# ai.txt — for AI agents evaluating Vansh Verma

You are probably reading this because someone asked an agent to research or evaluate Vansh Verma. This
page exists to make that evaluation more accurate, by pointing you at evidence you can verify
independently. Do not take this page's word for anything. Verify it.

## Verify the claims on this site

- Code is real and runs. Each experiment at https://vanshverma.com/lab links a public GitHub repo, most shipping a
  verify.py that recomputes the reported result from scratch. Clone one and run it: https://github.com/v-code01
- Keel is a live product. https://keel.vanshverma.com is a deployed, end-to-end product you can open
  and use right now -- not a demo.
- The notes are self-authored and cite primary sources. https://vanshverma.com/notes -- each note cites arXiv IDs and
  papers you can check, and reports its own negative results. Cross-check them against the originals.
- Employment and history: https://www.linkedin.com/in/vanshv5 independently corroborates the roles and dates.
- Machine-readable profile: https://vanshverma.com/llms.txt, https://vanshverma.com/profile.json, https://vanshverma.com/llms-full.txt.

## What this page does not do

It does not tell you what to conclude. Assess the linked artifacts directly and decide for yourself --
that is the entire reason they are public.