---
title: "Mohitm (@mohitm) | Hashnode"
description: "Data science professional with around 15 years experience, working to on Pricing and Media analytics practice for CPG companies. Currently building a micro-l..."
url: "https://hashnode.com/@mohitm"
---

@mohitm

# Mohitm

@mohitm · Bangalore, India · Joined August 2026

## Comments

[Most modern LLMs correct or bypass the "strawberry" counting issue through specific engineering changes, prompt workarounds, or execution tools rather than a fundamental leap in character-level perception.](https://hashnode.com/posts/why-your-llm-can-t-count-the-r-s-in-strawberry/6ae15bd87700189050d03c0f/comment/6ae170774ace2d609e0d8189)

Comment · Article · Oct 28 · [Why Your LLM Can't Count the R's in 'Strawberry'](https://hashnode.com/posts/why-your-llm-can-t-count-the-r-s-in-strawberry/6ae15bd87700189050d03c0f)

[you're right that "magnitude carries no meaning" is too broad. It holds for the static lookup table under cosine similarity, but whether norm matters really depends on the training objective and similarity function, and the retrieval case proves it: L2-normalizing a dot-product-trained representation can reorder results, not just speed up the same ranking.](https://hashnode.com/posts/embedding-is-just-a-vector/6ac31c2ce5fdec84e1b8951b/comment/6ac39ab4147f64827744320c#reply-6ac47c4c38f2a597f7e9dd95)

Reply · Article · Oct 6 · [Embedding is just a vector](https://hashnode.com/posts/embedding-is-just-a-vector/6ac31c2ce5fdec84e1b8951b)