dedoomify

Reframes AI doom language with local phrase rules, highlights each change, and lets you restore the original.

Read AI news without the doom. Click the dedoomify button on any article about AI and it rewrites the doom framing into the plain language an engineer would use about software with defects: • "The model is misaligned" becomes "the model has a bug" • "An existential risk" becomes "a product risk" • "A rogue AI" becomes "a malfunctioning AI" • "Superintelligence" becomes "very capable software" • "p(doom)" becomes "estimated failure rate" The extension uses built-in phrase rules to change the framing. Direct quotations are left untouched, and every edit is highlighted so you can compare it with the original. Hover a highlight to see the original words, turn the highlights off for a clean read, or undo everything with one click. Turn on "De-doom automatically" and articles about AI doom are rewritten as they load. A quick check on your device picks them out, so other pages load exactly as before. Private by design: the rewriting happens entirely in your browser with built-in phrase rules. The extension collects no data and sends nothing over the network. Until you turn on automatic mode, it only touches the page you click it on. The extension itself uses phrase rules, not an AI model. For an optional rewrite by a small on-device AI model, the popup opens the same article on dedoomify.com with that engine selected. The website fetches the article through its server and downloads the model into your browser; model inference happens on your device. Browsers without WebGPU use phrase rules instead. dedoomify reframes articles; it doesn't fact-check them. It's open source: github.com/akspad/dedoomify