Gleanings · Demo

Walks through setting up tokens, a type scale and ten core components, and argues small teams should copy a proven system instead of inventing one.

Explains caching long system prompts, batching background jobs and routing simple requests to a smaller model, with before-and-after cost numbers.

A weekend build using Tauri and SQLite; the author shares what made the app feel native and why they skipped accounts entirely.

Uses a single orders table to show ROW_NUMBER, running totals and moving averages, with the query for each step.

Argues that the first screen a new user sees is usually empty, and shows five patterns that turn it into onboarding.

Covers chunking by headings, hybrid keyword plus vector search, and forcing the model to cite passages, with an evaluation set of 200 questions.

A curated list covering animation, forms, testing and state management, with one line on why each one matters.

A week-by-week log of scope cuts, the one metric tracked, and why launching ugly beat waiting for polish.

Shows how removing half the charts, adding plain-language captions and fixing the color scale changed how the team used the dashboard.