
Scira is now a sandboxed research agent
Scira now runs inside a real isolated environment with a persistent filesystem, bash, a code interpreter, and a set of built-in playbooks. Ask it to research something and build charts from the data. It will.
Field notes · 8 issues
Notes from the team building an open source, minimal AI search engine. Launches, research, and the small details we obsess over.

Scira now runs inside a real isolated environment with a persistent filesystem, bash, a code interpreter, and a set of built-in playbooks. Ask it to research something and build charts from the data. It will.
Headless research CLI on Flue and a 19-question grounded-QA suite, both on GitHub. Sweep with eight model configs on Parallel and Exa: Parallel leads on mean score and mean latency; per-cell charts in the post.

Pick sources before you send, read the timeline after, use the refreshed voice bar—and web search now starts on Parallel AI unless you change it in Settings.

Connect GitHub, Notion, Figma, Stripe, and 55+ more services directly into your AI workflows. No copy-pasting, no context-switching.

The next evolution of Extreme Research: an autonomous agent that thinks, plans, searches documents, and concludes with traceable intelligence.

Simply email your query. We search academic papers, the web, Reddit, and X to send you a comprehensive and well cited reply.

Deep reintroduction to Extreme Research and its new real-time X Search integration.

Your AI-powered watchdog that monitors the web for you.
