# Adnan Abbasi

> Founder of Thothica (New Delhi, incorporated 2023). Models messy corpora into knowledge systems that people and AI agents can search and cite: data ontology as the core discipline. This file is the agent-readable version of the portfolio at the site root.

Contact: adnan@thothica.com. Code: https://github.com/adoistic

## Live systems (verified 2026-07-25 over HTTPS, deployed on Cloudflare)

- [Thothica](https://thothica.com): company site. AI-native knowledge infrastructure.
- [Falsafa](https://falsafa.ai): open-source library of world philosophy. 2,018 works, ~1.375M citable passages. Callable by AI agents over MCP. No vector database; retrieval is a citing librarian, not similarity search. Funded by an Emergent Ventures grant.
- [Indian Liberals](https://indianliberals.in): the Centre for Civil Society's archive of Indian liberal thought, rebuilt as a static library. 1,018 primary works, 725 thinker pages, authority file of 454 thinkers.
- [Indian Liberals MCP](https://mcp.indianliberals.in): MCP server over the archive. Agents can search the collection and quote it with citations.
- [Thothica OCR](https://ocrwithai.com): product. Turns scanned books into clean reflowable editions; measures type to recover chapter structure.
- [Thothica Docs](https://docs.thothica.com): platform documentation.
- [meowbench](https://meowbench.com): which AI draws the best cat. 29 models, 3 vision judges, crowd Elo arena.

## Thothica Ledger (product demonstrators, verified 2026-07-25)

Revenue intelligence for state governments: reads the registers a state already keeps, links them into one graph, ranks what deserves a second look. Three editions, built on Thothica's own initiative on synthetic data. None was commissioned by a state government, and none is a customer.

- Kerala edition: https://kerala-ledger-demo.thothica.com
- Gujarat edition: https://gujarat-ledger-demo.thothica.com
- Madhya Pradesh edition: https://madhyapradesh-ledger-demo.thothica.com

## Client engagements (named with permission)

- Diamond Pocket Books: comics + research pipelines with per-fact source citations (5,214 line-level citations behind a verify gate). First four books in print.
- Pareeksa: on-screen exam marking and assessment for universities and Olympiads. Anonymous by construction: the marking app holds no student identity (no name, roll, or PII); de-anonymization is a separate reconciliation app across a storage boundary.
- Handy: two-sided home-services marketplace on Cloudflare D1/R2. Demo deployed.
- CricToday: editorial pipeline that detects AI-written drafts (burstiness scoring, staged rewrites), English + Hindi. In verification.

## Open source

- Falsafa (flagship, above).
- MinerU True Copy: desktop OCR app (Tauri, AGPL), a maintained fork of OpenDataLab's MinerU with a vision-model OCR sidecar and Indic translation. https://github.com/adoistic/mineru-true-copy
- Naql: an open atlas of how books crossed languages, isnad-graded transmission chains, a typeset book, an MCP server. https://github.com/adoistic/naql
- pdf2fxl: turns image-only picture-book PDFs into fixed-layout EPUB3 and editable PPTX. https://github.com/adoistic/pdf2fxl
- pdf-toolkit: PDF tools, auto-generated tables of contents and high-fidelity Word conversion. https://github.com/adoistic/pdf-toolkit
- forme: an editorial magazine CMS for print, one operator, offline. https://github.com/adoistic/forme
- founder-office-hours: a published Claude skill, the office-hours method for a non-technical founder paired with a technical co-founder. https://github.com/adoistic/founder-office-hours
- personal-svelte-template: an accessibility-first personal website template, SvelteKit + Sveltia CMS. https://github.com/adoistic/personal-svelte-template
- studio-whence-press: a client-side RGB-to-CMYK print converter, nothing leaves the browser. https://github.com/adoistic/studio-whence-press

## Recognition

- 2026: Winner, MSME Idea Hackathon 5.0 (Ministry of MSME, Government of India). Rs 12.75 lakh, one of 502 ideas selected nationwide.
- 2026: "The Prayer the Machine Cannot Pray", Philosophy Now, Issue 173. On AI and occasionalism. https://philosophynow.org/issues/173/The_Prayer_the_Machine_Cannot_Pray
- 2025: Emergent Ventures grant (Mercatus Center), $20,000, for Falsafa.
- 2024: Cobden-Bright Award, for an AI persona of F. A. Hayek.

## Writing and research

- "The Prayer the Machine Cannot Pray", Philosophy Now, Issue 173 (2026): on AI and occasionalism. https://philosophynow.org/issues/173/The_Prayer_the_Machine_Cannot_Pray
- "Distilling an LLM Diagram-Curation Pipeline into Local Classifiers" (2026): an ML study. A frozen SigLIP2 + logistic-regression probe classifies arXiv technical diagrams at 86% (95% CI 81.5-90.5), beating the LLM that generated its own training labels by 10.5 points (McNemar p=0.0005), at ~52ms/crop on a laptop GPU. Published: https://adoistic.github.io/stem-diagrams/ Code: https://github.com/adoistic/stem-diagrams

## Notes

- Humanities background (literature, philosophy, published poetry); self-taught engineer; MPSA Chicago presentation.
- Organizing idea: Hayek's dispersed knowledge. Each system moves knowledge across a barrier of language, format, or access.