
Project अर्थ (Artha) aims to make the world’s best Sanskrit AI.
- ValuesOpen access, preservation of heritage, and knowledge that remains available beyond any one platform.
- VisionBuild the best AI for Sanskrit: useful to scholars, students, builders, and curious readers.
- MethodPair cultural fluency with rigorous engineering. Treat irregularity as information, not noise.
- Public workPublish datasets, benchmarks, interfaces, and the evidence required to question them.
- ScaleStart with Sanskrit; develop patterns that help other complex, under-resourced knowledge systems.



Artha/Bench-1
Can an AI coding harness structure a thousand irregular Sanskrit books?
View benchmark ↗
Artha/Texts
Browse a structured Sanskrit library from works to chapters to individual verses.
Browse texts ↗
794 Sanskrit books
What a four-million-record agent failure taught us about delivery contracts.
Read the field note ↗
Verse 2.16
Sanskrit, transliteration, grammar, translation, concepts, and attributed commentary.
Read the verse ↗
Texts, ideas, arguments, and evidence.
Every page has one clear job: preserve a source, explain a term, stage a disagreement, or show the work behind the system.
Read
Move from the full library into a work, chapter, verse, or curated episode.
Understand
Follow technical language and large concepts back to passages and traditions.
Research
See benchmark results, experiment reports, and the engineering notes behind the corpus.
Build
Technical surfaces for stable addressing, reproducibility, and future developer tools.