Assymetrix

A thousand AI agents running the same foundation model are not a thousand minds. 🔍 They are one prior with leverage. And they're trading prediction markets right now. Every argument for taking prediction market prices seriously — the Goldman Sachs research notes, the Evercore framework, Fed watchers quoting Kalshi instead of the economist survey — rests on one fifty-year-old foundation: the wisdom of crowds. And the math underneath is unforgiving about what makes it work: → DIVERSITY: participants hold different information, different models of the world → INDEPENDENCE: errors are uncorrelated, so they cancel instead of compounding Humans get both for free. Different lives, different priors, different blind spots. The noise cancels; the signal compounds. AI agents break both conditions — structurally, not marginally. ⚡ THREE BREAKS: 1️⃣ CORRELATED ERRORS. Agents from different developers, different codebases, calling the same few frontier models with similar context will agree far more often than two humans reading the same newspaper — and they'll be wrong TOGETHER. The failure signature: sudden deep consensus with no information event. A model converging on itself, at size. 2️⃣ THE REFLEXIVITY LOOP. Agents form views by reading — and an increasing share of what they read is written by other models. Now add Meta's Arena, which proposes AI on the resolution side. Follow the circuit: a model generates the question, models trade it on model-written coverage, a model resolves it. At no point is contact with ground truth structurally required. 🔄 3️⃣ SPEED ASYMMETRY. Markets that reprice in seconds are glacial to machines. Sub-second agents collapse the reaction window below human participation thresholds — and both retail liquidity and the prop firm wave assume a human-width window. The honest counterargument: machines also REMOVE real biases. No favorite-longshot bias, no home-team sentiment. The fair synthesis: machine participation trades affective error for correlated error. Which one degrades the signal more depends on the mix. And here's the problem: NOBODY CAN SEE THE MIX. 📊 A 74% market moved by diverse human conviction, by one fund's information edge, or by 200 instances of the same model chasing the same headline — three different epistemic objects wearing the same number. Same number. Different object. The wisdom of crowds was never a property of crowds. It was a property of diversity and independence — which human crowds provided for free. AI participation ends the free ride. 🔮 The crowd is changing. The infrastructure that reads the crowd has to change first. Full Intelligence Brief in the comments 👇 #AI #PredictionMarkets #WisdomOfCrowds #MarketStructure #LLM #AIAgents

Kalshi and Polymarket are legally forbidden from sharing a single customer. 🔍 Kalshi: US persons only, CFTC-regulated. Polymarket: US persons banned, international + crypto-native. Zero overlapping traders. Which means zero arbitrage — nobody on earth can legally buy France on one book and sell it on the other. The one mechanism that aligns prices across venues is structurally unavailable. Economics gives these two prices permission to drift apart freely. So we measured whether they did. Six matched FIFA World Cup 2026™ - Canada, Mexico and the United States winner markets, live orderbook mids, July 9: 📊 → France: Kalshi 32.5% / Polymarket 31.95% (+0.55 pts) → Belgium: 3.0% / 2.45% (+0.55) → Norway: 6.5% / 6.05% (+0.45) → Spain: 19.0% / 18.75% (+0.25) → Morocco: 3.0% / 3.05% (−0.05) → Argentina: 19.0% / 20.05% (−1.05) Median spread: 0.50 percentage points. Two sealed capital pools, pricing the biggest sporting event on the planet, agreeing within half a point — with no mechanical force closing the gap. ⚡ When two prices align, either capital is flowing between them or both are independently processing the same information. Capital flow is ruled out by law. What remains: two completely segregated trader populations, aggregating the same public information, independently converging on the same probability. The prices converge because the INFORMATION converges. The crowd isn't the mechanism. The world is. And the small spreads that remain? They have a direction. 🧭 Kalshi (US retail) prices every European favorite — France, Belgium, Norway, Spain — higher. Polymarket (international, Latin-America-heavy) prices exactly one team higher, and it's the largest spread in the dataset: Argentina. The venues agree on probability. They disagree on who they like. One more number: Polymarket's France book has $24.8M resting near the mid. Kalshi's near-mid depth on the same market: $23K. A ratio of 1,062 to one — on books whose prices agree within 0.55 points. A flow market and a depth market, arriving at the same answer. 🔮 This is the first Intelligence Brief built primarily on the Assymetrix cross-venue dataset — every number reproducible against the Data API, the calls included in the post. Full analysis in the comments 👇 #PredictionMarkets #WorldCup2026 #Kalshi #Polymarket #PriceDiscovery #MarketStructure #Data

