read
Build a consumer product that puts missing context back into digital media.
Registration opens September 1A clip without the full exchange. A headline without the article. A label without its meaning. A quote without the conversation around it. An AI summary that quietly drops the part that mattered.
People are not credulous. Most readers pick up tone, motive, and credibility, and can often sense when something has been cropped, exaggerated, loaded, or trimmed. But instinct only goes so far. The original video may be buried. The full article may be behind a paywall. A label may be doing more work than the evidence supports. A summary may sound neutral while leaving out the point that mattered.
Restoring the context takes longer than the next post takes to arrive, and nobody has time to do it for everything they see.
of Gen Z scored poorly on evaluating the shortcomings of AI technology in 2024. Growing up online does not come with the ability to audit what the feed hands you.
EY & TeachAI, AI literacy survey of 5,000+ Gen Z respondents, December 2024
government-imposed internet shutdowns hit 54 countries in 2024. When the record itself goes dark, the fragment is all anyone has left to judge by.
Access Now / #KeepItOn, internet shutdowns in 2024, February 2025
people, 32% of the planet, remained offline in 2024. Any tool that asks people to go check the source has to reckon with who can reach it.
ITU, Measuring digital development: Facts and Figures 2024
The user invokes it, the output is theirs, and nothing gets published about creators or anyone else. The architecture is open. These are entry points, not requirements.
An assistant that retrieves the fuller source behind a fragment and reports what it can and cannot verify.
Local inference for speed and privacy, with no round trip and no record of what the user looked at.
Context delivered in place, inside the page where the content is already being read.
A share-sheet or overlay path that works where most feed consumption actually happens.
Transcript retrieval, clip-to-full-exchange matching, and source tracing for a pasted link.
Gates check completeness, not quality. Substantive evaluation happens at the final rounds, where judges also weigh investability.
The pathway from access to voluntary use. Having the tool is not the same as reaching for it.
Whether the product can tell genuine debate from controversy that was manufactured to look like debate.
Whether the product would survive a platform turning hostile.
Whether every consequential output still traces to a linked source when the product is running at feed speed.
A consumer product in production, with users, telemetry, and a defensible business, on your record whether or not you win the investment.
Nobody has yet built a context product that people return to voluntarily. Whether a student team can is an honest question with an answer at the end of 18 months.
AI capability will keep moving. Part of the test is building something that gets better as models improve rather than something a model update makes obsolete.
The next 18 months are designed to find out what you can build when the problem is real and the framework gets out of your way.
Register at xfoundry.org/xperience