To make media framing visible and concrete, rather than just something people are told exists in the abstract. Reading the same real story rewritten two intentional ways, side by side, tends to land differently than a general warning to "watch out for bias."
Every story is generated in both directions on purpose — that's the whole design, not an attempt to hide a lean. The site's own editorial choice is which stories to cover and how the neutral "facts" summary is worded; we try to keep that summary as plain and unopinionated as possible, but no summarization process is perfectly free of judgment calls.
Reading two real outlets shows you their actual editorial choices, sourcing, and voice — which is more valuable than anything here, and we link to independent bias-rating resources specifically to encourage that. What this site adds is a controlled comparison: the same underlying facts, run through two intentional framings, so the technique itself is easier to isolate and name.
The daily pipeline pulls real, current headlines via web search and extracts a neutral factual summary before any framing is applied. Mistakes are possible — this is an automated process, not a fact-checked newsroom — so treat the site as illustrative rather than as a primary source for the facts themselves.
They're explicitly instructed not to — no new facts, quotes, or statistics beyond what's in the neutral summary, only emphasis and word choice. If you ever spot a rewrite stating something that isn't supported by the facts line above it, that's a failure worth noticing, not an intended feature.
It's an AI's own estimate, on a 1–10 scale, of how differently the left and right rewrites of a story ended up reading. It's a rough, automatically generated signal for which story to feature that day — not a rigorously validated metric, and not a claim about how divergent real-world coverage of that story actually was.
No. There's no original reporting, no correspondents, and no editorial desk making judgment calls the way a newsroom does. It's closer to an interactive teaching tool built around real headlines than a news outlet in its own right.
Pasted story text is sent to generate a rewrite and isn't saved afterward. Email addresses submitted for the future weekly digest are stored so we can notify you when that feature launches — nothing is sent to a mailing list yet, since that feature doesn't exist as of this writing.
Once a day, automatically. The homepage always shows the latest batch, with an archive available for past days.
They're AI-generated annotations of the AI's own rewrites — useful for building intuition about framing techniques, but not independent, human-verified media criticism. Treat them the way you'd treat any single automated analysis: a starting point, not a verdict.