Blanket does not execute trades. It does not hold funds. It is not Kalshi's product. Three negations, zero settlement logic, one unverified claim: that a small business can hedge operational risk with a binary event contract.
The specification documentation runs through regulatory positioning with precision. It lists tariff exposure, weather disruption, energy price variance, and election outcomes as hedgeable categories. It is silent on one parameter: the payout function. Does the recommended contract pay proportionally to loss? On the disclosed evidence, it does not. It pays a fixed amount or nothing.
I read product specifications the way I read smart contracts — I search for what is absent. The documentation explains what Blanket does not do: not execute, not custody, not settle. It never mentions basis risk. That omission is the most significant line in the specification.
In 2020, I spent three months stress-testing DeFi liquidation cascades on a local Ethereum testnet. I documented how oracle manipulation vectors emerge not in calm markets but in the precise scenarios protocols are designed to survive. Hedging systems fail the same way. They break under the exact conditions they are built to protect against, because those conditions expose the divergence between model assumptions and market reality. Kalshi's AI-powered hedge layer has this structural weakness embedded from the first contract design. The public conversation focuses on compliance status. It should focus on settlement mechanics.
Kalshi operates as a CFTC-regulated prediction market in the United States. It holds a compliance-first position in a sector dominated by offshore competitors like Polymarket, which operate without comparable regulatory encumbrances on the Polygon network. Blanket is a third-party tool: an AI engine that analyzes a small business's operational risk and recommends Kalshi event contracts as hedges. The intended user is a small enterprise with weather, energy, tariff, or political-event exposure but without the balance sheet or sophistication to access traditional futures or insurance markets.
The design is carefully scaffolded. Blanket reads market data through Kalshi's API, generates risk assessments, and outputs contract suggestions. It does not participate in order execution. It does not touch the payment rail. The construction deliberately locates the recommendation function outside the regulatory perimeter while the exchange absorbs the compliance burden. If the CFTC recharacterizes these outputs as solicitation or investment advice, the scaffolding collapses into registration requirements.
This is a liability firebreak. If Blanket's recommendations produce losses, Kalshi can plausibly claim the tool is not its product. The third-party developer absorbs reputational damage while the exchange preserves its regulatory goodwill. That division of responsibility works in normal circumstances. Its integrity under a mass-loss event is unverified.
The technology is unremarkable. An AI model consumes market data and emits contract suggestions. The real assets in this system are Kalshi's order book, settlement infrastructure, and event-result oracle. Blanket is a thin recommendation wrapper around a genuine matching engine. Thinness is not inherently a weakness. It becomes one when marketing language conflates contract recommendations with risk coverage. The word "hedge" does heavy lifting in that conflation.
Since Blanket handles no funds, KYC and AML obligations remain with Kalshi as the registered exchange. This is operationally clean but analytically wasteful. Event contracts covering tariff and political-event risk can function as alternative risk-transfer instruments with sanctions implications. An AI layer reading business exposure data could plausibly serve as a detection front-end for anomalies — a business hedging against a tariff outcome it has private knowledge of is a classic information-edge pattern. That surveillance feature is absent from the disclosed scope. The data is being collected. The analytics are not applied to market integrity. That gap will matter if the CFTC scrutinizes the relationship.
The competitive positioning matters. Kalshi sits at the intersection of two narratives: the compliance-first prediction market and the enterprise risk management play. Polymarket captures retail attention with permissionless, crypto-native event contracts. Kalshi wants the institutional narrative — regulated product, enterprise use case, defensible positioning. Blanket is the wedge: a productized story that says prediction markets are not gambling, they are risk infrastructure. That story is elegant. It also masks a structural mismatch in the underlying instrument.
The Payout Function Is the Product
The first audit question is fundamental: what is the payout function?
Kalshi event contracts settle on discrete outcomes. A weather contract pays a fixed amount or nothing, depending on whether a temperature threshold is crossed. A tariff contract pays if a policy event occurs, independent of magnitude. This is binary exposure. A two percent tariff increase and a forty percent increase produce identical payouts unless the contract ladder is structured with multiple thresholds. The documentation offers no evidence of such structuring.
Small business risk is continuous. A five percent tariff erodes margins. A forty percent tariff destroys the business. A hedge that cannot distinguish between these states is not a hedge. It is insurance without loss adjustment — a fixed indemnity that ignores the size of the claim.
