
What happened
Claude Code produced 29 article drafts — roughly 96,000 characters and 104 images in about two hours — and a separate AI, checking each against the source records, flagged 24 errors and overstatements.
Why it matters
The drafts were natural to read, so a reader alone would not have caught the mistakes; having a different AI verify each article against the source records appears to be what surfaced them.
Summaries like this, in your inbox every morning.
The author's goal was to hand both writing and fact-checking to AI, because publishing articles as a side job leaves no time to research each piece by hand. The approach was to use a different AI from the one that wrote the drafts. Two separate AIs were set up to compare each number and claim against the original records. The 24 problems were not limited to mistaken figures; plausible-sounding reasons and background absent from the records had slipped into the transitions between sentences. Some articles could be read as meaning the opposite of what the records concluded. The author notes that a follow-up article covers the types of errors and how much AI was used to check them.
Pick your industry and the AI tools you use, and get news related to your work every day.
Free · 30 seconds with Google · unsubscribe anytimeWhat is AIToday? →
Ask AI anything about this article. The AI reads this article, earlier AIToday articles, and Wikipedia, and cites its sources. Q&As are published on this page for other readers too.
Anthropic introduced Claude Duet, a proactive AI agent in research preview for Pro and Max subscribers, availa…
Moshi CEO Anekha Sokhal said on the AI Engineer podcast that her tool — 'if Bloomberg, Claude and Netflix had…

CFO Susan Li told Meta's third-quarter earnings call that "traditional enterprise companies" don't have enterp…

Rerunning the same images and prompts that broke Gemini 3.5 Flash Lite, the developer found Gemini 4 Argon enc…

A free Mac App Store app, aI'm Thinking, plays keyboard taps that speed up when Claude Code or Codex is writin…

Anthropic released Claude Duet, an agent for professionals and small businesses, built on Claude Opus 5.5 and…
