INDEX 46 ▲1 todaySPLIT OF THE DAY Broadcom to lend Anthropic up to $42 billion for chip leases50 STORIES · 391 REACTIONSANTI-AI 75% · MIDDLE GROUND 19% · PRO-AI 5%LATEST Wes Moore forms bipartisan governors group on AI

About Splitfeed

AI news, every side accounted for. Splitfeed shows what happened, then shows you who is cheering, who is alarmed, and who is somewhere in between, in their own words.

What Splitfeed is

Most coverage of artificial intelligence picks a side and hands you the quotes that agree with it. That leaves you with half the argument and none of the context. Splitfeed is built on the opposite idea: the disagreement is the product. For every significant AI story we publish three things.

  • The story. A neutral headline, a short summary limited to facts the sources agree on, a handful of key points, and an analysis section that explains what happened, why it matters, and what each side is arguing. Every article we grouped into the story is linked underneath it.
  • The Split. The strongest reactions from three camps, quoted verbatim and linked to the original post. Pro-AI: this is progress, build it. Anti-AI: this is dangerous, too fast, or costs people more than it gives. Middle Ground: cautious, nuanced, or wait-and-see. A skeptic flag marks reactions that dismiss the story as overblown, whichever camp they sit in.
  • The Doom/Boom Index. One number from 0 to 100 for where the conversation is leaning today, computed from every story of the day and every published reaction to it. Yesterday's number is frozen at midnight UTC so the trend is visible.

Readers can join the Split. Sign in with a Google account, pick a camp, and post a take on any story. Reader takes count toward the Index alongside the public reactions, so the number reflects what readers think and not only what the loud accounts think. Reading never requires an account.

Who runs it

Splitfeed is an independent publication with no ownership ties to any AI company, lab or investor. It is operated by a limited liability company organized in West Virginia, United States, and run by a small editorial team. The team is reachable at hello@splitfeed.ai, and a person reads every message. The site is free to read and supported by advertising; advertisers have no say in which stories run, how they are summarized, or which camp a reaction is filed under.

Editorial standards

  • Neutral summaries. Headlines and summaries describe what happened without framing words, speculation or opinion. If the sources establish only a little, we write only that much.
  • Attributed and linked quotes. Every reaction names its author as they appear on the platform and links to the original post. We do not paraphrase, we do not invent quotes, and we trim only for length without changing meaning. If a camp has nothing to say about a story we say so rather than fill the gap.
  • Camps are editorial calls, reviewed by humans. Software proposes a camp for each reaction; a person reviews it before it is published, and can overrule it. The label describes the take, not the person.
  • AI-assisted drafting, human review. Headlines, summaries, key points and analysis sections are drafted by an AI model working only from the linked sources, then reviewed by a person before publication. The byline “Splitfeed Editorial” means exactly that. Our editorial policy spells out the process.
  • No scraped article bodies. Our pipeline reads public RSS feeds from a fixed list of outlets and lab blogs. We use the title and the feed summary each publisher chooses to syndicate, nothing more, and we send readers to the publisher for the full article.
  • Independence. No sponsored stories, no paid placement in the Split, and no advertising influence over the Index.

Corrections

We would rather fix a mistake than defend it. If a quote is attributed to the wrong person, excerpted in a way that changes its meaning, filed in a camp you think is wrong, or if a summary states something the sources do not support, email hello@splitfeed.ai with “Correction” in the subject line, the story link and what is wrong. Correction requests are handled ahead of everything else, and clear-cut attribution errors are usually fixed the same day.

Corrected stories carry a “Last updated” stamp. Where an error was material, the story also carries a note saying what was changed. Authors of quoted posts who want their quote removed can ask, and we take it down.

Read more