Discover gists
The expanded prompt from Samraaj's TV makeover video. Identify your TV, give your coding agent eyes through a phone camera, then build and install a home screen for the platform you actually own.
- Have your TV, its remote, a computer, and a phone ready. Put the TV and computer on the same trusted home network.
- Open a coding agent on your computer with permission to work with local files and development tools, and the ability to inspect images. A text-only chat cannot deploy to your TV by itself.
- Open 01-TV-MAKEOVER-PROMPT.md below, copy its entire contents, and paste it into the agent. Fill in any preferences you already know; leave everything else as unknown.
- Let the agent identify the hardware first. It will then build the camera watcher, show you a QR code, and ask you to grant camera access and point your phone at the TV.
A pattern for building personal knowledge bases using LLMs.
This is an idea file, it is designed to be copy pasted to your own LLM Agent (e.g. OpenAI Codex, Claude Code, OpenCode / Pi, or etc.). Its goal is to communicate the high level idea, but your agent will build out the specifics in collaboration with you.
Most people's experience with LLMs and documents looks like RAG: you upload a collection of files, the LLM retrieves relevant chunks at query time, and generates an answer. This works, but the LLM is rediscovering knowledge from scratch on every question. There's no accumulation. Ask a subtle question that requires synthesizing five documents, and the LLM has to find and piece together the relevant fragments every time. Nothing is built up. NotebookLM, ChatGPT file uploads, and most RAG systems work this way.
I'm writing this on Github as it will give the reader wanting to get in contact some level of plausible deniability as opposed to visiting my website.
While both websites are encrypted, visiting a more generic website gives you more plausible deniability. Any institution that can wiretap or has metadata retention will only see you connecting to https://gist.github.com/ and not which page youre visiting.
I'm a technologist at Amnesty's Security Lab.
Last updated: 7 July 2026 (v1.5.1)
UIDrop is a Chrome extension that reads the visible CSS of webpages you choose to snap. Extraction runs entirely in your browser. We don't operate any servers, and we never see, collect, or transmit the pages you snap or the design data extracted from them.
| <!DOCTYPE html> | |
| <html> | |
| <head> | |
| <title></title> | |
| <meta charset="utf-8" /> | |
| <script src="Scripts/jquery-1.9.1.min.js"></script> | |
| <link href="Content/bootstrap.min.css" rel="stylesheet" /> | |
| <script src="Scripts/isRockFx.js"></script> | |
| <script> | |
| $(function () { |
| #EXTM3U | |
| #EXTINF:-1 group-title="𝐁𝐚𝐧𝐠𝐥𝐚𝐝𝐞𝐬𝐡𝐢" tvg-logo="https://i.postimg.cc/0215t9fT/btv.png",BTV | |
| https://btv.hridoytv-channel.workers.dev/btv.m3u8 | |
| #EXTINF:-1 tvg-logo="https://i.postimg.cc/28VMF8hx/btv-news.png" group-title="𝐁𝐚𝐧𝐠𝐥𝐚𝐝𝐞𝐬𝐡𝐢",BTV News | |
| https://btv.hridoytv-channel.workers.dev/news.m3u8 | |
| #EXTINF:-1 tvg-logo="https://raw.githubusercontent.com/ahan443/logo/refs/heads/main/atn%20bangla.png" group-title="𝐁𝐚𝐧𝐠𝐥𝐚𝐝𝐞𝐬𝐡𝐢",ATN Bangla | |
| https://tvsen5.aynaott.com/P3y2URgG7LDe/index.m3u8 | |
| #EXTINF:-1 tvg-logo="https://i.postimg.cc/W4SwWtd3/20240822-225840.png" group-title="𝐁𝐚𝐧𝐠𝐥𝐚𝐝𝐞𝐬𝐡𝐢",ATN News | |
| https://tvsen6.aynaott.com/da6WMXAk/index.m3u8 | |
| #EXTINF:-1 tvg-logo="https://i.postimg.cc/5NFp4xR7/20240822-225600.png" group-title="𝐁𝐚𝐧𝐠𝐥𝐚𝐝𝐞𝐬𝐡𝐢",Boishakhi |
