Package Development nbdev let's you develop and publish Python packages right from Jupyter Notebook. It generates documentation and publishes on GitHub Pages. You can also write tests and setup CI with GitHub Actions. github.com/fastai/nbdev

It really took me a while to "get it" when it comes to nbdev. But I gotta hand it to @jeremyphoward this way of working makes too much sense once you're used to it. As of today, I am working on tools that make this kind of work possible in @marimo_io. youtu.be/ZLg27UmAJbw

Andrej Karpathy → Founding member of OpenAI and the former Director of AI at Tesla. His "Zero to Hero" series is legendary. Posts once a year but that’s enough to create ripples in the AI industry. (Note: He taught Stanford's first Deep Learning course (CS231n), which trained the first generation of modern AI engineers) GenAI Explained → Turning complex AI breakthroughs into short, practical, and easy to understand videos that help you discover what matters, learn how it works and grow your skills in real-time. Backed by a global community of 13M builders and researchers Stanford Online → Ivy League education for free. Famous CS224N (Natural Language Processing) and CS231n lectures. Watch the world’s leading researchers (like Li Fei-Fei and Christopher Manning) deliver the same content their students pay thousands for. Serrano Academy → Luis Serrano (PhD in Math and Author) simplifies complex topics through hand-drawn cartoons and analogies. Learn attention mechanisms and LLMs with a friendly and high-energy vibe. Jeremy Howard → Top-down learning. Build the app first, figure out the math later. Teaches practical deep learning for coders. No PhD required. 3Blue1Brown → Visual poetry. Grant Sanderson uses his own custom animation engine (Manim) to visualize neural networks, backpropagation and linear algebra. He made the Transformer architecture (the T in ChatGPT) visually understandable for the first time. Hamel Husain → Learn the dirty work of AI: evaluations (evals), data curation and deployment. He is the Engineer's Engineer. Former Lead ML Engineer at GitHub and Airbnb and a major contributor to open-source tools like nbdev. Dave Ebbelaar → King of RAG (Retrieval-Augmented Generation). While others talk about models, he talks about pipelines and how to connect your private data (PDFs, Notion, SQL) to an LLM safely. Machine Learning Street Talk → Deeply technical and unedited. Interviews with AI rebels like Yann LeCun (former Meta's AI chief), who challenge the current direction of LLMs. Critical, non-hype perspective. Lex Fridman → Introspective interviews with everyone who matters in AI: Sam Altman, Demis Hassabis, Yann LeCun, Elon Musk. Vision and history of where the field is going.

My coding tools and infrastructure looks drastically different than 6 months ago. I used to optimize towards what fits my mental model. Now I use what the agents want. It doesn't make sense to swim upstream anymore. This is especially acute with niche tools (like nbdev)

I've been using nbdev since I found about it about a year ago. Great devX, you should try it out.

With the release of nbdev3 yesterday, now's a great time to discover why YOU should be using nbdev - and there's no better tldr than this thread that introduced nbdev2 in 2022, so take a look!:

nbdev v3 is released! The big change is that nbdev now uses pyproject.toml, not settings.ini, for all config. (nbdev v1 predates pyproject's support for project metadata, so we used our own settings file.) v3's `nbdev_migrate_config` automates migration nbdev.fast.ai/getting_starte…

This matches my experience. I used to do ~ 100% of my work in nbs and nbdev. Now it's pretty rare. I still use nbs some, but the scope is extremely limited, and so I don't use nbdev anymore. Or notebook environments in general for most things.

Yeah. I kind of use notebooks sometimes with claude code, which is ok for viz/data work. It kinda sorta is able to modify the json. However, nbdev is yet another level of friction (that made it untenable for most of my work).

💯 nbdev was beautiful while it served its purpose. Now can adopt tools/models/languages that agents are familiar with. The wall of languages and frameworks is fallen. My slight worry is that people will try to make do with what works best with AI and not innovate as much.

Wrote a blog about about why I no longer use nbdev. This is a story about: - Checking your ego to allow for what works best with AI - Our relationship with tools - The increasing importance of collaboration with others (and AI) - Being polyglot hamel.dev/blog/posts/ai-stac…

Yeah nbdev has a really cool way of converting ipynb to a bunch of different useful things: github.com/AnswerDotAI/nbdev

This is also why I can’t use nbdev anymore There isn’t a good **open source** ai Jupyter integration I really like (many exist but they all feel second class) So I’ve given up on using it to build software and am using more paved paths

