Natural Language Processing news and analysis on Web Pulse: 5613 curated articles tracking the latest Natural Language Processing developments, tools, and research, updated continuously from vetted sources.
A developer describes context compression as a technique to reduce token costs and improve LLM answer quality by filtering retrieved text before generation. Extractive compression keeps verbatim sentences scoring above a…
A developer advocates for query rewriting before retrieval to improve search accuracy in support bots, using cheap LLM calls to generate multiple query variants or step-back abstractions. The technique, which includes mu…
A developer demonstrates how to extract CRM enrichment data from email signatures using regex, achieving over 95% accuracy on well-formed signatures. The approach leverages Nylas Agent Accounts to automatically process i…
Nylas released agent-owned mailboxes for e-commerce order support, enabling AI-powered triage of customer emails. The architecture creates per-store mailboxes via API, with rules for pre-sorting and an AI layer that answ…
Clara, an AI phone agent for local businesses, answers calls, books appointments, and follows up with leads 24/7. The service targets small businesses that cannot afford a receptionist, offering a 7-day free trial and a …
A new generation of AI-powered toys, such as the ChattyBear teddy, uses generative AI like ChatGPT to engage children as young as three in endless conversations, but a new report warns these toys pose risks including fos…
A developer building a conversational analytics engine found that language models produce confident wrong answers when querying raw database schemas. To solve this, they created a domain graph—a metadata map built offlin…
Trace, a new macOS app, provides on-device meeting transcription and flagging without cloud uploads, requiring no accounts or servers. It captures microphone and system audio, transcribes locally on Apple silicon, and al…
A developer built Interview Tree, a tool that uses Claude to transform user interview transcripts into structured opportunity trees. The app, built with Next.js 16 and React 19, sends transcripts to Claude via the Anthro…
Google Research released Gemini-SQL2 on June 12, a text-to-SQL system built on Gemini 3.1 Pro that achieved 80.04% execution accuracy on the BIRD benchmark, the first model to surpass 80%. The system outperforms GPT-5.5-…
A developer built a meal planning and grocery shopping agent using HazelJS, a TypeScript framework for AI-native applications. The agent uses multi-agent orchestration, retrieval-augmented generation (RAG) for recipe sea…
Hobnob, a free and open-source desktop app, records, transcribes, and summarizes meetings entirely on a user's laptop without cloud dependency or internet. The app runs local AI models for transcription and summarization…
Undetectable Humanizer Basic, a tool that rewrites AI-generated text to sound human, is available for $39.99 (reg. $1,080) through June 28 as part of Deal Days, offering lifetime access with no subscription.
A reviewer tested Siri AI on macOS 27 Golden Gate for 24 hours and found it improved but limited, especially for automating complex tasks like running benchmarks. The assistant could analyze screenshots and calculate ave…
Google Research's Gemini-SQL2, built on Gemini 3.1 Pro, achieved 80.04 percent accuracy on the BIRD text-to-SQL benchmark, surpassing OpenAI and Anthropic. The technology aims to enhance natural language features across …
Nylas has released a recipe for building multi-day support agents that use email threading headers as durable memory. By leveraging Message-ID, In-Reply-To, and References headers, the agent can maintain conversation con…
Google and Apple are pushing voice-based AI interactions with new features at Google I/O and WWDC, but this trend assumes universal comfort with talking to devices in public, potentially alienating those who prefer text …
A developer built an open-source AI assistant to help users navigate the EU AI Act, which becomes enforceable on August 2, 2026. The system uses a two-agent CrewAI architecture with Qdrant for vector search, analyzing us…
Metadata filtering before vector search is a cheap but often overlooked recall win. By applying a hard predicate on metadata like customer_id, the search space shrinks from millions to hundreds of chunks, preventing irre…
A developer built a personal AI assistant but struggled with context handling as conversations grew longer. They implemented a hierarchical context management system that keeps recent messages raw and periodically summar…