The Problem with Knowing Things Publicly

An interactive breakdown of how Bench works

There is an old idea in markets that if you want to find out what something is really worth, you put people’s money behind it, Skin in the game and conviction at cost. It works, It’s why prediction markets consistently outperform polls and expert panels at forecasting outcomes.

But the version of this idea that has been deployed so far only solves half the problem. It surfaces accurate probabilities on binary outcomes. It does nothing for the far larger category of questions that don’t have a binary outcome and it actively destroys the value of private expertise the moment it’s made public.

Bench is an attempt to solve the second half.

The Leak Problem

Here is the structural failure of every public information market: expertise is only valuable while it’s private.

An A&R rep has spent three years in a specific regional scene shows, demos, DMs from producers, the whole pipeline. They hear a track from an unsigned artist that nobody outside that scene has touched yet. The conviction is specific and grounded: this artist, this sound, this moment in the cycle. It isn’t a guess. It’s pattern recognition built from thousands of hours of listening.

The moment they stake on that artist publicly, the information is out. Other A&R reps see the bet. Labels that weren’t paying attention start paying attention. The unsigned artist’s perceived value shifts not because anything changed about the music, but because someone credible just put money behind it on a public chain. The signing window closes. Competing offers appear. The rep’s edge, which was real and hard-earned, evaporates the second it became readable by anyone watching the market.

What made the conviction valuable wasn’t just being right. It was being right before anyone else was looking. Public staking collapses that gap immediately. The alpha was in the lead time, and the lead time is exactly what gets destroyed by on-chain visibility.

On Bench, the stake is encrypted. The rep commits their conviction and it stays private until the market closes. No other label sees the move. No trader can read the signal and act faster. The artist remains undiscovered to everyone except the market creator, who receives the aggregated conviction only after the window has closed. The rep’s three years of scene knowledge stays theirs until the moment it’s supposed to stop being theirs.

This isn’t a bug in prediction markets it’s a fundamental property of any public ledger. Transparency and information value are in direct conflict. You can have a system where everyone can verify the bets, or a system where genuine expertise retains its value, but not both at the same time.

That is why information markets have remained a theoretical curiosity for decades. The mechanism is sound, the infrastructure to run them privately hasn’t existed.

Until Arcium made encrypted compute practical on Solana, it still didn’t.

What Arcium Actually Does Here

Arcium’s encrypted compute environment allows computation to happen on private inputs without revealing those inputs to anyone, including the nodes doing the computation. In the context of Bench, this means a participant’s stake who they are, what they chose, how much they committed is processed in an encrypted state from the moment it’s submitted until the market creator closes the market and decrypts the aggregate.

No other participant sees anything. No one watching the chain sees anything. The protocol operators see nothing meaningful. The encrypted data is there, verifiable as existing, but unreadable to everyone except the intended recipient at the designated moment.

This changes what’s possible. If your bet is invisible, there’s no reason to obscure your genuine view. You stake your real conviction because staking it costs you nothing socially or strategically. The perverse incentive to hide what you actually think which ruins every public information market disappears entirely.

The Opportunity Market Structure

Bench calls its product an opportunity market rather than a prediction market, and the distinction is more than branding. The mechanics are genuinely different.

A prediction market asks a binary question with a verifiable outcome. A Bench market asks an open question with a best answer. The creator launches with a prize pool, a question, and an initial set of options. Participants stake on the option they believe best answers what the creator is looking for. That’s different from betting on what will happen it’s revealing what you believe is right.

Two things make Bench structurally different from anything adjacent.

First, the option set isn’t fixed. Participants can introduce options the creator never listed. A draft market opens with 40 prospects; someone adds a 41st. That name, the answer to a question the creator didn’t think to ask is often the most valuable thing the market produces.

Second, conviction has a price and a timestamp. Larger stakes carry more weight, earlier stakes earn a multiplier, the participant who commits before any consensus exists takes on the most uncertainty, and the reward structure reflects that.

The Solana Piece

None of this is economically viable at high transaction costs. A meaningful opportunity market needs many participants making considered stakes across a range of options. Markets with 200 participants each staking moderate amounts don’t work if the fee per transaction is meaningful relative to the stake.

Solana’s throughput and fee profile make the participation economics work. Small conviction bets are viable, large markets with many participants are viable. The infrastructure matches the use case in a way that higher-fee chains don’t.

Who uses this

The clearest way to understand who opportunity markets serve is to look at who currently can’t get what they need from existing tools.

A music label trying to make A&R decisions can run focus groups or trust internal instinct. Neither surfaces the kind of stake-backed, independent conviction that Bench can aggregate. A sports team can watch film and talk to scouts, but has no mechanism to get hundreds of people with genuine domain knowledge to reveal their real views simultaneously without contaminating each other. An investment firm can conduct expert network calls, but the experts know they’re being interviewed, the signal is curated, not raw.

Bench creates the conditions where people who know things reveal what they actually think, weighted by how confident they are, without that information bleeding out to anyone positioned to exploit it. The creator gets signal they can act on. The participants who were right get paid. The expertise stays valuable because it stays private until the window is closed.

This isn’t a better prediction market, it’s a different tool for a different problem one that has existed for long without a mechanism to solve it. The infrastructure to make it work, encrypted compute on a fast chain, is only recently real. Bench is the first serious attempt to build on top of it.