Weekend relief

Brought to you by:

Hey all, here is some weekend reading of interest, covering the IMF’s research on consensus mechanisms, AI models for predicting financial market stress, decentralized AI training, and crypto-native market making. Additionally, check out Tuesday’s podcast on Plasma, Pumpfun, and Aster. Have a great weekend!

— Luke

Rough week, huh? The second half of the week brought broad weakness across the board in risk assets, with equity indices reversing most of this week’s gains and crypto majors making new lows. Similarly, BTC broke recent lows, trading as low as $108,750 and reversing most of its September gains. ETH and SOL notched fresh lows, as well. The notable winners this week were Gold, which held flat, and Plasma’s XPL, which repriced over 50% higher on launch. 

Blockchain Consensus Mechanisms: A Primer for Supervisors

• The IMF published a working paper titled “Blockchain Consensus Mechanisms: A Primer for Supervisors (2025 Update).” This paper provides an overview of key blockchain consensus mechanisms—including Bitcoin’s Proof-of-Work (PoW), Ethereum’s Proof-of-Stake (PoS), and Solana’s hybrid Proof-of-History (PoH) + Tower BFT. It explores how these mechanisms impact supervisory goals like operational resilience, governance, market integrity, and settlement finality. It also surveys Layer 2 scalability solutions (e.g., rollups, state channels, sidechains), highlighting new risks related to slashing, validator concentration, staking-as-a-service, and restaking leverage. The paper stresses the need for supervisory familiarity with consensus design to mitigate financial stability risks as blockchain adoption in regulated markets grows. Read more.

Harnessing artificial intelligence for monitoring financial markets

• The Bank for International Settlements published a working paper titled “Harnessing artificial intelligence for monitoring financial markets.” The authors propose a hybrid approach combining a recurrent neural network (RNN) with a large language model (LLM) to forecast and interpret financial market stress. They focus on deviations from triangular arbitrage parity in the EUR‑JPY pair, using the RNN to generate interpretable, time‑varying weights on input variables. The weights guide the LLM to search for narrative explanations around high‑signal variables. In tests, the model successfully anticipates periods of market dysfunction (such as before March 2023 banking turmoil) and provides contextualized signals for regulators. Read more.

Decentralized AI Training: Architectures, Opportunities, and Challenges

• Lucas Tcheyan and Arjun Yenamandra of Galaxy published a research report titled “Decentralized AI Training: Architectures, Opportunities, and Challenges.” The report surveys emerging decentralized AI training protocols (e.g. Nous Research, Prime Intellect, Pluralis, Templar, Gensyn), outlining their technical designs, incentive schemes, and verification strategies. It highlights key challenges: communication overhead, correctness verification, compute heterogeneity, and tokenomics. The authors argue that decentralized training is transitioning from conceptual to operational, but further breakthroughs in optimization, governance, and trust layers are needed before matching centralized AI capabilities. Read more.

Adventures in Spot Market Making

• BlockBandit published an article titled “Adventures in Spot Market Making.” In it, the author chronicles their experience building a spot market-making bot on Aster: covering strategy design, parameter tuning, order execution, and performance tracking. They delve into challenges like inventory risk, spread management, adverse selection, and volatility. The post offers lessons learned (e.g. dynamic spread adjustment, position limits), real results data, and reflections on what works vs. what doesn’t in automated market making on decentralized exchanges. Read more.

Happy reading!

— Luke

Brought to you by:

Katana was built by answering a core question: What if a chain contributed revenue back into the ecosystem to drive growth and yield?

We direct revenue back to DeFi participants for consistently higher yields.

Katana is pioneering concepts like Productive TVL (the portion of assets are actually doing work), Chain Owned Liquidity (permanent liquidity owned by Katana to maintain stability), and VaultBridge (putting bridged assets to work generating extra yield for active participants).

On Tuesday’s 0xResearch livestream, we covered:

• Aster’s emergence as a Hyperliquid competitor: We examined Aster's rise following its token launch, noting its early $170–200M market cap and a 24h revenue figure (~$4.2M) that exceeded Hyperliquid’s (~$2.1M), though the product still lagged on a performance basis. Participants attributed success to wealth effects from BB token appreciation, targeted token distribution, and CZ’s strategic influence via Yeezy Labs and family office backing.

• CZ’s influence and market positioning: We noted that CZ played a central role in Aster’s positioning, with over 95% of supply controlled by insiders, and strategic tweets moving token prices (e.g., +67% following a tweet). His approach mimicked Hyperliquid’s success formula (requiring deposits for token access) while layering in retail incentives such as airdrops tied to user retention.

• Fleeting liquidity risks for perps DEXs: We explored the mercenary behavior of high-net-worth retail traders in perps markets, emphasizing their lack of allegiance and fast rotation to new venues (e.g., Aster, Lighter, Fogo) with better incentives or listings. This dynamic challenged Hyperliquid’s network effects and highlighted the need for stickier features like Builder Codes integrations (Phantom, Axiom).

• Token liquidity and listing strategy as user acquisition: We drew attention to how exclusive listings drive user traffic, citing Aster’s dominance in HYPE token trading due to early listing and 300x leverage. Similarly, Hyperliquid’s early HYPE listing helped it capture market share, reinforcing that exchange-specific token access serves as a core product differentiator.

• Pump.fun monetization and buyback model: We analyzed Pump’s revenue model, which included buying $1.2M of PUMP daily via 100% revenue buybacks, leading to a 7.2% supply repurchase to date. Creator rewards peaked at $4.3M/day before normalizing. The team acknowledged the buyback policy was short-term optics-driven rather than capital efficient, intending to signal token alignment.

• Plasma’s stablecoin chain thesis and early traction: We examined Plasma’s strategy of offering gas-free USDT transfers as a loss leader to bootstrap $2B+ in TVL, positioning itself as a stablecoin-centric DeFi chain. The launch of Plasma One (neobank) with 17k+ users on waitlist and partnership with Yellow Card illustrated efforts to target high-inflation, underbanked markets with peer-to-peer rails and USD-denominated yields.

• Debate over HYPE supply structure and FDV accounting: We reviewed the Jon/Hasu proposal to eliminate HYPE’s fixed cap and burn 45% of the supply to improve accounting clarity and align with equity-style market cap conventions. Participants debated the tradeoff between optionality (e.g., future community rewards) and clean capital structure presentation, noting the historical inconsistency of FDV data across crypto data providers.

Look for the full podcast on Spotify, Apple Podcasts and X.

Crypto is changing TradFi derivatives and rate markets as we know them.

DAS: London will feature all the builders driving this change. 

Get your ticket today with promo code: 0X100 for £100 off.

📅 October 13-15 | London