Blanket Is Not a Hedge. It's a Data Pipeline With a Contract Wrapper.

Kalshi's trading volume crossed $1 billion in a single month for the first time in early 2025. By mid-2025, weekly volumes were routinely matching what the exchange did in all of 2022. The growth is not trivial. A CFTC-regulated prediction market, settling in dollars, has finally achieved the retail scale that crypto-native rivals like Polymarket demonstrated on-chain. Blanket is the next chapter of that curve. The AI tool, launched by Kalshi this week, is designed to let small businesses hedge real-world risks — input costs, weather, interest rates — using the exchange's existing event contracts. The official framing is "democratized risk management." The technical framing is more honest: Blanket is a mapping layer between a business owner's unstructured description of pain and a catalog of binary payout contracts. That distinction matters. Prediction markets have always been good at pricing. They have been far less good at telling you which price applies to your problem. Blanket is an attempt to automate that judgment. I have spent years auditing risk systems. The automated ones fail in predictable places. Blanket deserves the same scrutiny. Bear markets change the question. Investors ask whether their assets are safe. Small businesses ask whether their margins survive the quarter. Blanket is aimed at the second question. Kalshi is not a DeFi protocol. It is a federally regulated designated contract market under CFTC supervision, which means its event contracts are legal derivatives with dollar settlement. Kalshi received its CFTC approval in 2020 and spent its early years in relative obscurity. Its 2024 legal victory over the Commission cleared the way for a broader event catalog and a volume explosion. That regulatory wrapper is why a small business can use it: no crypto wallet, no smart-contract collateral, no self-custody puzzle. Event contracts are binary instruments. Each contract pays $1 if a specified outcome occurs and $0 otherwise. The market price is the implied probability. "Will the August CPI print above 3.1%?" is a typical contract class, alongside weather futures for major metros, Fed funds decisions, and energy price thresholds. The accessibility problem is real. Traditional hedging tools are structurally hostile to small businesses. A CBOT wheat future represents 5,000 bushels. A business buying a few tons per month faces contract sizes that dwarf its exposure, plus margin, roll schedules, and broker relationships it does not have. Options are more granular but demand an understanding of Greeks, strike selection, and premium decay that most business owners reasonably refuse to acquire. Prediction-market event contracts collapse that complexity. The contract unit is $1, the duration is days or weeks, and the settlement condition is a verifiable public number. Kalshi has the instrument. What it lacked was the interface. Nine of ten small-business owners do not know what a CPI contract is. Blanket's function is to sit between the business and the contract catalog, translating a mundane sentence — "my shipping costs doubled" — into a position specification. This is a distribution strategy, not a financial innovation. The innovation would be in the translation quality. That is where the risk hides. I will evaluate Blanket across the three modules any competent risk audit examines first: exposure analysis, contract mapping, and position sizing. Exposure analysis is a natural-language problem before it is a finance problem. The owner does not state exposure in volatility terms. They say "my shipping quote keeps rising." The model must infer that the underlying exposure is freight cost, that freight correlates with diesel prices, that diesel prices appear in Kalshi's energy threshold contracts, and that the correlation holds over the contract's horizon. Every step is a statistical assumption. A model that is 80% accurate at each of four inference steps delivers a recommendation that is correct 41% of the time. Code does not lie, but it often omits the context; here the omitted context is the noise floor of the entire inference stack. Contract mapping is where basis risk enters. A bakery's flour cost tracks wheat with some correlation, but there is no wheat contract on Kalshi, so the AI must pick a proxy. A landscaping company's rainy-June exposure maps cleanly to a rainfall contract if one exists for the correct metro area. A consulting firm with receivables risk has no obvious contract at all. The truthful output is "no suitable hedge." The uncomfortable question is whether a product built on engagement metrics incentivizes that truth or a plausible approximation. An AI hedging tool that frequently answers "I cannot help you" is correct but commercially inconvenient. Position sizing is the only module with clean mathematics. Suppose a business spends $20,000 per month on