Sunny

← All weeks

What I’ve been reading (W37, 2025)

Doing things with AI, tips to solve hard problems, and crypto developments

  1. AI has decoupled doing from learning. In the past, the act of building or researching something forced you to understand it along the way. Now, with AI, you can quickly reach working results without actually deepening your knowledge. That convenience creates a gap: you get the output, but not the intuition. Writing is the antidote to it. Writing forces you to confront gaps in your knowledge, reflect on what you actually know, and retain the intuition that would otherwise have been built during the slower, pre-AI process. It turns passive consumption of AI outputs into active reinforcement learning. Without that step, you risk becoming efficient but shallow - producing results without growing smarter or faster in the long run https://www.youngmoney.co/p/write-it-down-write-it-down-write
  2. Some tips to solve hard problems https://sharif.io/solving-hard-problems a. Look for similar problems that’ve already been solved, and try to find an analogy to your problem b. Restate the problem in as many ways as possible and try to look at it from many different perspectives c. Generalise the solution to an already solved problem and see if it can apply to your problem d. The best way to solve a hard problem is by understanding the system and the principles that govern it
  3. Stablecoins enable instant on-chain transfers that avoid liquidity traps, reduce high card fees, ensure more reliable cross-border payments, and provide direct access to USD-based financial systems. They make money movement easier, faster, and more efficient for many businesses https://x.com/patrickc/status/1964119986227806502/photo/1, https://tempo.xyz/
  4. Apple’s products keep improving — beautifully engineered iPhone Air, AirPods Pro 3 and more — but the company has chosen safe profits over bold innovation. That was fine when smartphones drove the frontier, but now the frontier has shifted, and Apple is farther than they have been in years from changing the world https://stratechery.com/2025/iphones-17-and-the-sugar-water-trap, all new releases here https://www.apple.com/newsroom/
  5. MCP works best when tools reflect complete user goals. LLMs do not manage state the way developers do, so building tools around workflows produces better results https://vercel.com/blog/the-second-wave-of-mcp-building-for-llms-not-developers