Why you should write the code to use your Thermostat Yesterday, I tried following @karpathy 's lead and use CC to interface with my thermostat. The experience wasn't great for me. Today I tried the opposite, I built it step-by-step almost completely manually using @answerdotai 's SolveIT. It was great. I spent a much longer time, say maybe 4-5h. But it was worth it, even though I have a perfectly fine app already to handle the temperature (in other words, I might not reuse the code again). So, was worth it? Yes, why? Because there were a myriad small decisions and lessons learned throughout the process. Those are usually small enough and that don't feel significant but they do compound in the end. They make your tools much sharper. If you look at the final package I published to pypi you won't see that, and it looks like code an LLM could one-shot, the LLM could definitely NOT make you learn during the process. This is more or less how it went: Well, I started getting SolveIT to read the docs for me and list which endpoints does the Thermostat API have. Unfortunately, their docs require JS to render 😅. I took this chance to have a look a Zyte service to give SolveIT a tool to read the docs. With that setup I got a list of all endpoints for the API. I followed the instructions to get the API keys & tokens. Gave it a try, it worked, I went away, and when I came back the token had expired. No problem, solveIT created a super short _refresh method as part of the class. Next step, create the basic "Home Status" info endpoint. But call _refresh first to make sure we have an updated token... I thought we would have to do this for all endpoints, so let's actually create a `_request` that does this for us already. That way the methods simply look like this: At this point I took the chance to practice using dialoghelper. I had created two endpoints following a clear format: first markdown header, @ patch method into the class, try it out and display results. Because the structure was super clear, and the full list of endpoints available already, SolveIT could add each one of the endpoints in the exact format I liked, fast & straightforward. At this point, when I finished, it felt silly, why did I spend time doing this at all, it was done so fast! but it was fast mostly because everything was perfectly set up. I was reminded of @johnowhitaker said during the course: " ... using SolveIT, we spend most of our time sharpening our tools..." and that felt very true, once your tools are sharp and set up for the task at hand, the task was super fast to finish. Thanks to all this I learned things I wouldn't otherwise: - using Zyte to scrape (& having it ready to reuse as an LLM tool) - best way to design this API (I decided to name the methods exactly like the endpoints to avoid cognitive overload) - how easy it is to actually control my thermostat, and having a package to quickly interface with it - got better a publishing w/ nbdev - learned about _proc folder The API point was an important one, I tried many different ways to structure the endpoint calls, for example saving the home_id into the class, or using the first one by default, etc... The time in-between coding sessions, I could feel my brain in the background considering different approaches. I ended settling on mimicking the endpoints as close as possible, and letting the user built on top. I feel like a better engineer & API designer as a result of this. This won't happen if CC creates everything for you. There's the expression "Death by a thousand paper cuts" which describes how numerous small, seemingly insignificant problems accumulate over time to cause major failure The opposite happens with this approach: countless small decisions and lessons that felt insignificant on their own, but compounded into something valuable. @jeremyphoward often says to treat everything as a learning opportunity, and this definitely feels like it.

Package Development nbdev let's you develop and publish Python packages right from Jupyter Notebook. It generates documentation and publishes on GitHub Pages. You can also write tests and setup CI with GitHub Actions. github.com/fastai/nbdev

Donald Knuth, @jeremyphoward and nbdev would approve :D

Package Development nbdev let's you develop and publish Python packages right from Jupyter Notebook. It generates documentation and publishes on GitHub Pages. You can also write tests and setup CI with GitHub Actions. github.com/fastai/nbdev

【WebにLLM向けの入口を作るllms.txtの解説】 llms.txtはWebサイト直下の/llms.txtに置くMarkdownで、LLMやAIエージェントが要点に素早く到達できるよう最小構成を示す。H1見出しと要約の引用、補足説明、H2配下のファイルリストという順序を定め、必要なら各ページの“.html.md”版を提供する方針も含む。2024年9月3日に公開された提案で、開発ドキュメントやAPIの参照を効率化する狙い。 実装面では、FastHTMLがllms.txtを基にllms-ctx.txt/llms-ctx-full.txtを生成して活用している。nbdevは全ページの.md生成を標準化し、仕様の導入を容易にした。用途は推論時の参照が中心で、robots.txtやsitemap.xmlと競合せず補完関係にある。 llmstxt.siteは公開ディレクトリとして、AnthropicやPinecone、Klarnaなどのllms.txtを横断的に探せる。llms.txtは、サイト運営者がLLMへの提示情報を意図通りに整理し、ユーザーの質問文脈に合わせてAIが適切に深掘りできる状態を用意するものである。

A lot of appreciation to `nbdev`, `nbconvert`, and the @fastdotai @answerdotai teams for inspiration!

Package Development nbdev let's you develop and publish Python packages right from Jupyter Notebook. It generates documentation and publishes on GitHub Pages. You can also write tests and setup CI with GitHub Actions. github.com/fastai/nbdev