Paste a URL and get your page judged on the things Paul Graham actually tests for: whether a stranger can tell what it is, whether it is built for someone specific, and whether the words sound like a person.
Free, no login, no email. Nothing stored beyond a 24 hour cache. Markdown copy
Paste any URL
We load that one page the way a browser would — its text, its markup, and a screenshot. No login and no email.
Nine tests, applied to your page
The lens is nine recurring heuristics drawn from the essays, filtered down to the ones a single page can actually show evidence for. Every finding has to quote something on your page.
Get the verdict and the rewrites
One sentence on the most important thing, then the findings — each with the exact sentence it objects to and, for copy problems, the sentence we would have written instead.
You can already find out whether your contrast ratios pass, your meta description is the right length, and your images have alt text. Those tools are correct and they are useful, and not one of them will tell you the thing that is actually wrong with most landing pages, which is that a stranger cannot tell what the product is.
That failure is invisible to automated checks because nothing about it is malformed. The markup is valid. The page loads fast. It just does not say anything. What catches it is a reader with opinions, and the most useful opinions about startup pages have been written down in public for twenty years.
The obvious way to build this is to scrape the essays and search them at run time for whatever seems related to your page. We deliberately did not. Searching a corpus finds passages about your topic — point it at a pricing page and it returns the essay about pricing. What makes the review feel like the lens is not the topic, it is the recurring tests: does this say what it is, is it for someone specific, could I try it, would you say this sentence out loud.
Those tests are stable and there are only a handful of them, so they are written down in a file, distilled from the essays and cited back to them. That also means the same page gets the same verdict tomorrow, which a retrieval-backed version could not promise.
“Pick good cofounders” is the first item on his own list of thirteen. It is not in this tool, because a landing page cannot show evidence for it, and a reviewer asked to judge it from a landing page will invent something that sounds right. Every test in the lens has to be answerable from what is actually on your page. That constraint costs a few famous ideas and it is what makes the findings trustworthy.
No, and it does not pretend to be. It is a review lens distilled from his published essays — the recurring tests he applies, written down and applied to your page. It never writes as him, never claims to know what he would say, and never invents quotations. The essays it was built from are listed on the page, so you can check the lens against the source.
No login, no email, no ads. Same as this one.
Superflow agents review the pages you ship on every change, against the standards your team actually holds, and leave the findings as comments on the page itself.