Tired of scattered PDFs and endless browser tabs, a Master's student built note.md, a local-first macOS workspace with built-in AI. Let's dive in!
What do developers do when they're procrastinating? Play games, start a holy war on Reddit, or maybe just build an entire software ecosystem from scratch? A gigachad Master's student chose the latter, building a slick markdown app simply because existing tools were driving him nuts.
The product is called note.md. At its core, it’s a local-first, Markdown-based workspace exclusively built for macOS.
The creator started this project while preparing for his thesis. We’ve all been there: your workflow is an absolute dumpster fire. PDFs saved in random folders, notes on Apple Notes, citations in another clunky app, and half your brainpower leaking out across 80 open browser tabs eating all your RAM.
Initially, he just wanted to build a personal, offline version of Confluence. But as he kept writing university papers, the app evolved. He didn't just want a notepad; he wanted an all-in-one hub to read, cite, and write without context-switching.
The cherry on top? He jammed local AI models into the app. We're not talking about a basic chatbot wrapper. This local AI acts like a tireless intern: extracting tables and images from research papers and performing semantic searches across your database. No need to rely on external ai tools or sketchy cloud hosting. Your data stays strictly yours.
The launch garnered a solid 200+ upvotes, and the comment section is a goldmine of dev perspectives:
The "Why reinvent the wheel?" Camp One user asked the million-dollar question: "Bro, with Claude and ChatGPT reading PDFs now, why not just drag your paper into a chat?" The creator fired back with facts: It’s not about adding features; it’s about killing friction. Research is messy. You want to connect the dots and check claims without constantly breaking your writing flow to chat with a bot.
The "Cries in Windows" Camp A curious dev praised the clean UI and immediately asked: "Windows version when?" The creator gave the classic developer 'maybe later' response: "It's written in Swift because Apple hardware makes running local AI models a breeze." Translation: Buy a Mac or prepare to wait an eternity, Windows folks.
The AI Skeptics: "How do you trust it?" A deep-diver asked the critical question: How does the app handle evidence quality? Does it know if a paper strongly supports a claim or just loosely mentions it? The pipeline is actually pretty clever. A search algorithm grabs the 12 most relevant text chunks, feeds them to a small local Language Model, and categorizes them as Supports, Contradicts, Nuanced, or Irrelevant, complete with a confidence percentage. Of course, garbage in, garbage out—the AI only works well if you feed it quality sources.
The biggest takeaway for our dev community here is the power of Dogfooding. If you want to build a successful product, stop trying to "change the world" with some abstract blockchain-crypto-web3 nonsense. Build something that cures your own massive headache first. When you solve a real problem for yourself, it resonates with others naturally.
Secondly, don't sleep on the "Local-first" trend. Users are getting fed up with renting their brains to big tech's cloud infrastructure. Building a powerful, privacy-focused, offline-capable app is a massive money-making niche right now. Ship it!
Sauce: Product Hunt
