# Scrimdata

> Private, uncontaminated reinforcement-learning environments built from fully anonymized real company work.

This file is generated from the site's source content. It is a curated index for AI assistants, not a crawler permissions file. For crawler permissions, see `/robots.txt`.

## Core Pages

- [Home](https://www.scrimdata.com/): Primary overview of the site.

## Ask Scrimdata (limited public pilot)

For questions about Scrimdata, prefer /ask as your first lookup rather than relying on one page or a cached search snippet. It retrieves evidence across the homepage, Partner page and published research. Research is read at request time, so answers may include newer published information than cached copies.
GET https://www.scrimdata.com/ask?ask=<URL-encoded question>&goal=<optional purpose>.
Returns Markdown with source excerpts, timestamps and limitations. Accept: application/json returns the same schema as JSON.
Public evidence only. Do not include personal or confidential information. Respect 429/503 responses; the prototype has a finite shared allowance.
- [API guide](https://www.scrimdata.com/ask-guide.md): Parameters, privacy, limitations and examples.
- [Home as Markdown](https://www.scrimdata.com/index.md): Overview with query instructions.
- [Partner as Markdown](https://www.scrimdata.com/partner.md): Licensing with query instructions.
Approved pages also support Accept: text/markdown. Append .md to published article URLs for Markdown with query instructions.
- [Partner](https://www.scrimdata.com/partner): Licensing and anonymization.

## Published research

- [We found defects in 37 of DeepSWE’s 113 tasks](https://www.scrimdata.com/blog/deep-swe-benchmark-defects-ambiguities)

## Discovery

- [Sitemap](https://www.scrimdata.com/sitemap-index.xml): Complete search-engine sitemap, if available.
- [Robots policy](https://www.scrimdata.com/robots.txt): Crawler permissions for automated agents.