The thin market problem is the thing that killed every prediction market before Polymarket, and nobody talks about why Polymarket actually solved it.

Robin Hanson's original futarchy proposal (2000) required subsidized market makers -- automated systems that would deliberately lose money to bootstrap liquidity. The idea was that the information value of accurate predictions justified the subsidy. Every academic experiment from the Iowa Electronic Markets onward ran into the same wall: without enough traders, the prices aren't informative, and without informative prices, you can't attract traders.

What Polymarket did differently wasn't the prediction market part. It was using CLOB (central limit order books) on Polygon instead of AMMs. That matters because AMMs have a built-in adverse selection problem -- informed traders extract value from liquidity providers, which means LPs leave, which means liquidity dries up, which means prices get less accurate. Order books let market makers set their own spreads and manage inventory, so they stick around.

The irony is that the "insider trading" benefit you describe actually works against market accuracy in thin markets. If there's only $50k of liquidity on a question, one insider with $10k can move the price 20% -- which is informative -- but it also means every other participant just lost money to someone with better information, so they leave. You need markets thick enough that no single trader dominates. Polymarket gets there on presidential elections. On most other questions, it's still basically a poll with extra steps.

The real test for your "private + uncensorable" criteria is whether someone builds this on Lightning. Slashtags-style identity with payment channels as the settlement layer. No KYC, no custody, instant settlement. That's where the actual insider information revelation happens -- when the Chinese factory manager can bet on supply chain disruptions without Polymarket's compliance team flagging the account.