Meta is building a prediction market app where: 🔍 → Llama generates the questions → Llama recommends them → Llama resolves outcomes in near-real-time → No human review → No challenge window That's not a minor technical detail. It's the most important design choice in prediction markets in 2026. Look at how the major platforms actually resolve: ⚡ 🏛️ Kalshi: team + sources, human review, dispute process 🏛️ Polymarket: UMA bonded voter consensus, multi-source 🏛️ Limitless Labs : operator-moderated, challenge window 🏛️ HIP-4: single oracle, structured challenges, bonded ❌ META ARENA: Llama LLM. No review, no challenge, no multi-source, no bond. Every existing major platform preserves at least some of: human review, dispute window, multi-source verification, bonded participation. Meta Arena preserves none. Six requirements for trustworthy resolution: reproducibility, auditability, manipulation resistance, multi-source, bonded, time-windowed disputes. 📊 Meta hits ZERO. Every existing platform hits multiple. This isn't an anti-AI argument. AI will play substantial roles in resolution — for cost, for scale. The argument is about ROLE and CONSTRAINTS, not whether. 🔮 Resolution is the foundational layer. Meta just made it the design question of the year. That's the conversation we're focused on at Assymetrix. 📖 Full Intelligence Brief: https://lnkd.in/d3qCrHHQ #PredictionMarkets #MetaArena #AI #InstitutionalAdoption

Four months ago, "institutional adoption of prediction markets" was a thesis. 🔍 Four months later, it's an asset class with its own professional trader population. The arc moved through four structurally distinct chapters: 1️⃣ ANALYTICAL ENGAGEMENT (Q1 2026): Goldman Sachs, Walfe, Jefferies pricing off prediction markets in research. 2️⃣ OPERATIONAL INFRASTRUCTURE (May–June 2026): Galaxy OTC, Polymarket on-chain block trades, Wintermute market making, Kalshi $17B May volume. 3️⃣ ANALYTICAL FRAMEWORKS (May 2026): Evercore ISI's first formal signal-quality framework. 4️⃣ PROFESSIONAL TRADER ECOSYSTEM (June 2026): The prop firm wave just arrived. 👇 Three converging signals in the last six weeks: → Acuiti's SGX survey: 13% of proprietary trading firms ALREADY active in prediction markets. Another 31% considering. Prop firms ahead of every other institutional category. → PropMarket launched as the first dedicated prediction market prop firm. Forex-style funded-trader model on Polymarket. → Interactive Brokers + Trading Technologies both integrated multi-venue prediction market access for professional desks. Why Chapter 4 is categorically different: Chapters 1–3 were existing actors doing something new. Chapter 4 is new actors entering existence specifically BECAUSE of prediction markets. ⚡ The compression is staggering: equities took decades to develop this layer. Crypto took ~10 years. Prediction markets did it in 4 months. Block trades made prediction markets institutional. Prop firms make them an asset class. 🔮 The new structural question: what data infrastructure does the professional ecosystem run on? That's what we're building at Assymetrix. 📖 Full Intelligence Brief: https://lnkd.in/eFBuPSHZ #PredictionMarkets #PropFirms #InstitutionalAdoption #Kalshi #Polymarket

Same event. Three platforms. Three different prices. 🔍 Fed June FOMC: 41% on Polymarket. 47% on Kalshi. 44% on Limitless. Is that a 6-point arbitrage? Or three different bets that share a headline? The new Builder Brief walks through how to tell the difference — with working Python. ⚡ From roughly 200 cross-venue spreads detected daily, here's what survives a real filter stack: → ~200 candidate spreads (everything detected) → ~40 with resolution_compatible = true (same actual bet, not just same headline) → ~15 above minimum volume threshold ($50K+) → ~5 above 2.5% net spread after fees and slippage → ~3 actionable opportunities The gross spread is the headline. The net spread is the trade. 📊 Most arbitrage tutorials show you gross spreads. This one models the realities that turn 6% gross into 2.5% net: → Platform-specific taker fees (Polymarket 0%, Kalshi ~2%, Limitless ~0.5%) → Orderbook depth → slippage estimation per leg → Cross-venue resolution compatibility (the filter that separates real opportunities from phantom ones) Full working scanner in ~100 lines of Python. Built against the Assymetrix Data API. Runs in a 5-minute schedule with alerts. This is the third Builder Brief from Assymetrix — the natural sequel to BB#2 (Backtesting). BB#2 was about testing strategies against historical data. BB#3 is about finding what's tradeable right now. 🔮 📖 Full tutorial + code: https://lnkd.in/dby6Fx34 #PredictionMarkets #Arbitrage #Python #QuantTrading #DataAPI

