RTRoss Taylor@rosstaylor902:21 AM · Jul 24, 2026

824TECH

To predict the future, typically you backtest a strategy on historical data, eg financial time series. This is problematic for agents where ideally you’d like them to get world information from the web - but default web search tools leak future information. BackSearch is the answer for this: it has exactly the same shape as standard web tools, but the user can set an as_of date to simulate web information at a point in time. Note that for LLMs there is a dance to be had as LLMs can get information through pre-training - and are increasingly good at memorising it - so you typically have to configure your BackSearch date to align with the language model cutoff. We’ve been using this internally for RL environments to train our models, but we wanted to test if there is wider interest. For now we release a narrow index focused on news for 2026. Based on interest we can expand it to new domains and longer backdated horizons. Let us know what you think.

GR

General Reasoning@GenReasoning

🔍 Introducing BackSearch. LLMs are increasingly asked to predict the future, but a good backtest requires a snapshot of the internet at a point in time. BackSearch allows LLMs to search the web as it was on a particular date. It’s great for: 🔮 Forecasting and prediction markets. 📈 Quantitative finance. 🌍 RL environments that simulate the world. 🥶 Freezing websearch for benchmark reproducibility We’re releasing a narrow slice of our index to begin with, focused on the news domain for 2026. Based on feedback we’ll open up more of our index in subsequent releases. 👇 Try BackSearch out with the link below.