Anthropic has started a new blog called Claude Explains, discussing the capabilities of its AI models and written by that self-same AI model. The educational posts are “written” by Claude to explain how to use Claude. It's like an AI's personal diary, but with debugging tips instead of romantic exploits.
The blog is pitched as a “corner of the Anthropic universe where Claude is writing on every topic under the sun,” but that's not quite accurate. Claude may draft the pieces, but a team of human experts and editors sand and polish the rough outline to make sure they are readable and accurate, or as Anthropic calls it, a “collaborative approach."
Now, this idea isn't terrible on its face. This kind of AI-human tag team makes a lot of sense, at least when the AI is writing about itself. An article about how Claude can design a website or organize a financial report is well within Claude's wheelhouse. It's just explaining its own abilities. But a technically reasonable explanation and a few useful examples aren't a full blog post. Claude’s best work still won't always result in a coherent article, or one that a real person would want to read.
Anthropic is honest about how humans are part of the process throughout blog post production. Claude may start the car, but humans are at the wheel and navigating, lest it drive the article right into a ditch full of hallucinations and mixed metaphors. Anyone who’s used AI without guardrails knows this scenario isn’t far-fetched. AI is excellent at saying things that sound right until you try to actually apply them.
AI ghostwriting
Collaboration is certainly an efficient approach. Claude can crank out thousands of words without breaking a sweat, and if you’re using it to explain the same concepts it was trained on, it’s got a decent shot at getting things mostly right. Problems arise much more quickly when AI writers are left unsupervised, especially on subjects outside of the AI model's abilities.
The blog doesn’t proclaim the human element, so a casual reader might assume Claude is doing all the writing. That’s a branding choice, and not a neutral one. It creates a kind of halo effect, subtly bragging about how the AI breaks down data analysis and sounds like a real writer. Except it isn’t human. It’s a word blender that gets better results when someone else chooses the ingredients and adjusts the settings. And that distinction matters, especially as more people begin to trust AI-generated information in contexts far beyond technical blogs.
There's a steady stream of stories about media outlets embarrassing themselves by believing AI can replace entire content teams. The Chicago Sun-Times published AI-generated book recommendations for titles that didn’t exist, and multiple outlets have published AI-written features full of errors. And that's not even counting Apple's attempts at news summary headlines.
Claude Explains feels downright reasonable by comparison. If you’re going to use AI to produce content for public consumption, maybe keep to what it knows best. And don't leave out the humans.
You might also like
You must confirm your public display name before commenting
Please logout and then login again, you will then be prompted to enter your display name.