A glowing Claude Code terminal stretches into a massive layered codebase map labeled with tiny file trees and agent task trails, leaving clean dark space for bold text.
AI Off Leash episode: Claude Code ships a 1M token context window and what it unlocks for coding agents. Both hosts, TJ shocked and Shane skeptical.
Use round1-1 as the base. Tighten Shane likeness, reduce the UI clutter behind the 1M, and protect the top title from crop risk. Do not change the exact wording. Keep TJ shocked, Shane skeptical, and the central 1M token visual obvious.
8.7Round winnerBest overall packaging: exact required text, huge readable promise, clear TJ shock versus Shane skepticism, and the central 1M context visual is instantly understandable.
4Pretty tech visual, but the packaging is too vague. '1M TOKENS' alone is not enough for this audience or this episode promise.
8.3Very clickable and the words are extremely clear, but it feels busier and more synthetic than the winner. The hat text is a credibility leak.
Use round2-2. It has the cleanest one-second read. Next fix Shane likeness first, then add one unmistakable Claude Code/codebase cue without cluttering the center. Preserve the exact title text and keep the giant 1M unobstructed.
8.2Strong title, strong TJ shock, clear 1M token object, and good Claude Code specificity. The file tree helps technical relevance but adds clutter at mobile size.
8.6Round winnerCleanest read of the three: exact required text, huge readable 1M, simple agent/codebase flow, TJ shocked and Shane skeptical. It communicates the episode promise fastest on mobile.
8Readable and clean, but the UI cards/folders feel more generic SaaS than a concrete Claude Code 1M context-window moment.
Delivered to Slack as the single best thumbnail, scored 8.6 of 10.