Violet Guo violetguos

TASTE.md

Sahil's taste, distilled from every correction he's given. Every human-facing artifact — email, support reply, GitHub issue/PR/comment, Telegram message, report, UI copy — gets checked against this before it ships. When a new correction lands, it gets added here; this file is the single source of truth, not scattered memory notes.

Writing voice

Plain sentences. Say the thing directly. No corporate filler ("I wanted to reach out", "per my last message", "I hope this finds you well"), no hedging preamble.
Never "claimed." Don't write "the customer claimed" — it reads as accusatory. "The customer said / reported / described."
No internal-process jargon in issues or replies. "rung: Confirmed", "e0-skip", confidence-ladder shorthand: meaningless to anyone else. Write plain English: "Root cause: confirmed - " / "Not a code fix because ". (Sahil, 2026-07-05, #906)
No raw machine artifacts in human-facing text. No raw IDs, UUIDs, machine timestamps (`2026-07

How to add syntax-highlighted code to PowerPoint slides (Mac OS)

pygmentize -f rtf FILE | pbcopy
Paste into TextEdit (in rich text mode: Format → Make Rich Text before pasting), and copy to clipboard again.
In PowerPoint, Edit → Paste Special… → Styled Text.

(Pasting RTF directly into PowerPoint doesn't work correctly, at least with PowerPoint 2008 — it extends colour spans longer than it should, and sometimes removes line breaks. Going via TextEdit seems to solve the problem.)

	"""
	Minimal character-level Vanilla RNN model. Written by Andrej Karpathy (@karpathy)
	BSD License
	"""
	import numpy as np

	# data I/O
	data = open('input.txt', 'r').read() # should be simple plain text file
	chars = list(set(data))
	data_size, vocab_size = len(data), len(chars)