# Ooak Data

> Applied AI research lab building the data infrastructure for frontier AI. We
> turn real company data into RL environments where AI agents learn to work in
> the real world: anonymized digital twins of enterprise tools, expert-calibrated
> tasks, multi-step multi-tool workflows.

Every page on this site is available as Markdown via HTTP content negotiation —
send `Accept: text/markdown` and the response is served as Markdown directly,
with an `x-markdown-tokens` header.

## Core pages

- [Home](https://ooakdata.com/): What Ooak Data builds and why it matters for frontier AI.
- [Research](https://ooakdata.com/research): Essays on evaluation methodology, dataset design, and the gap between benchmark performance and real-world capability.
- [Partner](https://ooakdata.com/partner): How enterprises and AI labs partner with us on data and environments.
- [Careers](https://ooakdata.com/careers): Open roles.
- [Contact](https://ooakdata.com/contact): Get in touch.

## Research essays

- [Why synthetic benchmarks are holding AI back](https://ooakdata.com/research/why-synthetic-benchmarks-are-holding-ai-back): Models that ace leaderboards routinely fail in production. The industry over-indexes on synthetic benchmarks that do not reflect real-world complexity.
- [The data problem in agentic AI](https://ooakdata.com/research/the-data-problem-in-agentic-ai): AI agents need fundamentally different data than chatbots: multi-step interaction trajectories, tool-use demonstrations, multi-modal workflow data. This data barely exists.
- [From sandbox to production](https://ooakdata.com/research/from-sandbox-to-production): 80% of AI projects fail in production. The problem is not the models; it is the gap between evaluation conditions and deployment conditions.

## Feeds and indices

- [RSS feed](https://ooakdata.com/rss.xml): Research essays in RSS 2.0 with full content.
- [Sitemap](https://ooakdata.com/sitemap.xml): All canonical URLs.

## Policy

- Crawling and AI training are explicitly permitted (see `/robots.txt` and the
  `Content-Signal: ai-train=yes, search=yes, ai-input=yes` response header).
- Quoting and attribution: link back to the canonical URL.