Hyperresearch launches Claude Code research harness claiming DeepResearch-Bench RACE lead
Summary
Hyperresearch launches an agent-driven Claude Code research harness that claims the top spot on the DeepResearch-Bench RACE leaderboard, using a 16-step, tier-adaptive pipeline with adversarial critics and citation checks for runs ranging from 30 minutes to 80,000-word dissertations, though independent validation is still pending.
Key Points
- Hyperresearch launches as an agent-driven Claude Code research harness that claims to lead the DeepResearch-Bench RACE leaderboard, pending third-party validation.
- Its tier-adaptive 16-step pipeline uses adversarial critics, citation checks and patch-only edits to produce research reports from 30-minute queries to 25,000–80,000-word dissertation runs.
- The tool stores fetched sources in a searchable Markdown-and-SQLite vault, resumes crashed runs, and can recover legal open-access full text through Unpaywall, Europe PMC and CORE.