hishiryo

The Self-Evolving Multi-Model Memory Loop (SEMML), introduced in this paper, is a proposed approach to machine translation in which several large language models work with a shared collection of reviewed bilingual resources to create a recurring, fully traceable process that evolves with every request. Read the full paper here

In May 2026, I was invited to speak on “AI for Texts: Multilingual Translation and Interpretation” at the International Conference on Vedānta in Stony Brook, New York. A recording of the talk is available here

Since OpenAI will soon be retiring their CustomGPTs in favor of “plugins” I'm deliberately choosing not to adjust any of the GPTs and databases I have created. Therefore, I will deactivate all existing CustomGPTs on OpenAI and may consider developing them for “platform-independent” systems in the future. This proves once again that it is better not to rely on a particular ecosystem. This affects the following (custom) GPTs: Shankara Vedanta, SanGPT, VivaranaGPT, and ಸರ್ವತ್ರ - Sarvatra.

I apologize in advance and would like to thank everyone who, I hope has used them effectively.

BhashyaBrowser is a new study tool for navigating Śaṅkara’s Bhāṣya corpus in structured IAST format. It offers tree browsing, exact and fuzzy search, and concordance tools for source-driven Vedānta research.

Updated for use with minimum GPT-5.5 Thinking (standard) with access to a vast background corpus beyond normal OpenAI Custom GPT limitations.

A source-driven GPT for studying and defending Pure Adi Shankara Vedanta through the lens of Sri Sri Satchidanandendra Saraswati, with critical analysis of post-Shankara Advaita, Bhamati, Vivarana, Neo-Vedanta, and Neo-Advaita.

Academic Sanskrit–English translator combining streaming LLM intelligence with local lexical databases, Pāṇinian morphological verification, sandhi processing, and offline transliteration. Features five prompt profiles, samāsa analysis, padapāṭha display, and confidence scoring.

AI-powered OCR and translation studio for Windows. Combines image viewing, vision-model-based text recognition, and translation in a single workflow. Supports PDF rendering, multi-page navigation, image processing, and token tracking.

Command-line tool for searching Sanskrit texts on the Advaita Sharada corpus. Accepts Devanagari, IAST, or Roman input with automatic script conversion. Covers Bhāṣya, Prakaraṇa, and Vyākhyā collections.

A custom GPT built on OpenAI base models and enhanced with retrieval-augmented generation (RAG) over a curated translation-memory knowledge base.