Natural Language Processing news and analysis on Web Pulse: 5625 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
Google unveiled what it calls the biggest upgrade to its Search box in 25 years at Google I/O, expanding the input to accept text, images, files, videos, and Chrome tabs while providing AI-powered suggestions. The compan…
Google unveiled its AI-enabled smart glasses on Tuesday at Google I/O, partnering with Samsung, Gentle Monster, and Warby Parker to launch two styles that integrate Gemini AI for navigation, notifications, real-time tran…
Google announced new YouTube features at Google I/O 2026, including a conversational search tool called Ask YouTube that compiles relevant videos in response to complex queries. The company also introduced Gemini Omni AI…
Amazon has released design patterns for scalable voice agents using Amazon Nova Sonic, Amazon Bedrock AgentCore, and Strands BidiAgent to address challenges like high latency and multi-agent coordination. The three patte…
Meta AI glasses are enabling disabled users to navigate daily tasks hands-free, with veterans like Donald Overton, who lost his sight in Iraq, using the device to read menus and navigate airports, and quadriplegic Marine…
ChatFeatured, a Toronto-based startup that helps brands influence how they appear in AI search engines, migrated from PlanetScale Postgres to Postgres managed by ClickHouse to unify transactional and analytical workloads…
Vicki Boykis added tags to her static blog using BERTopic and LLMs including Claude Code and Deepseek, completing the project in 6-10 hours. She argues that LLMs excel at compression tasks like topic modeling and labelin…
Tom Tunguz, a writer of 16 years, tested an "AI editorial council" of three different models—Gemini, Claude, and OpenAI Codex—to edit his fourth draft, but the result was a "fingerpaint disaster" as each model imposed it…
The global market for AI voice agents is projected to surge from $2.4–$3.14 billion in 2024 to between $35 billion and $47.5 billion by 2033–2034, driven by lower costs and improved technology. Among small and medium ent…
The latest edition of Community Wisdom explores strategies for screening AI-generated content in hiring processes, alternatives to Wispr Flow for voice transcription, and a comparison between multi-agent pipelines and th…
Researchers have developed RLP, an information-driven reinforcement pretraining objective that integrates exploration and chain-of-thought reasoning into the pretraining phase of large language models, rather than reserv…
Meta India launched Business AI on WhatsApp for small businesses in India, enabling 24/7 AI-powered customer support, lead capture, appointment booking, and sales directly within the app without additional tools. The fea…
A developer's attempt to speed up his Voxize dictation system by replacing OpenAI's multi-phase transcription pipeline with ElevenLabs' Scribe v2 model backfired after testing revealed the new model produced significantl…
Production RAG systems often fail after moving beyond demo stage due to underbuilt indexing, retrieval, and observability layers. The indexing pipeline ingests documents into chunks and vector embeddings, while the query…
Abridge, the AI-native healthcare company founded in 2018, is projected to support over 80 million patient-clinician conversations this year across 250 U.S. health systems, saving doctors 10–20 hours per week on clinical…
Together AI has released Violin, an open-source video translation tool that uses speech recognition, large language models, and speech synthesis to convert video content across languages. The tool, powered by the Togethe…
Malte Buettner benchmarked bounding box accuracy for Document AI models using pages from the FlashAttention-3 paper, testing Qwen, Kimi, and Mistral via OpenRouter. The evaluation scored models on coverage, intersection-…
Archivists and historians are turning to large language models to rapidly transcribe centuries-old handwritten documents at scale, a task that previously required specialized training or custom software. Researchers at W…
Every interaction with a modern large language model is structured as a list of messages, each tagged with a role—system, user, or assistant—that shapes how the model responds and how context is managed across multiple t…
Granola introduced meeting templates and recipes to standardize documentation for repeatable meeting types, allowing users to structure AI-enhanced notes and run saved prompts for extracting insights or generating conten…