Chapter 3: Linear Methods for Regression
Linear regression is one of the oldest and most important statistical learning methods. The linear model makes strong assumptions about the structure of the data, and when these assumptions are met...
We discuss least squares fitting, subset selection, shrinkage methods, and methods using derived input features to improve the linear model's performance.
Chapter 2: Meaningful Names
Writing clean code is a skill that separates the professional from the amateur. Clean code is not written by following a set of rules. You don't become a software craftsman by learning a list of heuristics...
This book teaches you how to write code that is expressive, readable, and maintainable. You'll learn the difference between good and bad code, and how to transform bad code into good code.
Abstract
The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also...
We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely.
Executive Summary
As AI systems become more powerful and widespread, ensuring their safety and alignment with human values becomes increasingly critical. This report examines current risks...
We analyze governance frameworks, technical safety measures, and international cooperation efforts to mitigate potential harms from advanced AI systems.
Introduction to Deep Learning
Deep learning has revolutionized artificial intelligence by enabling machines to learn complex patterns from data. Neural networks with multiple hidden layers can approximate any function...
We'll explore convolutional networks for vision, recurrent networks for sequences, and transformer architectures that power modern AI systems.
Your files know things
together that they
can't know alone.
Not another file manager. FileGrind connects files across formats.
Every app handles one file type. FileGrind breaks all your files down and finds how they connect. Which pages across all your books talk about the same concept? Which documents cluster by topic? Navigate your knowledge, not your folders. Your files stay on your Mac.
We'll email you when it's ready.
Break down barriers between files
PDFs don't talk to EPUBs. Your research papers don't know about your notes. FileGrind grinds everything down to structured pieces—chips—then shows you how they connect. Find clusters. Map concepts. See which pages across 50 books discuss the same idea. This wasn't possible before.
Not a PDF reader. A file grinder.
FileGrind doesn't replace your PDF app or ebook reader. It breaks files down—any file type—into structured pieces called chips. Launching with PDFs and EPUBs. More formats coming. Each chip is searchable. Linkable. Ready to connect with chips from other files.
Mix chips from anywhere
A block is chips from one file or many, organized by your schema. Pull page 45 from a PDF, Chapter 3 from an EPUB, your notes from a text file. Now they're one thing. Query it. Export it. Use it. This is the insight files couldn't give you separately.
Caps: What you can do to files
A capability—cap—is an action you can perform on your files. Extract text from a PDF? That's a cap. Classify an image? Cap. Parse 3D model metadata? Also a cap. Drag a cap onto any content. It runs. You get structured output you can search and organize. FileGrind ships with essential caps built-in. The community builds specialized ones.
Search ignores file boundaries
One search. All your files. Every format. Find where concepts appear across your entire library, regardless of file type. Semantic search clusters related ideas even when the words differ. Navigate knowledge, not folders.
The right model for the job
General-purpose LLMs are great. Specialized models are better. Search, download, and run task-specific models on your content. Translate chapters. Proofread paragraphs. Find every giraffe image across PDFs and PNGs. Latest open-source models. Running locally. On exactly the pieces you choose.
Browse. Download. Run.
FileGrind finds compatible models from HuggingFace and other sources. One click to download. Runs on your hardware. Apply to a single chip, a whole block, or your entire library. No API keys. No usage limits. No sending your content anywhere.
More file types. More capabilities. Always growing.
We're launching in beta with PDFs, EPUBs, text files, and images. But FileGrind isn't limited to these. It's designed to handle any file type through caps. Need 3D models? Audio files? Code analysis? New caps get added by the community. Install what you need. Or build and share your own for specialized workflows. FileGrind grows with its users.
Not Blender. Not Photoshop. A bridge between file types.
FileGrind doesn't replace specialized tools. It sits between them. It's not Blender, but it can analyze 3D models and extract properties across your collection. It's not a DAW, but it can find patterns across audio files. It's not an image editor, but it can classify and label thousands of images automatically. Then it gives you that data. For research. For organizing. For insights no single app could provide.
Opens files in whatever app you use
Want to read that PDF? FileGrind opens it in Preview, Acrobat, or whatever you've set. Edit a 3D model? Opens in Blender. FileGrind finds the connections. The specialized apps do what they do best. You move between them seamlessly.
Your files. Your Mac. That's it.
No cloud uploads. No API calls with your documents. No "processing on our servers." Every LLM runs locally using MLX. Search happens on your machine. Your library stays yours. Unplug your internet—FileGrind keeps working.