Latest Tech News & Insights

A newly discovered vulnerability in AI-powered CLI and IDE tools exposes sensitive browser data to theft. When these tools automatically open HTML files without confirmation, malicious repositories can exfiltrate cookies, localStorage, and sessionStorage contents, including API keys and authentication tokens.

While structured outputs promise reliable data extraction from LLMs, new research reveals they come at a significant cost to response quality. This trade-off between format compliance and accuracy could impact critical applications relying on AI data extraction.

Google Research introduces an AI-powered feedback system using Gemini 2.5 Deep Think to help theoretical computer scientists rigorously verify their papers before submission. The tool, tested at STOC 2026, identified critical errors in over 97% of papers and earned high marks from researchers for its ability to augment the peer review process.

A new AI writing assistant, BetterWrite, aims to revolutionize technical documentation by integrating directly into development workflows, promising to reduce documentation overhead while improving accuracy and consistency.

The Metrik API delivers hourly updates on Time to First Token (TTFT) across 26+ large language models from OpenAI, Anthropic, Google, and xAI. By exposing provider averages, performance comparisons, and change tracking, it equips teams with the data needed to fine‑tune model selection and optimize user experience.

A new open‑source library, StringZilla, delivers up to 150× speedups over ICU for Unicode case‑insensitive substring search by leveraging AVX‑512 kernels and a rigorous, spec‑compliant folding engine. The article explains how the library navigates the complexities of UTF‑8, showcases performance benchmarks across languages, and outlines future extensions for ARM and other scripts.