<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="3.10.0">Jekyll</generator><link href="https://samitmohan.github.io/feed.xml" rel="self" type="application/atom+xml" /><link href="https://samitmohan.github.io/" rel="alternate" type="text/html" /><updated>2026-04-03T16:58:51+00:00</updated><id>https://samitmohan.github.io/feed.xml</id><title type="html">samit</title><subtitle>my home on the internet</subtitle><author><name>Samit Mohan</name></author><entry><title type="html">the annotated microgpt</title><link href="https://samitmohan.github.io/tech/2026/04/02/microgpt.html" rel="alternate" type="text/html" title="the annotated microgpt" /><published>2026-04-02T06:30:00+00:00</published><updated>2026-04-02T06:30:00+00:00</updated><id>https://samitmohan.github.io/tech/2026/04/02/microgpt</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[Karpathy's GPT from scratch - 340 lines of pure Python, annotated from autograd to inference. No frameworks. No magic.]]></summary></entry><entry><title type="html">what happens when you press ‘submit’ on chatgpt</title><link href="https://samitmohan.github.io/tech/2026/03/26/how-chatgpt-works.html" rel="alternate" type="text/html" title="what happens when you press ‘submit’ on chatgpt" /><published>2026-03-26T18:30:00+00:00</published><updated>2026-03-26T18:30:00+00:00</updated><id>https://samitmohan.github.io/tech/2026/03/26/how-chatgpt-works</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[from raw internet text to a streaming response in your browser - pretraining, alignment, inference, and everything in between]]></summary></entry><entry><title type="html">from residual connections to attention residuals</title><link href="https://samitmohan.github.io/tech/2026/03/18/attention-residuals.html" rel="alternate" type="text/html" title="from residual connections to attention residuals" /><published>2026-03-18T12:30:00+00:00</published><updated>2026-03-18T12:30:00+00:00</updated><id>https://samitmohan.github.io/tech/2026/03/18/attention-residuals</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[The residual connection solved deep learning in 2015. Ten years later, Moonshot AI noticed it's been sabotaging deep networks the whole time. The fix is 30 lines of PyTorch.]]></summary></entry><entry><title type="html">how to use claude</title><link href="https://samitmohan.github.io/tech/2026/03/18/claude.html" rel="alternate" type="text/html" title="how to use claude" /><published>2026-03-18T08:36:04+00:00</published><updated>2026-03-18T08:36:04+00:00</updated><id>https://samitmohan.github.io/tech/2026/03/18/claude</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[most people use claude code like a chatbot. here's how to actually drive it.]]></summary></entry><entry><title type="html">building pytorch from scratch</title><link href="https://samitmohan.github.io/tech/2026/03/11/minitorch.html" rel="alternate" type="text/html" title="building pytorch from scratch" /><published>2026-03-11T08:36:04+00:00</published><updated>2026-03-11T08:36:04+00:00</updated><id>https://samitmohan.github.io/tech/2026/03/11/minitorch</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[Building pytorch from scratch in ~1300 lines - reverse-mode autograd, conv2d, optimizers. Trains MNIST. No C++, no CUDA, just closures and numpy.]]></summary></entry><entry><title type="html">favourite interview questions</title><link href="https://samitmohan.github.io/tech/2026/02/13/interview-questions.html" rel="alternate" type="text/html" title="favourite interview questions" /><published>2026-02-13T08:36:04+00:00</published><updated>2026-02-13T08:36:04+00:00</updated><id>https://samitmohan.github.io/tech/2026/02/13/interview-questions</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[Five questions I'd rather ask than 'derive Kadane's algorithm in 30 minutes pretending you've never seen it.']]></summary></entry><entry><title type="html">first failure of 2026: solving 100+ deep learning questions</title><link href="https://samitmohan.github.io/fun/2026/01/30/first-failure.html" rel="alternate" type="text/html" title="first failure of 2026: solving 100+ deep learning questions" /><published>2026-01-30T08:36:04+00:00</published><updated>2026-01-30T08:36:04+00:00</updated><id>https://samitmohan.github.io/fun/2026/01/30/first-failure</id><author><name>Samit Mohan</name></author><category term="fun" /><summary type="html"><![CDATA[Made it 20 days into my '1 ML problem per day' resolution before a concert in bombay killed my streak. Still solved 100+ though.]]></summary></entry><entry><title type="html">building rag for my website</title><link href="https://samitmohan.github.io/tech/2026/01/27/rag.html" rel="alternate" type="text/html" title="building rag for my website" /><published>2026-01-27T08:36:04+00:00</published><updated>2026-01-27T08:36:04+00:00</updated><id>https://samitmohan.github.io/tech/2026/01/27/rag</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[My blogs are too long so I built a RAG over them. Chunking, FAISS, reranking, streaming via Groq - and an eval pipeline because vibes aren't metrics.]]></summary></entry><entry><title type="html">all the math you need for ai</title><link href="https://samitmohan.github.io/tech/2026/01/21/math.html" rel="alternate" type="text/html" title="all the math you need for ai" /><published>2026-01-21T08:36:04+00:00</published><updated>2026-01-21T08:36:04+00:00</updated><id>https://samitmohan.github.io/tech/2026/01/21/math</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[Every derivative, gradient, and matrix operation you need to understand deep learning - with code alongside every concept.]]></summary></entry><entry><title type="html">numpy &amp;amp; pytorch for dummies</title><link href="https://samitmohan.github.io/tech/2026/01/07/torch.html" rel="alternate" type="text/html" title="numpy &amp;amp; pytorch for dummies" /><published>2026-01-07T08:36:04+00:00</published><updated>2026-01-07T08:36:04+00:00</updated><id>https://samitmohan.github.io/tech/2026/01/07/torch</id><author><name>Samit Mohan</name></author><category term="tech" /><summary type="html"><![CDATA[Ground-up numpy and pytorch - we build MNIST classifiers two ways to see how the pieces fit together.]]></summary></entry></feed>