roughly 12-16% throughput from program-aware scheduling over KV-aware routing alone. The post also describes experimental shared-pool indexing and a proposed KvHint interface for session-aware cache policies in vLLM and SGLang.

Why it matters: The post explains how session identifiers let an inference stack track agent working sets, with measured throughput gains on SWE-bench and agentic RL rollouts.

AIGoogle DeepMind has released EmbeddingGemma 2, an open 740 million parameter model that maps text, images, audio, and video into one embedding space. It is built on the Gemma 4 architecture under an Apache 2.0 license and supports an 8K token context window. The company reports a code benchmark gain from 68.76 to 78.68 on MTEB Code and says the model can run on-device with about 567MB of active RAM for the full multimodal version on a Google Pixel 11 Pro.

Why it matters: The release shows how a 740M-parameter embedding model can cover text, code, images, audio, and video on local hardware, with memory and storage figures to compare against other on-device options.

AIGoogle released EmbeddingGemma 2, an open embedding model under the Apache 2.0 license that maps text, code, images, video, and audio into a shared 768-dimensional space. Developers can load a 270M-parameter text and code setup, or add vision and audio encoders up to a 740M-parameter full multimodal model. Matryoshka truncation to 256 or 128 dimensions reduces vector storage, with the guide noting quality losses on image, video, and speech retrieval at lower dimensions.

Why it matters: The guide gives concrete encoder sizes and dimension-storage tradeoffs, showing how to choose a configuration for text, code, image, video, and audio retrieval.

AIGoogle DeepMind launched EmbeddingGemma 2, an open-weight 740M parameter model that maps text, images, video frames, and audio into one vector space. The model can run on-device, with about 567MB active RAM for the full multimodal model on a Google Pixel 11 Pro, and is available through Google AI Edge Gallery, Google AI Edge Foresight on Mac, and MediaPipe Tasks, with ML Kit support coming in the weeks ahead.

Why it matters: The post names concrete on-device apps, memory footprints, and latency figures, showing how a multimodal embedding model can power local search without cloud calls.

AICloudflare introduced a Web Search API through AI Gateway, partnering with Ceramic.ai, Exa, and Linkup to give agents fresh web results instead of guessed URLs. Requests appear in AI Gateway logs and draw from AI Gateway credits, with partners committing to Cloudflare's Verified bots crawling standards and including source links in results. Partners at list API pricing without markup are available via a REST endpoint or a Workers binding, with native server tools planned.

AIAi2 released AstaBrief 8B, a model that turns a research question and retrieved literature excerpts into a cited report, along with its training data. In Asta's Generate a report feature, Fast mode averages 51.1 seconds per report versus 178.5 seconds for Thinking mode, about 3.5x faster. The model is built on Qwen3-8B with supervised fine-tuning and DPO, and institutions can run its open weights on their own infrastructure.

Why it matters: The post explains the data filtering and one-pass generation choices behind a fast open-weights report model, showing what worked and what did not.

AIMeta is testing Forum, a standalone iOS and Android app in the US that syncs with users' Facebook Groups to consolidate their conversations in one place. The update adds a new top-contributor role replacing previous badges, an AI-powered Ask feature that surfaces group posts and comments, and topic labels for exploring interests.

AIThe United Nations system launched UN System Data Commons, an open-source platform built on Data Commons by Google that integrates siloed global statistics into one AI-ready knowledge graph. Users can query it in natural language, browse by location or theme, and use MCP-enabled AI agents to fetch verified figures and draft charts or reports. The UN plans to add more datasets, aiming to include 80% of UN system statistical datasets by 2027.

AIBaseten has launched Hosted Tools, starting with Baseten Grounded Inference, a server-side web search capability for models hosted on Baseten. Developers enable it by adding a hosted search tool to a Messages, Chat Completions, or Responses request, and the platform runs the search loop with partners Exa, Keenable, Parallel, and You.com. In Baseten's benchmarks, agents using the hosted tools saw a 15% reduction in end-to-end latency compared with client-side tools, and the feature is in playground preview with 25 RPM rate limits and $2 of free credits.

AIMistral models now power Firefox Smart Window, Mozilla's AI browsing assistant in beta. Mistral will serve users in France and North America first, with the United Kingdom and Germany expected later this year. Conversations are not saved on Mozilla's servers by default, and partners such as Mistral agree to zero data retention.

AIAlibaba NLP has released UEmbed-4B, a decoder-only multimodal embedding model built on Qwen3.5 4B that outputs both dense and sparse embeddings from one forward pass. It handles text, image, video, and mixed-modal inputs for retrieval and visual-document search, and sparse activations map to vocabulary terms usable with inverted indexes. The model is available on Hugging Face in a family that also includes 2B and 9B variants.

AIFireworks AI describes fine-tuning Qwen3-Embedding-8B on private (query, positive) pairs using bidirectional InfoNCE loss through its Training SDK, then serving the model via an OpenAI-compatible embeddings endpoint. The post reports that around 150 training steps was enough, that rank-32 LoRA landed within about one point of full-parameter fine-tuning, and that gains were largest where the base model struggled, while tasks like CoSQA and FiQA2018 showed flat results.

AIAlibaba NLP released LaSER-Qwen3-8B, an 8B-parameter dense retriever built on Qwen/Qwen3-8B that internalizes explicit reasoning into latent space through continuous latent thinking tokens. The model scores 29.3 nDCG@10 on the BRIGHT benchmark, ahead of the rewrite-then-retrieve pipeline's 28.1, and carries a 4096-dimension embedding with an 8192-token maximum sequence length. It is licensed under MIT and adds about 1.7× latency over standard single-pass dense retrievers.