Article: Ethereum Tokenomics Insights for Web3 Entrepreneurs, a Quantitative Study
Date: 2025-01-26
Publisher: Zishan Ashraf Mohammad; Nicolas Harkiolakis, PhD (École des Ponts Business School)
Score: ₿₿₿
Read time: 50-60 min
Summary:
- Overview: This paper studies how Ethereum’s token design relates to the ETH price over time. It uses bi-weekly data from August 2021 to September 2025, starting after the EIP-1559 burn upgrade. The study tests supply changes, token burns, staking and liquidity, network fees, and wealth concentration. Results show that ETH price is mainly linked to real network use and locked capital. Supply expansion lowers price pressure, while burns and fees support price. Wealth concentration does not show a meaningful price effect.
- Liquidity And Utility: Total value locked (TVL) has the strongest positive relationship with ETH price. Higher TVL reflects real usage and long-term commitment to the network. This suggests value comes from capital staying in the system.
- Supply Expansion Effects: Token unlocks are negatively related to ETH price. When new supply enters the market, price pressure increases. Careful control of issuance is important for stability.
- Fees And Burns: Gas fees and burned tokens both show positive links to price. More activity leads to higher fees and more burns. This ties network usage directly to supply reduction.
- Governance Concentration: The Gini coefficient does not show a significant link to ETH price. Ownership inequality does not appear to drive short- or medium-term pricing. This result assumes transparent and observable on-chain data.
Article: A Control Theoretic Approach to Decentralized AI Economy Stabilization via Dynamic Buyback-and-Burn Mechanisms
Date: 2026-01-15
Publisher: Zehua Cheng; Wei Dai; Zhipeng Wang; Rui Sun; Nick Wen; Jiahao Sun
Score: ₿₿₿
Read time: 35-40 min
Summary:
- Overview: This paper studies how to stabilize token economies in decentralized AI networks. It argues that static buyback-and-burn rules increase volatility and fail during market stress. The authors propose a Dynamic-Control Buyback Mechanism (DCBM) using control theory. The system adjusts buybacks continuously based on price deviations and treasury health. Agent-based simulations show lower volatility and lower operator churn than existing methods. The study focuses on long-term sustainability rather than short-term price support.
- Problem With Static Models: Most token buyback systems use fixed or threshold-based rules. These rules react too late or too strongly during market swings. The paper shows they behave pro-cyclically and worsen downturns.
- Dynamic Buyback Design: The DCBM treats the token economy as a control system. A PID controller adjusts buybacks using price trends rather than spot prices. Treasury spending is strictly capped to prevent insolvency.
- Stability And Solvency: The model guarantees the treasury never fully depletes. Buybacks shrink automatically when funds are low. This creates counter-cyclical behavior that saves reserves in bull markets and supports prices in downturns.
- Simulation Results: In high-volatility scenarios, DCBM reduces price volatility by about 66%. Operator churn falls from roughly 19.5% to 8.1%. The system also shows stronger resistance to manipulation than threshold-based approaches.
Article: Tokenomics: Quantifying Where Tokens Are Used in Agentic Software Engineering
Date: 2026-01-20
Publisher: Mohamad Salim; Jasmine Latendresse; SayedHassan Khatoonabadi; Emad Shihab
Score: ₿₿
Read time: 20-25 min
Summary:
- Overview: This paper studies how tokens are consumed in multi-agent systems used for software engineering. It analyzes 30 software tasks executed by the ChatDev framework using a GPT-5 reasoning model. The authors map internal agent phases to standard software development stages. They measure how many input, output, and reasoning tokens are used at each stage. Results show that most tokens are spent during review and refinement, not initial coding. The work helps explain where costs come from in agentic software workflows.
- Code Review Costs: The code review stage uses the most tokens across tasks. On average, it accounts for about 59.4% of total token usage. This shows that iterative refinement is the main cost driver.
- Input Token Dominance: Input tokens make up the largest share of usage at about 53.9% overall. Agents repeatedly pass large contexts to each other during collaboration. This creates a high communication overhead.
- Stage Cost Profiles: Different development stages show different token patterns. Coding is output-heavy, while review and documentation are input-heavy. These profiles help predict costs by task type.
- Practical Implications: Token costs mainly come from verification, not creation. Teams can reduce costs by limiting repeated reviews or adding human checkpoints. The findings guide more efficient agent design.
Article: Economic Censorship Games in Fraud Proofs
Date: 2025-02-27
Publisher: Ben Berger; Edward W. Felten; Akaki Mamageishvili; Benny Sudakov
Score: ₿+
Read time: 35-45 min
Summary:
- Overview: This paper studies how economic censorship can affect fraud proofs in optimistic rollups. It focuses on attackers who bribe block proposers to delay or block honest challenge transactions. The authors model fraud proofs as multi-round games between an attacker and a defender, each with limited budgets. The goal is to understand how long the challenge period must be to ensure honest success. The analysis covers single-proposer, special-round, and multi-proposer settings. Results give clear budget and time conditions for rollup security.
- Economic Censorship Model: The attacker spends money to censor transactions instead of controlling consensus. The defender spends money to get transactions included. The challenge period creates economic pressure on the attacker over time.
- Budget Ratio Thresholds: The paper derives exact conditions linking attacker and defender budgets. If the attacker’s budget is large enough relative to the defender, censorship can succeed. Otherwise, the defender is guaranteed to win the fraud proof.
- Special And Multiple Proposers: Some blocks are harder to censor due to default inclusion rules. These “special rounds” increase attacker costs. With multiple proposers per block, censorship becomes even more expensive.
- Design Implications: Long challenge periods protect against economic censorship. Defender costs can be much lower than attacker costs under realistic settings. The results justify week-long challenge periods used by major rollups.
Article: Selfish Mining under General Stochastic Rewards
Date: 2025-08-08
Publisher: Maryam Bahrani; Michael Neuder; S. Matthew Weinberg
Score: ₿+
Read time: 45-55 min
Summary:
- Overview: This paper studies selfish mining when miners earn rewards from more than just fixed block issuance. It shows that modern incentives include transaction fees and unpredictable extra rewards such as MEV. The authors build a general model where rewards can vary over time and be random. They analyze how miners change behavior when optimizing for total rewards instead of block rewards alone. Results show selfish mining becomes profitable at much lower hash power under realistic reward mixes. This means current security assumptions may be too optimistic.
- General Reward Model: The paper introduces a reward model that includes fixed rewards, time-based fees, and random extra rewards. These rewards are treated as stochastic but structured. This better reflects real miner incentives today.
- Static Reward Properties: The authors define conditions under which rewards depend only on time since the last block. Many real rewards, including linear transaction fees and MEV spikes, fit this definition. This makes the analysis tractable while staying realistic.
- Lower Attack Thresholds: When block rewards and fees are combined, selfish mining becomes profitable with about 25% hash power. With additional random rewards, the threshold can drop below 20%. This is much lower than the classic one-third assumption.
- Protocol Design Impact: Security depends on the full reward mix, not issuance alone. As block rewards shrink, MEV-like rewards grow in importance. Protocols must account for these incentives to remain secure.
