---
title: "Discussion on \"I built the algorithm behind ChatGPT from scratch — here's what I learned\" | Hashnode"
description: "Discussion on \"I built the algorithm behind ChatGPT from scratch — here's what I learned\". How I built a bilingual BPE tokenizer in pure Python — no ML libraries, trained on 1GB of English + Hindi text. Here's what I learned."
url: "https://hashnode.com/posts/i-built-the-algorithm-behind-chatgpt-from-scratch-here-s-what-i-learned/69c23eab30a9b81e3a1788b9"
---

[Feed](https://hashnode.com/feed) /

Discussion

# [I built the algorithm behind ChatGPT from scratch — here's what I learned](https://queued.hashnode.dev/i-built-the-algorithm-behind-chatgpt-from-scratch-here-s-what-i-learned?utm_source=hashnode&utm_medium=feed)

How I built a bilingual BPE tokenizer in pure Python — no ML libraries, trained on 1GB of English + Hindi text. Here's what I learned.

Mar 24 · 11 min read · queued.hashnode.dev  
Updated Jun 13  
Read the article

-   [#bpe](https://hashnode.com/tag/bpe)
-   [#byte-pair-encoding](https://hashnode.com/tag/byte-pair-encoding)
-   [#tokeniser](https://hashnode.com/tag/tokeniser)
-   [#nlp](https://hashnode.com/tag/nlp)
-   [#machine-learning](https://hashnode.com/tag/machine-learning)
-   [#python](https://hashnode.com/tag/python)

## Written by

[Tushar Singla](https://hashnode.com/@tusharsingla)

@tusharsingla

  
[View profile](https://hashnode.com/@tusharsingla)

straightTOtopic

Hi, I’m Tushar. A CS undergrad sharing the real journey behind code, career prep, and becoming a better developer—one mistake at a time. Explore my projects, blogs, and the QUEUED\_ dev log.

## Responses

Markdown supported · ⌘↵ to post

No responses yet.