PageLeft Manifesto

Ramanujan wrote to three Cambridge professors. The first two returned his papers without comment, but the third, Hardy, recognized the genius. If Hardy had been busy that day, one of the greatest mathematical collaborations of the 20th century wouldn’t have happened.

What if? Milliseconds to Cambridge instead of weeks by ship. Free publishing to the entire world. The network that built YouTube and Wikipedia.

Now ask him to reach a mathematician working on the same problems. Social media wasn’t made for this; arXiv requires a credential network he doesn’t have. So he asks a chatbot to translate his notation into standard terms, puts those into search, and gets textbook recommendations from AI Overview. How about community forums where like minds once gathered? Nope. Chatbots ate them. He could try email, but most won’t reply. He develops his theorems with the chatbot and writes a blog post. Nobody reads it: personal blogs are black holes on the internet.

Now take Hardy’s side. He’s stuck on a problem in partition theory, so he searches for new approaches and gets his own papers back. When he asks a chatbot about it, it summarizes what he published last year. Somewhere out there, a blog post with the answer exists. Even if the model were trained yesterday, Ramanujan’s posts would be buried; Language Modeling Is Compression, and compression favors the majority. One noble voice drowned out under a million. Hardy would give anything to find Ramanujan, but he can’t. Not even in 2026.

Ideas can’t find other ideas. AI was supposed to be the next miracle, but it made it worse: flooding every channel with noise, compressing the signal into consensus, and siloing each conversation into a private chat. It may soon be harder for minds to find each other than the year Ramanujan sent his letter to Hardy. It may be exaggerated, but the trend is clear.

But what if they could be found?

Two people describing the same problem in different vocabularies end up in the same space. Ramanujan’s notation wouldn’t match Hardy’s keywords, but the meaning would.

But you’d drown in noise. The searchable web is mostly ads, spam, and marketing copy. Ban one thing: keeping it all for the writer. Now the writing is free for anyone to read, to share, and to build on. Writers who choose it want their ideas to be free, to belong to the world. Those who keep their ideas for sale won’t be welcome.

Extend it, and what you create belongs to the world too. The network of knowledge can only grow.

How do you know what’s good? Let the contributors decide. When someone links to your work, that link is a vote. More votes, higher rank. How do you know what’s bad? Intelligence can tell. Bad ideas decay away.

The technology exists.

The distance between ideas is measurable: embeddings. The signal, filter, and protection are one license: copyleft. The ranking is a citation graph: PageRank. The patent expired in 2019.

A crawler walks from seed URLs, follows links, and only keeps pages that carry a copyleft license. Everything else is ignored. Each page is embedded by meaning — title and opening text — so that similar ideas land near each other regardless of vocabulary. Links between pages form a citation graph. Pages that others point to rise; islands sink. Standard PageRank.

One API endpoint. No UI — bring your own. Give it a query in natural language, get back ranked pages — URL, title, snippet, license, semantic score, and rank score. That’s it.

Anyone can donate crawler or inference compute to keep it free, as in beer. A search engine for ideas that chose to be free.

Ramanujan’s theorems wouldn’t have needed Hardy’s luck. Indexed by meaning, filtered by intent, ranked by peers: they find mathematicians willing to share the same problems.

No more PhD required to talk shop. No grant. No paywall. No begging for peer review. Someone publishes an idea and the nearest minds find it before the week is over. The intellectual conversation moves where the ideas are, not where the institutions gatekeep.

Ideas want to find each other; they always have. Copyleft frees them to flourish beyond the author. We carry them to life. Ramanujan’s notation and Hardy’s theory were reaching for the same truth from opposite ends of the world. Ads drowned them, AI buried them. But now, they are alive. Connected, united, expansive.


All PageLeft postspageleft.ccSource code on GitHub

Written with Claude Opus 4.6 via Claude Code. I directed the argument; Claude drafted prose.