mach-e

Andrej Karpathy builds a GPT from scratch in code, following Attention Is All You Need and GPT-2/3, in about two hours using PyTorch. He connects it to ChatGPT and assumes familiarity with his earlier makemore videos.

Jeremy Howard's 90-minute practical tour of language models. It covers how they work, GPT-4's capabilities and limits, using the OpenAI API, code writing, data analysis and fine-tuning.

Martin Gorner's Devoxx talk introduces deep learning with TensorFlow, walking through neural networks for MNIST, how to choose an architecture and how to diagnose networks that fail to train.

Andrej Karpathy builds a bigram character-level language model from scratch, introducing torch.Tensor and the framework of training, sampling and loss evaluation. It is the first step toward a GPT-style Transformer in his makemore series.

David Beazley's PyCon 2019 tutorial derives the lambda calculus from the ground up in Python, using only single-argument functions. He builds booleans, numerals and more with no other language features.

A roughly 90-minute film of John Dobson building a Dobsonian telescope step by step, including hand-grinding the mirror.

Tsoding live-codes a JSON parser from scratch in Haskell in about 111 lines, using only the standard library. It is an edited Twitch stream that explains the intuition behind parser combinators, functors, applicatives and monads.

A long livestream of George Hotz live-coding a toy SLAM implementation, twitchslam, over two days in May 2018. It is an unedited session of Python programming with computer vision and visualization.

In a long live-coding session, Andreas Kling works on making the Cookie Clicker game playable in the Ladybird browser from SerenityOS. He tracks down bugs in the browser engine and fixes them.

Andreas Kling live-codes a prototype JavaScript JIT compiler for the LibJS engine used in SerenityOS and Ladybird. He builds a standalone bytecode VM, a simple interpreter, and then a simple JIT in about 80 minutes.

David Beazley's workshop screencast builds asynchronous programming in Python from scratch, starting from callbacks and continuations and replacing threads to show how concurrency works underneath.