# 2505 Labs

> UK research laboratory for AI and robotics. We build reliable, useful, and accountable systems — and publish what we learn.

2505 Labs researches and ships systems at the frontier of intelligence and the physical world, with care for reliability, usefulness, and long-term benefit.

## Sites

- [2505 Labs](https://2505.in): Research laboratory homepage — approach, products, company, research papers, blog
- [Nomos](https://nomos.wtf): Agentic software factory built on our DADG research — repo-scoped agents, codebase memory, multi-file execution
- [Research paper — DADG / ASG](https://2505.in/papers/asg.pdf): Dynamic Abstract Syntax Tree Graphing for Infinite-Context Agentic Coding (Roshan Raghavander, Naren Kumar)

## Pages

- [Home](https://2505.in/): Mission and overview
- [Research](https://2505.in/research): Research papers
- [Blog](https://2505.in/blog): Notes from the lab
- [Company](https://2505.in/company): Purpose, team, values
- [Contact](mailto:hello@2505labs.com): hello@2505labs.com

## Products

- **Nomos** (live): AI software factory. Powered by Dynamic AST Dependency Graph (DADG) retrieval research.
- **Kaz** (coming soon): Banking infrastructure for AI agents — spend controls, settlement, audit.

## Approach

1. Capability with constraint — raise agent ability inside real engineering limits
2. Autonomy with accountability — permissions, measurable outcomes, human authority
3. Embodiment with care — robotics where safety, robustness, and usefulness hold together

## Optional

- [GitHub](https://github.com/2505-lab)
- [X / Twitter](https://twitter.com/2505lab)
- [LinkedIn](https://linkedin.com/company/2505-lab)
- [Sitemap](https://2505.in/sitemap.xml)
- [robots.txt](https://2505.in/robots.txt)