Gayathri Seshadri·Oct 27“I Only Love You” vs “I Love Only You”: Why Word Order Breaks AI (and How Positional Embeddings Fix…Same words. Same letters. Completely different meaning. Here’s the tiny trick that teaches ChatGPT-style models the difference.
Gayathri Seshadri·Oct 21Why Every Engineer Starting Out in AI Should Understand TokenizationFive practical ways the first step of an LLM affects cost, product quality and your interviews
Gayathri Seshadri·Oct 16How LLMs Learn From Raw Text: Building Input–Target Pairs with a Sliding WindowPart of a series on building a large language model from scratch. You can read this post on its own: every term is explained as it comes…
Gayathri Seshadri·Oct 12Why Your LLM Doesn’t Read Words: A Plain-English Guide to Token EmbeddingsComputers only understand numbers. Here’s how large language models turn words into numbers that actually carry meaning.
Gayathri Seshadri·Oct 6Byte Pair Encoding: How GPT Models Read TextWhy GPT doesn’t read words or letters, but something in between.
InStackademicbyGayathri Seshadri·Sep 29LLM Tokenization, Explained From ScratchIf you do not have Medium paid version here is the Friends Link.
InStackademicbyGayathri Seshadri·Sep 16What Are Transformers? A Beginner’s Guide to the Architecture Behind ChatGPTHere is a Friends link if you have not subscribed to Medium.
InStackademicbyGayathri Seshadri·Jul 13Beyond Search-and-Replace: 5 Surprising Ways OpenRewrite is Automating the Death of Technical DebtThe 15-Year Migration Nightmare
InStackademicbyGayathri Seshadri·Jul 6From Demo to Disaster: 5 Hard Truths About Scaling AI Agents in 2026Here is a Friends link if you have not subscribed to Medium.
InStackademicbyGayathri Seshadri·Jun 19The Evolution of Autonomy5 Impactful Truths About the New Era of AI Agents