---
title: "Michał Z (@sixfingers) | Hashnode"
description: "Exploring AI, Big Data, machine learning, NLP, and linguistics. Read the latest articles by Michał Z on Hashnode."
url: "https://hashnode.com/@sixfingers"
---

@sixfingers

# Michał Z

@sixfingers · Warsaw, Poland · Joined November 2026

Exploring AI, Big Data, machine learning, NLP, and linguistics

[Website](https://huggingface.co/miczac)

### [How to Build a Large Language Model from Scratch? Toktumi 3.0 – Part 3 (The Final Part)](https://sixfingers.hashnode.dev/toktumi-3-en?utm_source=hashnode&utm_medium=feed)

Toktumi 3.0 – a Polish-language LLM based on GPT-2. Fine-tuning, 100% synthetic training data. Conclusion: LESS CAN BE MORE.

34 min read · [sixfingers.hashnode.dev](https://sixfingers.hashnode.dev/toktumi-3-en?utm_source=hashnode&utm_medium=feed)

[8h ago](https://hashnode.com/posts/toktumi-3-en/6af4da9a63e3e8f3d91d5cde)

### [How to Build a Large Language Model from Scratch? Toktumi 2.0 – Part 2/3](https://sixfingers.hashnode.dev/toktumi-2-en?utm_source=hashnode&utm_medium=feed)

Toktumi 2.0 – a Polish-language LLM based on GPT-2. How to learn from the first training and do it better.

24 min read · [sixfingers.hashnode.dev](https://sixfingers.hashnode.dev/toktumi-2-en?utm_source=hashnode&utm_medium=feed)

[8h ago](https://hashnode.com/posts/toktumi-2-en/6af4da49e9132b8921350837)

### [How to Build a Large Language Model from Scratch? Toktumi 1.0 – Part 1/3](https://sixfingers.hashnode.dev/toktumi-1-en?utm_source=hashnode&utm_medium=feed)

Toktumi 1.0 – a Polish-language LLM based on GPT-2. From web scraping to a working model.

27 min read · [sixfingers.hashnode.dev](https://sixfingers.hashnode.dev/toktumi-1-en?utm_source=hashnode&utm_medium=feed)

[9h ago](https://hashnode.com/posts/toktumi-1-en/6af4d9f763e3e8f3d91c7e69)

### [Jak zbudować duży model językowy od zera? Toktumi 3.0 – Część 3 (i ostatnia)](https://sixfingers.hashnode.dev/toktumi-3-pl?utm_source=hashnode&utm_medium=feed)

Toktumi 3.0 – polski model LLM oparty na GPT-2. Fine-tuning, 100% danych syntetycznych i jeszcze lepsze wyniki.

25 min read · [sixfingers.hashnode.dev](https://sixfingers.hashnode.dev/toktumi-3-pl?utm_source=hashnode&utm_medium=feed)

[9h ago](https://hashnode.com/posts/toktumi-3-pl/6af4d978fbcb453115f6282b)

### [Jak zbudować duży model językowy od zera? Toktumi 2.0 – Część 2/3](https://sixfingers.hashnode.dev/toktumi-2-pl?utm_source=hashnode&utm_medium=feed)

Toktumi 2.0 – polski model LLM oparty na GPT-2. Jak wyciągnąć wnioski z pierwszego treningu i zrobić to lepiej.

19 min read · [sixfingers.hashnode.dev](https://sixfingers.hashnode.dev/toktumi-2-pl?utm_source=hashnode&utm_medium=feed)

[9h ago](https://hashnode.com/posts/toktumi-2-pl/6af4d8bde9132b892131277e)