BackSearch lets agents query the web as-of a past date to avoid future leakage
brianrkelly · x · 2026-07-24
BackSearch is a web search tool for agent backtesting that lets users set an asof date, so the model only sees information available at that point in time.
- Built to avoid future information leakage from normal web search in forecasting and prediction workflows.
- The team says they’ve been using it internally for RL environments to train models.
- The first release is a narrow news index for 2026, with plans to expand based on interest.
- It works with the same interface shape as standard web tools, but requires aligning the backtest date with the LLM’s pretraining cutoff.
Related event: BackSearch: Time-Travel Web Search Tool to Prevent Data Leakage(2 posts)→
More from coding & agent
- Claude Code vs Codex: the harness may matter more than the model — al_kRicha · 2026-07-24
- Which model should power each agent task in production? — Logical-Silver-272 · 2026-07-24
- Offloop says its 4-person multi-agent harness beat Claude Code and Codex on GDPval — rohanpaul_ai · 2026-07-24
- Graph engineering, in one line: deterministic multi-agent loops become DAGs — vincent_koc · 2026-07-24
- BackSearch’s development was led by @ranko3000 with feedback from several collaborators — rosstaylor90 · 2026-07-24
- Default Codex CLI with GPT-5.5 scores 92.3% on XBOW, but the paper says the harness matters — evilsocket · 2026-07-24