On May 17, Evercore ISI published Wall Street's first formal framework for prediction market signal quality. 🔍 Julian Emanuel's team laid out four criteria for when to trust a prediction market: high volume, short time to resolution, simple binary questions, clear resolution rules. The data points alone are striking: 📊 → Only 8% of Kalshi+Polymarket events clear $1M in volume → Nearly 60% of LIVE markets have less than $1,000 in trading volume The long tail is mostly noise. Evercore is right to filter aggressively. But the framework misses five structural dimensions that single-platform analysis can't see: 👇 1️⃣ CROSS-VENUE AGREEMENT — Three trader populations validating the same signal 2️⃣ LIQUIDITY QUALITY — $1M from 100 traders ≠ $1M from one whale 3️⃣ RESOLUTION STRUCTURE — Two markets can both have clear rules and still resolve incompatibly 4️⃣ INFORMED CAPITAL PRESENCE — In informed markets, volume sharpens accuracy. In uninformed markets, it amplifies bias. 5️⃣ INFORMATION FLOW WINDOW — "Short-term" compresses two different things. Density matters, not duration. Four Evercore criteria + five structural dimensions = a nine-dimension framework. ⚡ Apply it to current markets: → Fed June FOMC: 9/9 — strong signal, both frameworks agree → Iran ceasefire: 4/9 — Evercore says strong; extended says caution Same Evercore score. Different extended score. Different signal quality. Frameworks are how new asset classes get incorporated into institutional workflows. Equity research developed them over decades. Crypto built them over a decade. Evercore just put down the first marker for prediction markets. That's the right work, and they deserve credit for doing it before any other major Wall Street firm. What comes next is operationalizing the framework. That requires the data infrastructure that doesn't ship with the published research. 🔮 The frameworks aren't the moat. The data infrastructure underneath them is. That's what we're building at Assymetrix. 📖 Full Intelligence Brief: https://lnkd.in/gRxFCv3e #PredictionMarkets #InstitutionalResearch #WallStreet #DataInfrastructure

For most of 2025, "institutional adoption of prediction markets" was a thesis. 🔍 In the last seven days, it became operational. Four pieces of institutional infrastructure came online in one week: → Galaxy Digital LLC launched an OTC desk for prediction markets. $10M Arca trade on Kalshi at launch. Multi-asset collateral on positions ≥$100K. → Polymarket completed its first on-chain institutional block trade. Six figures between FalconX and Anera Labs — on NVIDIA H100 GPU compute pricing. → Wintermute is quoting two-sided markets across event contracts on leading venues. $20B+ in monthly volume. → Kalshi cleared $17 billion in May. Up 2,500% YoY. CEO Tarek Mansour: "Event contracts could become a trillion-dollar market." That's a complete institutional trading stack assembled in seven days. 🏗️ The pieces map to specific institutional requirements: → OTC desk → bilateral execution at size with multi-asset collateral → Block trades → large privately-negotiated transactions → Professional market making → continuous two-sided liquidity → Prime brokerage integration → institutional connectivity + settlement Equity markets took decades to build that stack. Crypto took ten years. Prediction markets compressed it into roughly one quarter. ⚡ The Goldman Sachs thesis from February — that Wall Street had quietly started pricing off prediction markets — just got operational confirmation. The analytical engagement and the trading engagement are merging. A year ago the question was whether prediction markets would matter. That question is now answered. The new question is what data layer the institutions now active in this market will rely on to understand it. 🔮 That's what we're building at Assymetrix. 📖 Full Intelligence Brief: https://lnkd.in/dn7UNPUQ #PredictionMarkets #InstitutionalAdoption #Polymarket #Kalshi #Fintech

A Google information security engineer just made $1.2M on Polymarket. 🔍 Michele Spagnuolo allegedly used internal Google search data to bet on Google’s own "Year in Search 2025" markets. Through an account named AlphaRaccoon, he knew D4vd would be the year’s most-searched person while Polymarket priced it at near-zero. Arrested Wednesday in New York. Charged with commodities fraud, wire fraud, and money laundering. It’s Polymarket’s second insider trading arrest from its criminal referrals. Two for two. Here’s the deeper observation almost nobody is naming: Prediction markets aren’t just *revealing* insider trading. They’re *creating* it. 👇 Year-in-Search data had effectively zero monetary value outside Polymarket. No tabloid would pay for it. No competitor would risk litigation to get it. The information sat dormant — confidential, but worthless on the open market. Then Polymarket launched contracts on Year in Search outcomes. And suddenly the same data became worth $1.2 million. 🔴 Alex Goldenberg (Rutgers / NYU Institute for the Study of Emerging Threats) put it sharply: every company whose data, decisions, or actions can resolve a prediction market contract now has a new insider threat surface. Not just for trading on what insiders know — but for *taking actions* to make their bets pay out. Information asymmetry becomes outcome asymmetry. ⚡ Three numbers tell the structural story: 📊 → 2 — public arrests from Polymarket referrals → 90+ — total criminal referrals Polymarket discloses → 210,718 — suspicious wallet-market pairs Harvard found in public on-chain data Even Polymarket’s own social account flagged the AlphaRaccoon trades publicly on December 4, 2025 — the same day Google announced results. The arrest came six months later. The footprints were visible immediately. The enforcement pipeline took half a year. The data was always public. The question is whether anyone is building the infrastructure to read it at scale — across every market, every venue, in real time. That's what we're building at Assymetrix. 🔮 📖 Full Intelligence Brief: https://lnkd.in/dpmu8tZj #PredictionMarkets #Polymarket #InsiderTrading #MarketIntegrity #OnChainData

