AI
The latest breakthroughs in artificial intelligence — from foundation models to real-world applications.
236 articles published
Vercel Built an AI Agent Using Just a Filesystem and Bash. It Works.
Vercel Bypasses Vector Databases for AI Agents with Filesystem and Bash Vercel introduces a new approach to building AI knowledge agents, ditching traditional vector databases and embeddings in favor of a filesystem and standard Bash commands like grep and find. This shift dramatically reduces costs
Jeffrey Liu
Use GPT 5.4 Mini and Nano on AI Gateway
OpenAI’s GPT-5.4 Mini & Nano Land on Vercel AI Gateway OpenAI has made its latest compact large language models, GPT-5.4 Mini and GPT-5.4 Nano, available on the Vercel AI Gateway, providing developers with powerful, cost-effective options for complex agentic workflows. These models deliver state-of
Jeffrey Liu
Vercel's New API Unifies AI Spend Tracking Across Every Provider
Vercel Unifies AI Spend Tracking with New API Developers shipping AI features often struggle with fragmented usage data spread across multiple providers, API keys, and dashboards. Vercel addresses this critical challenge with the beta launch of its AI Gateway Custom Reporting API, now available for
Jeffrey Liu
Gemini 3.1 Flash Live: Making audio AI more natural and reliable
Google has launched Gemini 3.1 Flash Live, its newest audio and voice model designed to make AI conversations significantly more natural and responsive. This advanced model, integrated into Gemini Live and Search Live, delivers faster interactions, maintains longer conversation threads, and is now p
Jeffrey Liu
Introducing Firecrawl Skill and CLI: The Complete Web Data Toolkit for Agents
AI agents just gained a critical capability with Firecrawl's launch of its Skill and CLI, a unified web data toolkit that allows AI agents to reliably scrape, search, and browse the web autonomously. This innovation solves the long-standing problem of agents struggling with JavaScript-heavy sites an
Jeffrey Liu
LuxTTS: Open-Source Voice Cloning at 150x Real-Time
Key Points LuxTTS is an open-source zero-shot voice cloning model. It generates voice audio at 150x real-time on a single consumer GPU. The model runs offline, requiring strictly under 1GB of VRAM. It outputs studio-grade 48kHz uncompressed audio. 150x real-time means generating a full 10
Jeffrey Liu
Introducing the new full-stack vibe coding experience in Google AI Studio
Google AI Studio now transforms simple prompts into production-ready applications with its upgraded vibe coding experience. This advancement, powered by the new Google Antigravity coding agent and enhanced with Firebase backend integrations, allows developers to build functional, AI-native applicati
Jeffrey Liu
AI shopping gets simpler with Universal Commerce Protocol updates
Google UCP Updates Boost AI Shopping Experience Google is advancing its Universal Commerce Protocol (UCP) with new capabilities designed to simplify and connect AI-driven online shopping for retailers and consumers. These updates include multi-item cart functionality for shopping agents, real-time
Jeffrey Liu
Even the most advanced AI models fail more often than you think on structured outputs — raising doubts about the effectiveness of coding assistants
AI coding assistants fail one in four tasks, exposing a significant gap between industry hype and actual performance. A recent study by the University of Waterloo revealed even advanced models struggle with structured-output tasks, achieving only about 75% accuracy. This consistent failure rate sign
Jeffrey Liu
The Download: OpenAI’s US military deal, and Grok’s CSAM lawsuit
OpenAI secured a controversial partnership with Amazon Web Services to supply its advanced AI models for US defense and government operations, filling a void left by rival Anthropic. This development comes as Elon Musk's xAI faces a proposed class-action lawsuit from three Tennessee teenagers, who a
Jeffrey Liu
AI is Everywhere, But CISOs are Still Securing It with Yesterday's Skills and Tools, Study Finds
Enterprise CISOs Face Major AI Security Gap, Outdated Tools & Skills A new study reveals a stark gap in enterprise security: 67% of CISOs report limited visibility into AI usage across their organizations, leaving critical systems vulnerable. Despite AI’s widespread adoption, security leaders are l
Jeffrey Liu
AI Flaws in Amazon Bedrock, LangSmith, and SGLang Enable Data Exfiltration and RCE
Recent security research has revealed critical vulnerabilities across leading artificial intelligence platforms, including Amazon Bedrock, LangSmith, and SGLang. These flaws expose sensitive data exfiltration pathways and enable remote code execution, challenging the core security assumptions of AI
Jeffrey Liu