The Research Problem
Manual research is one of the most time-consuming tasks in any knowledge worker's day. You open a dozen browser tabs, scan articles, cross-reference facts, and try to synthesize everything into a coherent narrative. What should take minutes often stretches into hours. We built the Iris research pipeline to solve exactly this problem, delivering thorough, cited research at a pace that keeps up with how fast you think.
How the Pipeline Works
When you ask Iris to research a topic, it kicks off a multi-stage pipeline that operates largely in parallel. Here is what happens behind the scenes:
- Query Expansion — Iris decomposes your question into multiple targeted search queries designed to maximize coverage across different angles and subtopics
- Parallel Web Search — Using integrated search tools, Iris retrieves results from across the web simultaneously, prioritizing authoritative, recent, and diverse sources
- Browser Automation — For sources that require deeper reading or live behind interactive pages, Iris uses browser automation to navigate sites, extract content, and capture information inaccessible through standard search APIs
- Content Extraction — Raw web content is cleaned, structured, and distilled into the most relevant passages for your query, discarding boilerplate and advertisements
- Synthesis — Finally, the language model synthesizes all extracted information into a coherent, well-organized response with inline citations
Balancing Quality and Speed
One of the hardest engineering challenges was balancing thoroughness with latency. Researching more sources improves quality but increases wait time. We solved this through intelligent source prioritization. Early in the pipeline, Iris evaluates the likely relevance of each source and allocates more processing time to high-value results while quickly discarding low-quality ones. The result is that Iris typically processes 50 or more sources in under 30 seconds, delivering research that would take a human analyst an hour or more to compile manually.
Source Transparency and Citations
We believe research is only useful if you can verify it. Every research output from Iris includes source citations so you can trace any claim back to its origin. This transparency is essential for professional use cases where accuracy is non-negotiable, from market analysis and investment due diligence to academic literature reviews and competitive intelligence.
Our research pipeline continues to improve with every iteration. We are actively working on support for specialized databases, academic paper repositories, real-time data feeds, and image-based source retrieval to make Iris the most capable research assistant available anywhere.