Last year I shipped Extreme Mode, an autonomous research system that plans, searches, and synthesizes. It worked well, but I kept asking: what if the agent could actually think out loud, process your documents, and know when it’s truly done?
Today I’m launching Extreme Agent, the next evolution that transforms Extreme Mode from a search pipeline into a reasoning-first research partner.
What’s New
Visible Reasoning
The agent now explicitly thinks before every action. Instead of blindly executing searches, it:
- Records a brief thought explaining why it’s about to search
- Plans the next step with a clear action description
- Analyzes results between searches to decide what to explore next
You’ll see these thinking steps in the timeline, making the entire research process transparent and auditable.
Faster, Parallel Searches
Each search call now runs 3-5 queries simultaneously instead of one at a time. The agent strategically varies queries to capture different angles on the same topic, dramatically speeding up research while maintaining depth.
Document Search
Upload your PDFs, spreadsheets, Word docs, or CSVs and the agent can search through them alongside the web.
Supported formats:
- PDF documents
- Excel spreadsheets (.xlsx, .xls)
- Word documents (.docx)
- CSV files
Example:
“Compare the revenue projections in my Q4 report against the latest analyst coverage on [company].”
The agent searches your uploaded Q4 report and the web, synthesizing both into a single analysis.
Clear Completion
The agent now signals when it finishes. No more ambiguous endings. You’ll see:
- A clear completion signal in the timeline
- A brief summary of what was researched
- Confirmation that all planned steps were executed
Smarter X Search
The X (Twitter) search now supports:
- Include handles: Filter to posts from specific accounts
- Exclude handles: Filter out specific accounts
- Date ranges: Default 15 days, customizable window
- Multi-query: Multiple X queries per search
Date-Filtered Web Search
All web searches now filter by date. The agent defaults to recent results for time-sensitive queries, ensuring you get fresh information for breaking news, product launches, and trending discussions.
Example: Research with Documents
Prompt: “Analyze how our product roadmap aligns with current market trends in AI agents. I’ve uploaded our internal strategy doc.”
How the agent works through it:
- Reads the uploaded strategy doc for key themes
- Searches the web for current market trends
- Pulls sentiment from X about AI agents
- Compares internal plans against market signals
- Concludes with gaps and insights
Tips
Do:
- Upload relevant documents when you have internal context
- Ask for specific comparisons (“our plan vs market”)
- Specify time bounds (“last 2 weeks”, “since announcement”)
- Request X sentiment when community reaction matters
Important
The agent’s thinking is visible but not editable. If you see it heading in the wrong direction, start a new query with clarified intent.
Quick Start
- Go to scira.ai and start a new chat
- Toggle Extreme Mode (Pro subscription required)
- (Optional) Upload documents via the attachment button
- Enter your research brief
- Watch the thinking + search timeline unfold
- Review the completion summary and sources
Powering Scira Lookout
Extreme Agent is now the default search mode for Scira Lookout, the automated monitoring system that watches the web for you.
When you create a Lookout, it runs Extreme Agent on your schedule (daily, weekly, or monthly) and delivers research-grade reports straight to your inbox.
Example Lookout use cases:
- Daily AI news digest with sentiment from X
- Weekly startup funding roundups
- Monthly competitor analysis
Set it once, and Extreme Agent handles the rest.
Built for Scale
Extreme Agent is powered by xAI’s Grok 4.1 Fast, a reasoning model with a 2 million token context window. This means the agent can hold entire research sessions, multiple documents, and hundreds of search results in memory without losing context.
The infrastructure runs on Vercel’s Fluid Compute, which dynamically scales compute resources based on demand. Combined with the Vercel AI SDK for streaming and tool orchestration, Extreme Agent delivers fast, reliable research at any scale.
Try Extreme Agent Today
The future of research isn’t just faster search. It’s an agent that thinks alongside you, reads your documents, and knows when the job is done.
Questions or feedback? Reach me at zaid@scira.ai.