Markets evolve quickly, but the underlying infrastructure evolves even faster. This week’s roundup explores the ideas shaping the next phase of crypto and AI, from tokenized equities and DeFi lending to Physical AI, prediction markets, memecoins, and AI agents. Here are some of the most insightful articles I came across this week. — @stacy_muur explores what makes equity perpetuals viable beyond 24/7 access, stating that deep liquidity and efficient execution determine a good trading experience. The article highlights how @OndoPerps improves capital efficiency by letting tokenized stocks serve as collateral, making them more than just assets held in a wallet. However, users must also understand the added risks that come with leverage and collateral management. x.com/stacy_muur/status/2075… — @Gyokeres_eth shares a practical guide to using @arbital_xyz’s market-making bot to automate trading volume on Perpl, for more efficient farming of rewards. The goal is to keep trading costs low while taking advantage of @perpltrade’s ongoing incentives. While MM bots can simplify perp farming, managing costs, monitoring positions, and only participating when rewards outweigh risks are of utmost importance. x.com/gyokeres_eth/status/20… — @MahoneDeFi explains why tokenized stocks need lending infrastructure built for more than simple collateral, arguing that current DeFi models fragment liquidity and limit capital efficiency. @aave V4’s Hub-and-Spoke architecture separates risk management from liquidity while supporting margin lending and securities lending within the same system. Aave V4 is positioned to scale onchain securities finance without sacrificing liquidity or risk controls. x.com/mahonedefi/status/2077… — Physical AI is expanding far beyond humanoids, with funding flowing into AI models, robot manufacturers, hardware, training data, and real-world deployments. @muskan_kalra24 highlights companies like @Figure_robot as leaders across different parts of the ecosystem. While humanoids have the spotlight, the biggest commercial success today still comes from specialized robots, with long-term winners likely emerging from the companies building the full Physical AI stack. x.com/muskan_kalra24/status/… — @Tanaka_L2 examines @OndoPerps’ early performance, highlighting how using tokenized stocks directly as collateral improves capital efficiency for onchain derivatives trading. The platform gained early traction with strong trading volume and open interest, particularly in commodities, while positioning tokenized assets for more active use. Although the model is promising, long-term success depends on broader liquidity, sustained user activity, and performance beyond incentive-driven trading. x.com/tanaka_l2/status/20766… — @imlaurieowen explains the essence of storytelling for VCs and how great fund narratives focus on long-term vision and market shifts, not short-term execution. The article outlines 4 storytelling frameworks, including Contrast, Insight, Solution, and Path, and how they help funds communicate a clear point of view and stand out. The strongest VC stories consistently combine character, conflict, stakes, and a compelling vision of the future. x.com/imlaurieowen/status/20… — Memecoins have become crypto’s most effective user acquisition tool, helping new chains attract liquidity, onboard users, and build early network activity. @Eli5defi cites @solana, @base, and @RobinhoodApp’s Chain as examples of how meme-driven demand can evolve into broader DeFi and infrastructure growth. While memes are powerful for bootstrapping ecosystems, long-term success depends on converting speculative activity into lasting utility and user retention. x.com/eli5defi/status/207695… — @lzminsky explains hedging in prediction markets and how different it is from a trader’s perspective of hedging. While prediction markets can price and transfer event risks, they cannot automatically serve as effective corporate hedges without proper structuring, risk matching, and documentation. Ultimately, prediction markets are valuable financial infrastructure, but adopting a one-size-fits-all approach could undermine their long-term credibility. x.com/lzminsky/status/207678… — @meta_alchemist introduces a framework for turning AI workflows into self-improving systems, arguing that most AI loops automate tasks without getting better over time. He outlines 5 stages of AI loops, using memory, benchmarking, independent evaluation, and iterative testing to drive continuous improvement. Effective AI workflows don’t just repeat tasks; they learn from past mistakes and evolve with every iteration. x.com/meta_alchemist/status/… — DeFi lending is becoming commoditized, with borrow rates and protocol margins converging across major platforms, making scale the key differentiator. @DefiVaults argues that @aave is best positioned with a profitable integrated model, while @Morpho is sacrificing fees to gain market share and bet on future monetization. As the market matures, the biggest advantage may come from controlling liquidity, stablecoins, and user distribution rather than lending margins alone. x.com/defivaults/status/2077… — That’s it for this week. Stick around for more insightful articles next week.