This is basis risk: the divergence between an instrument's payoff and the economic loss it purports to cover. It is the most under-discussed failure mode in the prediction-market-as-risk-management narrative. The marketing framing claims hedging functionality. The contract mechanics deliver only event exposure.
The empirical consequence is predictable. A small business buys the contract before a tariff ruling. The ruling arrives. The contract pays its fixed amount. The actual loss is three to four times that amount. The business declares the hedge failed. The complaint lands at a state regulator. From the regulator's perspective, a retail customer was sold an opaque derivative-like product under a misleading risk-management label. That characterization carries enforcement consequences.
I audited yield aggregation protocols in DeFi with the same architectural flaw: mathematical models that appear correct in isolation and break under real-world loss distributions. The model assumes one distribution. The market delivers another. In prediction markets, the binary contract is the source of the mismatch. The AI risk engine cannot repair a structural gap in the underlying instrument. It can only recommend contracts that either pay a fixed amount or nothing. That is the entire product surface.
The Liquidity Paradox
Prediction markets exhibit a liquidity paradox. Order books for niche event contracts — an El Niño index, a specific tariff ruling, a regional temperature threshold — are thin in normal conditions. Spreads are wide but workable. In the exact event the contract is designed to hedge, liquidity does not increase. It evaporates. Exits become one-sided.
My DeFi stress-testing surfaced the same pattern. In liquidation cascades, the deepest books failed first because everyone exited simultaneously. The counterparty pool disappears exactly when risk materializes. A small business holding a weather contract during an extreme event cannot exit at fair value. The market is structurally one-sided at that moment.
Blanket recommends entries. There is no disclosed mechanism for managing exits. For a long-tail event contract, this creates a one-way door. The user enters, news breaks, the market gaps, and the position cannot be closed except at a punitive price. The hedge becomes a sunk cost.
Consider the settlement mechanics dispassionately. A prediction market contract is a cash-or-nothing digital option. The payout is fixed. The premium is a function of implied probability. When a small business buys a tariff contract, it is acquiring a binary digital option position. The convexity profile is stark: bounded upside, full downside on the premium. Traditional proportional instruments do not share this profile. A futures contract gains and loses with the underlying. A binary contract does not — it snaps.
Tariff announcements are discrete catalysts. They cause immediate repricing. The number of users wanting to hedge at that instant is the highest it will ever be, while the available liquidity is the lowest it will ever be. This is not an edge case. It is the core scenario of the product's stated use case. A hedging tool that cannot guarantee exit capability under stress is a conditional claim on a market that does not exist when the condition triggers.
The Unit Economics of Fear
Hedging is an episodic decision. Small businesses do not hedge tariff risk daily. They evaluate exposure quarterly, or when news forces their attention. This creates an uncomfortable economic structure: meaningful customer acquisition cost against episodic transaction volume. The source material does not disclose pricing or fees. Three revenue models are plausible: subscription access, per-recommendation fees, or commission from Kalshi on referred volume. The first two face the frequency constraint directly. If a business engages the tool four times per year and pays a subscription, the effective cost per engagement is high. The third model creates an alignment problem: the AI engine generates revenue per recommendation regardless of whether the recommendation is structurally sound. Volume incentives do not produce risk-optimized outcomes.
I trust the null set, not the influencer. A recommendation engine monetized by referral volume is structurally indistinguishable from affiliate marketing.
The Firebreak Is Thinner Than It Appears
The "does not execute, does not hold funds" construction is designed to keep Blanket outside the broker-dealer definition. The argument is historically defensible: recommending contracts without executing orders or touching funds falls short of statutory brokerage triggers. The instability lies in recharacterization risk.
The CFTC has latitude to revisit the event contract category. It has already demonstrated willingness to act on political event contracts. If it determines that event-based contracts used as hedges function like retail commodity options, the entire framework shifts. Retail commodity options carry suitability obligations, intermediary registration requirements, and disclosure mandates. A third-party AI recommendation engine sits squarely inside that perimeter.
