For certain use-cases it may be preferable instead to annotate titles with AI-generated tags that annotate the content, such as social media comments. Use-case: there is major political interference currently on websites like Twitter where AI-generated humans are creating tweets and replies intentionally designed to divide. We effectively need to augment the web with 'syntax highlighting' for semantics so we can completely optimize and route our attention correctly.
We need to be careful that the tags are objective representations and not subjective. "Disinformation" for example is not a possible tag, at least not until MCP web search is integrated into the system.
The insertion mechanism needs to be customizable on a per-site basis, i.e. each site can have its own DOM crawling and mutation schema for structured injection. Simply appending a prefix to the HTML element may not be preferable when operating on tweets, where instead we may wanna work it in this specific frontend and make a more intuitive integration.
I know all of this is probably beyond the scope of this project. I'm just laying down the bricks for record purpose, in case we can get a more solid community movement around Unhype, or a different more advanced fork aimed at reclaiming control over information.
For certain use-cases it may be preferable instead to annotate titles with AI-generated tags that annotate the content, such as social media comments. Use-case: there is major political interference currently on websites like Twitter where AI-generated humans are creating tweets and replies intentionally designed to divide. We effectively need to augment the web with 'syntax highlighting' for semantics so we can completely optimize and route our attention correctly.
We need to be careful that the tags are objective representations and not subjective. "Disinformation" for example is not a possible tag, at least not until MCP web search is integrated into the system.
The insertion mechanism needs to be customizable on a per-site basis, i.e. each site can have its own DOM crawling and mutation schema for structured injection. Simply appending a prefix to the HTML element may not be preferable when operating on tweets, where instead we may wanna work it in this specific frontend and make a more intuitive integration.
I know all of this is probably beyond the scope of this project. I'm just laying down the bricks for record purpose, in case we can get a more solid community movement around Unhype, or a different more advanced fork aimed at reclaiming control over information.