LAPIS

ML Research Notes Hub

3 min read

Welcome to our Quartz-based home for ML research and experimentation logs. This is a lightweight, local-first, markdown-driven setup designed to keep our ideas flowing and our knowledge sharp. For setup details, check the official Quartz docs.

Why This Exists

After two-plus years on Lark, we’ve hit some walls. Here’s what’s been dragging us down:

  • Search slog: Slow searches make digging up old ideas a chore.

  • Page lag: Waiting on loads kills quick comparisons—think 100 context switches at 10 seconds each, that’s 16 minutes lost daily.

  • LLM roadblock: Lark’s clunky API locks our data away, stifling AI-powered insights or web tie-ins.

  • Dull vibes: It’s functional but corporate-bland—boring tools sap motivation to log more.

  • Typing delay: Web latency (worse than, say, Google Docs) adds virtual brain fog—every millisecond slows your mind, and that stacks up.

This site flips the script: local markdown files, fast edits, and a workflow that plays nice with LLMs and human creativity.

What We’re Aiming For

  1. Fast & Fluid Editing
    Local edits in VSCode beat any online editor. Update logs on the fly, keep the momentum going.

  2. LLM-Powered Boost
    Write raw, then let AI rephrase, reformat, or query the full knowledge base. Future-proofed for retrieval-augmented generation (RAG).

  3. Rock-Solid Persistence
    Flat files dodge platform risks (like Lark’s regulatory woes). Portable, durable, done.

  4. Looks Good, Feels Good
    A clean, sharp interface makes writing and browsing a win—more notes, better sharing.

How It’s Structured

The content folder (see GitHub) powers this Quartz site, built for research flow:

/  
├── index.md                   # You’re here: site overview  
├── experiments/               # Where the action happens  
│   ├── active/                # Live experiments (e.g., [YYYY-MM]-cool-idea.md)  
│   ├── archive/               # Past work by domain (nlp/, vision/, etc.)  
│   └── techniques/            # Reusable tricks (data-transforms.md)  
├── outcomes/                  # Results that matter  
│   ├── reports/               # Final write-ups ([YYYY-MM]-project-report.md)  
│   ├── insights/              # Patterns we’ve cracked (training-dynamics.md)  
│   └── figures/               # Visuals that stick (embedding-visualizations.md)  
├── tools/                     # How we roll  
│   ├── environment/           # Setup guides (python-environments.md)  
│   ├── aws/                   # Cloud know-how (s3-configuration.md)  
│   └── obsidian/              # Workflow hacks (plugins-setup.md)  
├── reference/                 # Quick lookups  
│   ├── commands/              # Snippets (git-workflows.md)  
│   ├── packages/              # Libs we lean on (pytorch-patterns.md)  
│   └── cheatsheets/           # Fast guides (markdown-formatting.md)  
├── resources/                 # Outside inspo  
│   ├── papers/                # Key reads ([YYYY]-descriptor.md)  
│   ├── datasets/              # Data we use (twitter-favorites.md)  
│   └── learning/              # Growth fuel (tutorials.md)  
└── _templates/                # Starters for experiments, reports, etc.