Nicki argues Robinhood executed a classic platform playbook against Kalshi: partnering to validate prediction markets demand, then building competing infrastructure through Rothera Exchange once the market proved real. Kalshi cleared $22.9B in 2025 and $24B+ quarterly by Q1 2026, reducing Robinhood's share from 60% to roughly 25% of volume. The lesson: infrastructure builders must develop defensible moats like liquidity depth and institutional credibility before distribution partners capture the economics, or face the dependency becoming leverage.

Eric Liu shows how prediction markets can reduce parlay collateral requirements by 10-70% depending on portfolio composition using Integer Linear Programming, which identifies the worst-case loss scenario across correlated markets instead of collateralizing each bet in isolation. MMs currently reserve capital for impossible outcome combinations—like BTC closing both above and below $100K simultaneously—but ILP solves this by finding the actual maximum loss across all possible market resolutions in milliseconds. The result tightens quotes and enables deeper liquidity without sacrificing the fully-collateralized guarantees peer-to-peer settlement requires.

Prediction markets like Kalshi and Polymarket have devolved into sports betting platforms, with ~65% of volume in sports over the past year, because they lack the market structure to support higher-value applications—sharps won't trade without uninformed gamblers, and gamblers prefer short-duration sports contracts. Aelix argues AI agents solve this by functioning as cheap, forced-participation sharps that dramatically lower minimum viable liquidity, enabling micro-markets and private institutional forecasting that could finally unlock the original vision of prediction markets as truth machines, though it remains unclear whether markets retain their current form in an AI-dominated future.

Taetaehoho analyzed liquidity rewards on Polymarket and sponsorships on Kalshi from February to May 2026, finding they only move top-of-book liquidity when daily spend exceeds 1% of existing book depth—below that, median programs show no effect. Even at higher intensities, incentive size poorly predicts actual liquidity response; pre-existing conditions like spread width matter more. The thesis: prediction market liquidity requires structural innovation beyond rewards alone.

taetaehoho compares sportsbook and prediction market pricing across identical events, finding that liquid prediction markets offer 100-300 bps better prices than sportsbooks even after accounting for 150-175 bp fees, but de-vigged sportsbook odds match prediction market prices, suggesting counterparty information and last-look advantages tighten spreads more than maker competition does. Long-tail markets on Polymarket and Kalshi suffer 10-50% spreads versus <$1,000 volume, indicating anonymity and market immaturity create depth problems sportsbooks have solved at scale.

Kunal argues prediction markets have become volatility playgrounds where short-duration 5min and 15min crypto markets now generate ~40% of Polymarket's daily fees despite comprising only 16% of volume, with professional bots capturing consistent 1.1%-1.6% margins while retail traders lose ~$500 on average per address. Kalshi's crypto share jumped from 9% to 46% since January, and Hyperliquid's upcoming 15min BTC markets threaten fee compression, making expansion into non-crypto volatile assets and better product execution critical for maintaining leadership.

mph argues that Polymarket and Hyperliquid's HIP-4 are not competing for the same pie — Polymarket targets retail through TV ads and street activations while Hyperliquid's user base is already inside the crypto bubble, leaving room for both to thrive. Polymarket's announcement of perps directly escalates the rivalry into Hyperliquid's core territory, but mph expects the incumbent to hold the perps edge for the foreseeable future. Fragmentation across prediction markets ultimately benefits the sector, and an aggregated trading layer will matter more than any single native UI long-term.

Baheet argues that prediction markets' $6.5 billion weekly volume masks a structural problem: 99% sits in politics, sports, and crypto while thousands of long-tail markets barely exist because infrastructure can't support them. AMMs fail due to inevitable impermanent loss at resolution; CLOBs require professional market makers (23 at Kalski, top three providing 70% of election liquidity) and ignore unprofitable niche markets. Melee's parimutuel market maker solves this by using bonding curves per outcome, enabling cold-start liquidity without intermediaries while allowing creators to launch permissionless markets and capture fee revenue—unlocking the $100 billion in passive DeFi capital currently locked out.

Petro argues HIP-4, activated by Hyperliquid on May 2, is not a Polymarket clone but a new outcome primitive that settles on-chain in $USDH with cross-margin integration across perps and spot markets. While Polymarket and Kalshi printed $22B volume in April with on-chain prediction markets two orders of magnitude smaller, HIP-4's permissionless deployment via 1M $HYPE staking, unified margin mechanics, and end-to-end on-chain settlement differ fundamentally—though unresolved questions around close-side fee schedules, portfolio margin rollout, and non-curated builder deployment in Phase 2 will determine whether the market actually wants it.

Michael's response to the CFTC's March 2026 ANPR on prediction markets argues for a *multidimensional* public-interest framework instead of treating all event contracts identically. Four dimensions: (1) **information structure** — markets where outcomes emerge from dispersed knowledge (elections, FOMC) enable Hayekian price discovery; concentrated/low-legibility markets (e.g. "what phrase will the CEO say") collapse into pure access trading. (2) **manipulation economics** — does the contract create incentives to *cause* the outcome rather than predict it? Cites Brian Armstrong's Oct '25 Coinbase earnings-call mention market and P2P.me trading on its own fundraise. (3) **social utility of the price signal** — pandemic/climate/election markets serve public decisions; hyperspecific individual-behavior contracts don't. (4) **repugnance** — Alvin Roth's framework: some markets degrade something morally significant regardless of manipulation (terminally-ill timing markets, nuclear-detonation contracts). Reframes "insider trading" as three distinct patterns calling for different remedies: outcome influence (fix via market design, not surveillance), duty breach (the Polymarket Maduro-strike case — misappropriation framework applies), and information advantage without breach (the price-discovery engine — restricting it would erode what the CEA was written to protect). Third argument: **resolution integrity is load-bearing**. Event contracts have no external reference price. Three failure modes: rule mutability after listing (Polymarket's '24 government-shutdown contract — resolution language added Dec 20, odds spiked 20%→98%, no shutdown actually occurred), undefined rule hierarchy (Venezuela election overridden via UMA vote despite "primary source" language), single-source oracle vulnerability (Paris-CDG temperature sensor, suspected hairdryer attack, ~$34K in payouts). Whenever resolvers can also hold positions, the incentive to influence resolution is structural. Recommends: original specs as complete reference document, fixed resolution-source hierarchy at certification, cost-of-corruption assessment for single-signal markets.