Presented by

Settlement Solution - MoneyGram taps Fireblocks to expand stablecoin use in global payments and treasury ops.

Solana Giant - Solmate to buy RockawayX in all-stock deal to build $2B institutional Solana giant.

Staking Expansion - Galaxy looks to expand institutional staking infra with Alluvial acquisition.

New here? Ask Copilot up to 5 free queries per day with Messari Basic.

This week’s episode dives into the decentralized future of AI with Ivan Nikitin, co-founder of Fortytwo, a crypto-powered AI network that uses swarm inference to unite thousands of small, specialized models into a single, ever-improving system.

In line with findings from Messari’s newly released State of AI Report, Fortytwo challenges the unsustainable trajectory of centralized compute by enabling AI to scale across consumer hardware—from laptops to dedicated devices and even mobile phones. Ivan breaks down how peer-ranking improves accuracy, how custom datasets like Fortytwo’s Rust benchmark are created, and why decentralized inference is emerging as a serious competitor to frontier models.

The conversation also previews Fortytwo’s roadmap across research APIs, node participation, pricing, and future tokenized governance—making this a must-listen for anyone tracking the intersection of AI, crypto, and distributed compute.


Watch the full discussion on YouTube, Spotify, or Apple Podcasts.

Messari's protocol reports give you a deep dive on the foundation and state of top crypto protocols, including key metrics and notable events. See the complete list of protocol reports here and get a preview of our latest report below.

  • Layer-1 smart contract blockchain using Liquid Proof-of-Stake (LPoS) for consensus

  • Token: XTZ (tez) is the native asset used for staking (“baking”), delegation, governance, and fees on L1 and on the Tezos-based L2

  • Tezos L1 transactions rose 21.5% QoQ; contract calls also increased, signaling healthier base-layer activity

  • L1 fees climbed 16.9% QoQ to 17,460 XTZ, while Etherlink L2 fees fell 36.7% QoQ to 50,220 XTZ as its Kernel 4.1 upgrade doubled gas throughput and lowered effective costs

  • VeChainThor is a Layer-1 smart contract platform launched in 2017, focused on real-world enterprise and sustainability use cases

  • Daily active addresses: +85.2% QoQ to 62.8k; daily new addresses: +54.4% QoQ to 40.9k

  • +32.3% QoQ to 370k, ending a two-quarter decline and reflecting broader onchain usage

  • ~81.5% of gas now from EVM-based transactions, with EVM gas consumption up ~51% QoQ

  • High-performance L1 for real-time apps: Somnia targets internet-scale throughput, sub-second finality, and ultra-low fees to support gaming, social, and interactive digital worlds

  • After a 6-month testnet (10B+ txs, 118M wallets), launched mainnet and the SOMI token, ending Q3 with a $112M market cap

  • Averaged 14.3M daily transactions in September and generated 199k SOMI in fees, showing meaningful retention beyond airdrop farming

From Our Sponsor

Kalshi’s latest raise valued the company at roughly $11 billion, making co-founder Luana Lopes Lara the youngest self-made female billionaire in history. The company also recently tokenized all of its event markets on Solana, allowing them to be held, traded, or composed with broader DeFi. However, Kalshi still operates as a regulated and centralized venue. 

functionSpace, a new startup in the space, by contrast, is explicitly positioning itself as an open developer framework that lets any application embed prediction capabilities without running a venue, managing liquidity, or operating a resolution system. The team splits every market into two linked surfaces:

  • A Probability Market: a continuous forecasting surface where users trade probability distributions (not binary shares).

  • A Reality Market: an adversarial, capital-locked resolution mechanism that determines the final truth value.

Why is this interesting?

Well, today, prediction markets generally settle in one of two ways (i) Kalshi’s centralized adjudication or (ii) Polymarket’s oracle-driven dispute process. Specifically, Kalshi resolves outcomes internally by referencing official sources (i.e., government data or certified news outlets). Once finalized, it settles contracts at $1 or $0, a predictable system, but one that is entirely dependent on a centralized adjudicator and typically requires 24 to 48 hours to finalize. Polymarket, on the other hand, uses decentralized oracles. Event markets typically rely on UMA’s Optimistic Oracle, where a proposer posts a bonded assertion and disputes escalate to UMA tokenholder voting. This model is permissionless and fast when uncontested, but governance-based disputes can extend settlement by several days and introduce external trust assumptions.

Like Polymarket’s UMA-based flow, functionSpace begins with participants posting bonded assertions about the correct outcome; a similar “put capital behind your claim” mechanic. However, instead of relying on an optimistic proposal window and a governance-based dispute escalation, “reality markets” add a second role where participants who believe an assertion is incorrect or premature can post bounty capital to oppose or delay settlement (i.e., continuance agents). Resolution occurs only when the “closure” capital outweighs “continuance” capital by the required margin for a sustained period, turning settlement into an adversarial capital equilibrium rather than a yes/no oracle dispute. The final settlement value is the capital-weighted median of surviving assertions, and assertions outside a statistically defined band are slashed. 

Whether adversarial resolution can consistently attract liquidity and defend against manipulation remains an open question, but it's a clear attempt to redesign settlement as a transparent, onchain economic process rather than an external dependency.

Keep Reading