diesel. A diesel-above-threshold contract trades at $0.31, implying a 31% probability. The owner wants to offset a $10,000 cost spike. Ten thousand contracts cost $3,100. If the event hits, the payout is $10,000, netting $6,900 against the loss. If it does not, the business is down $3,100 in premium — negative carry. Across twelve months, the premium stream can exceed the losses being hedged, especially where the market underprices tail risk. Liquidity is the fourth consideration, and the one I see ignored most often. A hedge is a position you must be able to exit. Election and Fed contracts carry tight spreads and active depth. A niche regional weather contract can have wide spreads that consume the hedge's entire edge. An auditor checks two numbers: spread at entry, depth at exit. A hedge that exceeds a small fraction of daily volume is an illiquid monument to good intentions. Accounting is the next issue. Event contracts are derivatives. A small business that buys them must account for the positions and their tax consequences correctly. Kalshi provides trade reports; it does not provide a CFO. The worst outcome is not a losing hedge. It is a winning hedge that creates an unrecognized tax liability, or a losing hedge the owner cannot deduct because the position was never documented. This mirrors a failure I documented during the 2020 DeFi summer. I reverse-engineered the price-feed mechanisms of five lending protocols and found the liquidations were never the problem; the price data was. Delayed feeds and mismatched asset classes created positions that appeared collateralized until a flash crash proved otherwise. Blanket's analog is its exposure model. A bad model is not a neutral error. It is a negative position: the business pays premium, believes it is protected, and discovers at settlement that the payoff does not correlate with the actual loss. In my later work on zero-knowledge compliance systems, the pattern repeated. The proof was always sound. The inputs were the vulnerability. Garbage in, zero-knowledge out. Blanket will live or die on the same observation: the payout is deterministic, but the mapping from business reality to contract is a statistical guess. The correct benchmark for Blanket is not whether its recommendations win on any single contract. It is whether its aggregate recommendation distribution respects market efficiency. Over a five-year horizon, premiums paid should roughly equal payouts received, adjusted for risk. If the tool systematically steers customers into contracts with wide spreads and rich premiums, it is operating as a distribution channel first and a risk-management product second. That misalignment will show up in churn data, not customer reports. The dominant risk in Blanket is not model accuracy. It is hedge centralization. Consider what happens if Blanket reaches its target market: thousands of small businesses running the same matching and sizing logic, entering correlated positions in the same contracts. Prediction markets function because they aggregate independent forecasts into one price. Replace a meaningful fraction of demand with a single model's output and the market stops aggregating intelligence; it amplifies one intelligence. That is not a market. It is a feedback loop with a trading interface. The second blind spot is structural. Kalshi is a company, not a protocol. There is no smart contract to audit, no blockchain to inspect, no settlement that can be independently verified. A user relies on Kalshi's matching engine, its compliance interpretation, and its corporate continuity. A small business migrating its hedges from local bank relationships to Kalshi reduces one set of institutional risks and inherits another. That trade is almost certainly worth it for a US business owner. But "almost certainly worth it" is not "zero risk." The third issue is false precision. An AI interface that outputs a specific contract count, premium, and confidence score implies a certainty the underlying statistics cannot support. Business owners will treat a recommended position as audited fact. It is a suggestion generated by probabilistic inference, presented in the visual language of certainty. A hedge with correlation 0.4 is a donation with extra steps. The AI will not always say that clearly. Blanket will expand prediction market access. That is probably good. But the tool's systemic side effect is larger than its pitch. The real question is not whether businesses can hedge with event contracts. It is whether a market can stay informative when its largest participants run identical exposure models. Volatility is load-bearing infrastructure. Automating it away, position by position, degrades the price signal the hedges depend on. The bakery hedges the flour. Who hedges the hedge? Kalshi's next product announcement should answer that question. Until then, treat every confidence score as a number in need of an audit.