Blogs Should Talk Back
Publishing on the web has been a monologue for thirty years. An author writes, a reader reads, and the only feedback channel is a comment box at the bottom that the author may read next week and the reader will never revisit. Every real conversation the article provokes happens somewhere else — in chats, in meetings, in someone's head on a commute. The page itself stays inert.
The page you're reading right now is our attempt to fix that. Next to this text (or behind the button, on your phone) is an AI that has read this article and is waiting to discuss it with you — while you read, not after. This piece explains what that is, why we built it, and what we think it means for written content generally.
Reading generates questions in real time
Watch yourself read anything substantive: objections and questions surface mid-paragraph. "Does that claim hold for my industry?" "What does that term actually mean?" "This contradicts what I read last week." In the monologue model, those thoughts evaporate — the cost of chasing each one (open a tab, search, lose your place) exceeds the value of most of them. So readers finish articles with their questions unasked, which means half the value of the text is never collected.
An embedded companion drops that cost to a sentence. Ask, get an answer grounded in the article, keep reading. Disagree with a claim and argue — the model knows what the author said and can defend it, concede its limits, or point out what the article doesn't cover. The article becomes the opening move of a conversation instead of the entire game.
Honest boundaries make it useful
We've given the companion narrow rules, because narrow is what makes AI trustworthy: it grounds answers in the article first and says when it's going beyond it; it's told to admit when the article doesn't cover something rather than improvise an answer; and your conversation is yours — sign in, and it's saved per article, so returning to a piece next month means resuming a thought, not starting over.
The interesting unit isn't the article or the chat. It's the article plus your conversation with it — a version of the piece that exists for exactly one reader.
| Traditional | Interactive | |
|---|---|---|
| Feedback loop | A comment nobody reads | A conversation, in the moment |
| Return visits | Start over each time | Resume where you left off |
| Editorial signal | Pageviews only | The actual questions readers ask |
What this does to writing
Writing for a page that talks back changes the author's job. You no longer need to preempt every objection in the text — the companion handles the long tail, and the article can stay sharp and opinionated instead of bloating into defensive completeness. The questions readers actually ask become the best editorial signal a writer has ever had: they tell you precisely where the argument was unclear, which claims provoked pushback, and what the next article should be.
This is a v1, and we're building in the open. Coffee Corner — topic rooms where readers discuss with each other, not just with the AI — is next. If the conversation model here works, or annoys you, tell the companion; we read the transcripts of what people ask (that's the point). The monologue era of publishing didn't end because someone declared it over. It ends one page at a time, whenever a page learns to answer back. This one just did.