What if AI failures were described like ordinary software issues? Paste a link to an AI article and dedoomify turns down the apocalypse: "the model is misaligned" becomes "the model has a bug". Less Skynet. More software. And every change is highlighted with the original text one hover or tap away. You can try an example without signing up or installing anything!

The default engine runs on quick phrase rules or you can run a local AI model in your browser via WebGPU; there's also a Chrome extension that applies the rules to the page you’re reading. And all the code is open source! https://github.com/akspad/dedoomify

There's so much AI doom in the world right now that I wanted to see what articles would look like reframed to be less doom and gloom and instead more matter-of-fact (sprinkled with some sunshine and roses).

Let me know what you think - suggestions and improvements are always welcome :)