There is also the solicitation theory. Under securities law, solicitation can trigger broker-dealer registration even without execution authority. The legal question is whether Blanket's AI output constitutes individualized investment advice or generic content. The answer depends on facts not disclosed: whether recommendations are tailored to a user's specific exposure data, whether the AI claims directional accuracy, and whether the output is structured as a solicitation to buy specific contracts. The specification suggests tailoring. Tailored recommendations are harder to defend as mere content.
State-level enforcement is another vector. The CFTC is not the only regulator. State securities regulators have jurisdiction over retail solicitation. The small business owner is a retail investor under most state definitions. An investment-adviser registration exemption does not automatically extend to AI-generated individualized recommendations. The third-party structure that protects Kalshi from federal claims may not protect it from a fifty-state regulatory patchwork.
This uncertainty is not priced into the product's risk narrative. Compliance is described as settled. It is provisional.
The Data Extraction Layer
The most undervalued asset in this arrangement is not the hedge recommendation. It is the operating data the AI collects.
Small businesses feeding revenue figures, supplier relationships, cost structures, and risk exposures into a third-party AI engine generate a dataset of significant analytical value. The privacy disclosures are absent. No data processing agreement is described. No third-party sharing restrictions are stated. No model training disclosure exists.
This is a material omission. A dataset of small business exposures, correlated with subsequent hedging behavior and realized outcomes, is precisely the input required to sharpen prediction market pricing models. The tool presented as serving the retail user may also function as a data collection layer that improves market-maker pricing. The economic value of that feedback loop accrues to the platform, not the user.
Metadata is just data waiting to be verified. The privacy structure of Blanket determines whether this is a benign service or an extraction mechanism. The documentation does not permit that verification.
The Wedge Story
The compliance positioning is the strategic asset. Kalshi's regulated status attracts a different user class than Polymarket's permissionless model. Small business owners will not deposit funds into a jurisdictionless platform. They will consider a CFTC-regulated venue. Blanket extends this reach by reducing the cognitive cost of entry: an AI tool that translates business exposure into contract recommendations.
This is a real wedge. Insurance brokers ignore the long tail of micro-enterprises. Futures brokers demand margin ratios and account minimums the target user cannot satisfy. Prediction markets, intermediated by an AI layer that speaks in risk terms, fill a genuine gap.
The constraint is breadth. The disclosed scope covers weather, energy, tariffs, and elections. These map to insurance and futures product categories that already exist under a different regulatory regime. The disintermediation risk is real: if Kalshi's event contracts functionally substitute for licensed risk-transfer products, insurance regulators will take notice. The wedge that opens the market can also trigger the intervention.
The public conversation around Kalshi centers on political event contracts and CFTC tolerance for election markets. That focus misses the actual systemic risk. The retail small-business narrative is the blind spot. An economic protection instrument is being sold with the vocabulary of insurance while possessing the payout profile of a lottery.
The failure scenario encounters no structural obstacle. A small business hedges a tariff event through Blanket. The event occurs. The contract pays its fixed indemnity. The loss exceeds that indemnity by a factor of three. The business complains to a state securities regulator with a theory: the AI tool's recommendation constituted investment advice from an unregistered adviser. Blanket has no assets. Kalshi does. Under an apparent-agency theory, the exchange inherits the liability.
Verification is the only trustless truth. This product is not verifiable at the settlement level. The AI model is proprietary. The validation methodology is undisclosed. The settlement mechanics of recommended contracts are not explained in any available material. Proofs don't care about marketing narratives; they are either valid or invalid. The proof for Blanket's hedge thesis has not been submitted for verification.
Silence in the code speaks louder than hype. The regulatory firebreak is beginning to look like a game of hot potato with a regulator that has repeatedly preferred to intervene after concentrated retail losses rather than before.
The next twelve to eighteen months will produce one of two resolutions. The CFTC reclassifies retail event contracts as commodity options requiring suitability obligations and professional intermediaries, or a concentrated small-business loss event triggers state enforcement under an investment-adviser theory. Both paths terminate the current construction.
The engineering correction is a shift toward proportional instruments — contracts calibrated to loss magnitude rather than binary event occurrence. The infrastructure for such instruments exists. The market demand is unproven. Until that correction, Blanket remains a recommendation engine with an insurance vocabulary and a lottery payout structure.
The institutional clients I brief apply a simple filter: does the instrument reduce the variance of the underlying exposure? For binary event contracts carrying basis risk, the answer is conditional at best. The basis risk is the product.