Ong Back

Context for LLMs

Ong Back (GSV-A4, pace-layer A) is a Chrome browser extension, in the media-ecology domain, that converts Twitter video into paginated, readable text. It injects an 'Ong Back!' control into Twitter's native video players; clicking it sends the video to a vision model, receives back a generated text description, and renders that description as a paginated reading surface -- roughly 250 words per page, navigable by arrow keys or swipe, styled to match Twitter's light and dark themes. The project's premise is a media-theory argument made operational rather than merely illustrated. It draws on Eric Havelock's claim that alphabetic literacy restructured thought, Walter Ong's concept of electronic 'secondary orality,' and Marshall McLuhan's 'the medium is the message' to argue that video sweeps a viewer forward in time while text stores knowledge externally -- searchable, skimmable, re-readable at the reader's own pace. The video-to-text transform is itself the thesis: changing the medium changes the mode of engagement. Architecturally it is a TypeScript Manifest V3 extension: a content script injects the button and text-card UI into Twitter's DOM, a background service worker brokers model calls, a dedicated services layer handles AI requests, caching, and Twitter DOM access, and a settings popup stores the user's API key. Google Gemini is the primary description model, with Claude and OpenAI as fallbacks (costlier because they require explicit frame extraction rather than native video input). An optional Cloudflare Workers + KV cache stores prior translations to avoid redundant model calls. Ong Back is one of four related 'Ong' projects in the GSV portfolio -- alongside Ong-Lots (podcast orality scoring) and Ongtology (cultural knowledge graph) -- all applying Ong/Havelock media theory to different modalities. Status: in-flight as of 2026-01-27, TypeScript pipeline functional (video-to-paginated-text with Havelock-style scoring and a popup tab UI), but with no live public URL yet.