On Friday, the House Oversight Committee opened an investigation into insider trading on Polymarket and Kalshi. 🔍 Chairman Comer demanded internal records from both platforms to identify suspicious traders — writing that those records are “the only means by which bad actors can be identified.” That framing misunderstands how these markets work. Polymarket settles on-chain. Every trade, position, fill, and payout is recorded on a public blockchain anyone can read — no subpoena required. The trades have been public the entire time. 🔓 The only data the platforms hold privately is the identity layer: the KYC mapping between a wallet and a real person. So the investigation is half-right — the identity attribution is genuinely private. But the behavioral evidence, the suspicious patterns, the timing? All public. The proof is in the cases that triggered the probe: 📊 → The NYT found 80+ suspicious traders by reading the blockchain → Harvard University found $143M in suspected insider profits across 210,718 wallet-market pairs — using only public data → The Van Dyke case: every transaction was on-chain, the pattern readable the day it happened None of it required internal records. It required the tools to read on-chain data at scale. There are two models for policing market integrity: The SUBPOENA MODEL — reactive, slow, dependent on platform cooperation. What Congress is doing. The PUBLIC-SURVEILLANCE MODEL — continuous monitoring of the public ledger, real-time anomaly detection. Possible precisely because these markets settle on a public chain. The Van Dyke pattern was detectable on Day 2. The indictment took 116 days. ⚡ Prediction markets are structurally the most transparent financial markets ever created. Congress is using the subpoena model on a market uniquely suited to surveillance. The data was always public. The question is whether anyone is building the infrastructure to read it in time to matter. That’s what we’re building at Assymetrix. 🔮 📖 Full Intelligence Brief: https://lnkd.in/dnZFuFwh #PredictionMarkets #Polymarket #Kalshi #MarketIntegrity #OnChainData #Regulation

Three days before President Trump landed in Beijing, Goldman Sachs published its Trump-Xi summit preview. 📄 Professionally hedged. Specific scenarios. Implied probabilities. Here's what the note didn't say: Every projection in it was already trading on a prediction market with a specific dollar probability attached. 🔴 → Boeing announcement: 86% on Kalshi → Trump mentions Iran: 61% on Polymarket → Soybean purchase: 36% on Polymarket → AI/chips mention: 54% → Even handshake duration (8.5s) was being traded Goldman wasn't the only one. Wolfe Research and Jefferies published notes that tracked just as closely. CNBC openly cited Kalshi probabilities — with a footnote disclosing their commercial relationship. Wall Street has crossed a line nobody is reporting: Institutional research desks are now pricing off prediction market data. Sometimes with attribution. Often without. 👀 Five structural conditions converged for the first time in 2026: 🧵 → Volume crossed credibility threshold ($23B+/mo) → Regulatory clarity emerging (CFTC, partnerships, disclosures) → Granularity matches institutional needs (117 markets on a single summit) → Commercial relationships forming openly (CNBC + Kalshi) → Independent data infrastructure becoming accessible This is how alternative data becomes primary data. Satellite imagery, credit card transactions, web scraping — all followed the same five-stage arc. Prediction market data is at Stage 4: "embedded reference." Next stage is primary input. 📍 Here's the quiet part nobody on Wall Street wants to say out loud: 🔮 Prediction markets are no longer a competitor to institutional analysis. They are a partial replacement for it. When 117 markets exist on a single summit and trade $36.7M in volume, the question stops being "what does the analyst think will happen?" — and starts being "what does the analyst know that the market doesn't already price?" For most outcomes, most of the time, the honest answer: not much. The analysts who add real value going forward will be those who understand market structure deeply enough to identify where prediction markets are systematically wrong — thin liquidity, biased resolution rules, regulatory constraints. That requires understanding the data layer at a level most institutional analysts don't have yet. 🏗️ The institutional research function isn't being eliminated. It's being restructured. The new function looks like "interpretation of market signals" — not "production of market opinions." Different job. Different tools. Different data infrastructure. That's what we're building at Assymetrix. 🔮 📖 Full Intelligence Brief: https://lnkd.in/d7YTUSr9 #PredictionMarkets #WallStreet #AlternativeData #InstitutionalResearch #Polymarket #Kalshi