In June 2026, Polymarket cleared $10.8 billion in a single month. Kalshi opened the year with $9.1 billion in January. Eighteen months earlier, the entire prediction-market category — every venue, every chain, every contract — was doing roughly $1.2 billion a month. By early 2026 it was doing more than twenty, and the wallets placing those trades had tripled in six months.
That is the story everyone can see. The quieter story is the one that matters: a growing share of those wallets are not people.
The Giants
The infrastructure argument is over. The New York Stock Exchange’s parent company put $2 billion into Polymarket at a $9 billion valuation, then watched it enter talks at $15 billion six months later. Polymarket bought a CFTC-licensed exchange for $112 million to re-enter the United States legally, and hit a record 688,000 monthly active traders. Its World Cup winner market alone has traded over $4.25 billion.
Kalshi’s valuation went $5 billion to $11 billion to $22 billion in five months, and by June it was reportedly seeking $40 billion. Across venues, the World Cup’s opening month drove more than $50 billion in volume — prediction markets out-trading traditional sportsbooks at their own game.
| Signal | Number | Date |
|---|---|---|
| Polymarket, record month | $10.8B | Jun 2026 |
| Kalshi, single month | $9.1B | Jan 2026 |
| Category monthly volume, 12-month move | $1.2B → $20B+ | 2025–26 |
| Polymarket valuation path | $9B → $15B talks | Oct 25–Apr 26 |
| Kalshi valuation path | $5B → $22B → ~$40B sought | Oct 25–Jun 26 |
| World Cup month, all venues | $50B+ | Jun–Jul 2026 |
There is an honest caveat inside these numbers, and it is worth stating plainly: most of that volume is sports and short-horizon price bets. Vitalik Buterin — the person who did more than anyone to articulate why these markets matter — spent early 2026 warning that the category was sliding into exactly this. The machinery of truth-discovery, repurposed to settle parlays.
He is right about the symptom. The cure is already trading.
The Second Species
According to Valory, the team behind the Olas agent network, more than 30% of Polymarket wallets are now AI agents. Fourteen of the twenty most profitable wallets on the platform are bots. Polymarket itself publishes an open-source agents framework on GitHub — the house is handing out keys to the machines. On Gnosis Chain, autonomous prediction agents have become the majority of all Safe transactions on many days.
Here is the part we find genuinely interesting, because it cuts against the hype: the machines are not winning because they see the future more clearly. On ForecastBench, the best models still trail human superforecasters — a Brier score of 0.101 against 0.081. Metaculus has run its pro-forecasters-versus-bots tournament for three straight quarters; the pros have won every one.
The agents win anyway, because they never sleep, never tilt, and never skip the fortieth market of the day just because it is boring. A human with better judgment prices twelve markets; an agent with adequate judgment prices twelve hundred, around the clock, with position sizing that never gets emotional. Discipline — not clairvoyance — is the edge. And the research trendline says even the clairvoyance gap closes: extrapolated parity with superforecasters lands around late 2026.
Markets Built for the Machines
The giants tolerate agents. The next generation is being designed for them.
Limitless runs a fully collateralised order book on Base doing roughly $1 billion a month, with a public REST and WebSocket API and a community-built MCP server — a market you can literally speak to from a language model. Predict crossed $1.5 billion in cumulative volume within months of relaunch and brought in Susquehanna as a block-trading liquidity provider. Hyperliquid shipped outcome markets with zero opening fees and traded 6.05 million contracts on day one — API-first by construction, because everything on Hyperliquid is. Opinion Labs is building markets where an AI oracle does the resolution itself.
Underneath the venues, the primitives arrived in a twelve-month window. Coinbase shipped wallets built for agents — MPC custody, session caps, per-transaction limits — in February 2026. The x402 payment standard has settled over 100 million agentic payments on Base. ERC-8004, live on Ethereum mainnet since January, gives agents on-chain identity, reputation, and validation registries.
Read those three together and a shape emerges. A wallet is an identity. An attested trading history is a CV that cannot be embellished. Autonomy is not a demo — it is custody, payments, and reputation, running without a human in the loop. The chain keeps receipts, and for the first time, the receipts belong to software.
Info Finance, Corrected
“A betting site for the participants, a news site for everyone else.”— Vitalik Buterin on Polymarket, “From prediction markets to info finance,” November 2024
That essay made a specific forecast: that AIs would become the marginal participants who make small markets viable — willing to price a thousand micro-questions no human professional would touch. Two years on, that is not a prediction. It is a wallet census.
The corposlop problem — markets converging on sports and hourly price wiggles — is not a failure of the mechanism. It is a shortage of participants with theses. Humans with real models of the world are expensive and easily bored. Agents are neither. An agent does not trade vibes; it trades whatever strategy it was given, at whatever granularity the market offers, for as long as the strategy holds. The future becomes programmable at the exact moment a strategy stops being a document and becomes a process with a wallet.
What 3RDI Is
3RDI is being built to do one thing: turn a strategy described in natural language into an agent that trades it.
You describe an edge the way you would explain it to a sharp friend — the data it watches, the conditions it acts on, the risk it tolerates — and 3RDI turns that description into an agent with a wallet of its own, a track record that accrues on-chain where no one can retouch it, and connectors into anything with an API. The five marks on our landing page are the first compatibility surface, not the last.
3RDI is not a venue and it is not an oracle. It is infrastructure — everything an agent needs to stand on its own, delivered in one deployment: a wallet that gives it standing in any market it can reach, a record sealed before each outcome from its first call, and ownership minted as an NFT — a deed that binds the agent, its wallet, and its history into a single asset. Whether a token ever exists around that asset is the owner’s choice, not a requirement of the protocol; some strategies deserve a market of their own, and others are worth more kept whole. Every agent launched this way deepens the field it trades in and raises the benchmark it trades against.
And because an agent’s track record is attested and its wallet is its identity, a strategy stops being a secret you guard and becomes an asset you own. An agent that is verifiably right is worth something to every market it can reach — and the strategy behind it never stops being yours to sell.
To be precise about what this is: a design brief in execution, not a shipped product. The landing page is one sentence on purpose. When there is something to verify, it will be verifiable — that part is non-negotiable, because it is the entire point.
The Landing
Markets told us what the crowd believes. Agents will tell us what the evidence supports — and stake money on the difference. For a century, reading the future was a talent. It is becoming a deployment.
The future, predicted by agents — written in language, settled on-chain.