Natural Language Processing news and analysis on Web Pulse: 5573 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
Spotify launched an AI-powered conversational assistant in beta on July 15, 2026, letting eligible Premium subscribers aged 18 and above in the United States, Ireland, and Sweden control the streaming service with natura…
A developer's guide demonstrates how to integrate an AI humanizer API into n8n, Zapier, and MCP-capable agents using the ToHuman API as a reference. The API rewrites AI-generated text to bypass detection tools like GPTZe…
A developer spent a month building linkedin-skills, an open-source system of ten AI-powered skills for Claude Code and Codex that automates the LinkedIn content grind—from planning and writing to engagement analysis—whil…
Domain-specific languages (DSLs) make LLM code generation production-ready by shrinking the output space so models produce verifiable results instead of plausible noise, according to Rachel Goldstein. Unconstrained gener…
Spotify launched a new AI chatbot called 'Talk to Spotify' for Premium subscribers in select regions, allowing users to create custom playlists, ask questions about songs and listening habits, and interact with audiobook…
Rime, a San Francisco startup building text-to-speech models for live phone calls, raised $24 million in Series A funding led by M13, with participation from Twilio Ventures, Corazon Capital, and Unusual Ventures. The co…
A free, open-source AI video clipper that runs entirely in the browser without uploading files has been released. The tool uses OpenAI Whisper for transcription and Qwen3.5-2B for clip selection, all executing locally vi…
Voice AI's biggest challenge is accurately detecting when a speaker has finished talking, a problem known as speech endpoint detection that affects response timing and user experience. Developers must optimize computatio…
A new pure-Go, cgo-free implementation of the Wyoming voice-assistant protocol, called gowyoming, has been released, providing client, server, and satellite roles for Go programs to interact with Home Assistant's voice p…
A voice agent is not a chatbot with a microphone bolted on; it is a real-time system with a dozen moving parts, and the person on the other end feels every one of them. The author, following the success of FastAPI, break…
Scikit-ollama, a library based on scikit-llm, bridges scikit-learn with locally running Ollama models to perform zero-shot text classification without a cloud API. The library enables users to load a movie review sentime…
Hume released the 'Real World VoiceEQ' benchmark, the largest human evaluation of voice AI to date, finding that Google Gemini leads in text-to-speech and speech understanding, OpenAI's GPT Realtime Mini leads in speech-…
OpenAI announced that data science teams can now use ChatGPT Work to generate root-cause briefs, KPI memos, and other analytics outputs directly from raw work inputs such as logs, Jira tickets, and Slack threads, cutting…
A developer demonstrates a bounded JSON repair loop for LLM output in Python, using only the standard library to classify syntax, shape, and semantic errors. The pattern follows a generate-parse-validate-repair-accept-or…
OpenAI has unveiled three new large language models named after celestial bodies, signaling a strategic focus on balancing performance with computational efficiency. The naming reflects the company's ambition to achieve …
AI chatbots are increasingly deployed by companies seeking cost savings, but 42% of consumers still prefer human interaction, according to a cited statistic. Poor implementation frustrates users, leading to lost sales an…
Researchers from Tel Aviv University and the University of Ljubljana have introduced ConlangCrafter, a fully automated multi-stage pipeline that uses large language models to construct coherent artificial languages with …
RESOURCE2SKILL, a new framework from an unnamed research team, transforms tutorial videos and multimodal content into executable skills for AI software agents, boosting agent performance by an average of 11.9 percentage …
A new framework called Anchor Supervised Revocable Decoding (ASRD) improves accuracy by up to 6.4% and boosts inference throughput by up to 7.2 times in diffusion large language models without additional training, accord…
A new benchmark called FlipSet reveals that most vision-language models fail at Level-2 visual perspective taking, with 75% of errors stemming from egocentric bias. In an evaluation of 103 models, the majority performed …