judge2020

Logging into the Cloudflare CLI (`cf`) on a headless / SSH box

cf auth login (the Cloudflare CLI, not Wrangler or cf the Cloud Foundry tool) assumes it's running on a desktop with a browser. On a remote SSH host with no display, the default flow fails. This explains why, and gives a small helper that logs you in from a headless box using a browser on your local machine.

Verified against cf v0.3.0.

Why the default flow breaks over SSH

Grok Build (`grok`) — full codebase breakdown by concern

An investigation of xAI/SpaceXAI's open-source terminal coding agent ("Grok Build"), published as a single squash commit (b189869 "Publish harness and TUI open-source") synced from their internal monorepo.

Scale: ~1.32M lines of first-party Rust across 77 crates, plus ~26K lines of vendored third-party Rust. About 45% of first-party Rust is test code once inline #[cfg(test)] modules are counted (see the repo-composition page in this gist; dedicated test files alone are ~16%). The root Cargo.toml is generated by monorepo tooling and marked read-only.

LOC by concern

Trim transparency from bounds of a gif

Sometimes you have a gif that has a transparent pixel border. This is inconvenient when you don't want this for chat-app emojis or stickers.

Courtest of chatgpt here's a script that takes in a gif via python input and outputs the trimmed gif.

prerequisites

ffmpeg
gifski (no video mode needed)

pkill -HUP UniversalControl

Thanks to mmac_m3a on macrumors forum.

python3 -m pip install discord.py
copy the script into a file (e.g. initialize.py)
run python3 initialize.py
quit once you see "We have logged in as <user#1234>"

Setup

Sign into https://tiktok.com and go to your profile
Click favorites or likes
scroll down till all your videos load
paste the script in console and run it
Script will download the file containing newline-delimited links
download via yt-dlp like so:

	/*
	Collect and export a list of Twitter usernames you're following, along with follow-back status, to a CSV file every 80 seconds. Done client-side using Javascript running in the Console window of the web browser.

	Instructions:
	1) Modify the 'LANGUAGE' variable to match your Twitter interface language.
	2) Customize 'SKIP_USERS' with usernames you do not wish to include in the CSV.
	3) 'PRINT_FOLLOW_INFORMATION' determines whether follow-back status is included.
	4) 'SKIP_FOLLOWERS' and 'SKIP_NON_FOLLOWERS' settings filter out respective user groups.
	5) 'MS_PER_CYCLE' controls the frequency of data collection cycles.
	6) Load the Twitter page showing users you're following, open the console (F12), paste this code, and run it.

	/* eslint-disable camelcase */
	import React, { useState } from "react";
	import {
	FormInputRange,
	FormInputSelect,
	FormItem,
	FormLabelRange,
	} from "@tesla/design-system-react";
	import { Form, Input } from "@tesla/informed-tds";
	import { arrayOf, func, number, shape, string } from "prop-types";

	import glob
	import json
	from datetime import datetime
	from matplotlib import pyplot

	age_ts = []
	age_keys = ["prob_13_17", "prob_18_24", "prob_25_34", "prob_35_over"]
	age_lists = {k:[] for k in age_keys}

	gen_ts = []

	# python -m pip install pyperclip
	# credit gpt4
	import pyperclip
	import time
	from urllib.parse import urlparse, urlunparse

	# Function to check if the content in the clipboard is a URL
	def is_url(content):
	try:
	result = urlparse(content)