made sam’s aquarium real with @cursor_ai

designing in layers the best software doesn’t overwhelm you on day one or abandon you when you get sophisticated. it grows with you through carefully designed “layers” that reveal themselves exactly when you need them. think of it like learning to drive. you start with acceleration, brake, steering. you don’t need to know how the engine works or aerodynamics. but as you get more curious about cars, those deeper systems are always accessible, within the same car. at Notion, we built this philosophy into everything. you could start typing and immediately have a working doc. then add a table. then views. then automations. Cursor does this beautifully too — starts as a code editor that just works, then you discover Tab, then Agents, multiple models, then advanced rules & team workflows. each layer reveals itself naturally as you code more. most software gets this wrong — they either dump everything on you at once, or kill the advanced stuff so you can’t do things your way. the real challenge isn’t designing each layer, it’s designing the transitions between them. that moment when someone outgrows the defaults shouldn’t feel like hitting a wall. it should feel like discovering a secret door that was always there. alan kay once said users should be able to “open up the hood” of their word processor and tinker with it. when you design in layers, you’re building that kind of learning environment where users can discover their own potential. the goal isn’t to hide complexity — it’s to sequence its reveal. give people exactly what they need when they need it, and trust them to grow into the full power, pushing the boundaries your concepts in ways you’ve never imagined before.

on not changing for the sake of change: the best design doesn't announce itself. it just works, quietly doing what it needs to do without asking for applause. when we redesign something, the question isn't "how can we make this look different?" it's "how can we make this genuinely better?" better means the person using it gets to their goal faster, with less friction, with more delight in the small moments. UI should disappear. when someone is writing, they shouldn't think about the text editor. when someone is paying for coffee, they shouldn't think about the payment flow. the interface becomes invisible when it's doing its job right. constraints are not limitations—they're the foundation of good design. the 140 character limit made twitter what it was. the grid made swiss design timeless. the rectangle made the iphone possible. subtlety is harder than flash. it takes restraint to not add that extra animation, that additional color, that one more feature. but restraint is what separates craft from decoration. too many designers optimize for screenshots instead of experiences. they chase the latest trend instead of understanding the core problem. they simplify for the sake of simplifying instead of clarifying what matters. good design has soul because it serves people, not egos. it emerges from deep understanding of the problem, not surface-level aesthetics. it gets better with use, not worse. the best compliment you can get is when someone says "i didn't even notice the design—it just worked perfectly." that's when you know you've made something good.

Using Cursor well = fast, clean code. Using it wrong = AI spaghetti you’ll be cleaning up all week. Here’s how to actually use it right: 1. Set 5-10 clear project rules upfront so Cursor knows your structure and constraints. Try /generate rules for existing codebases. 2. Be specific in prompts. Spell out tech stack, behavior, and constraints like a mini spec. 3. Work file by file; generate, test, and review in small, focused chunks. 4. Write tests first, lock them, and generate code until all tests pass. 5. Always review AI output and hard‑fix anything that breaks, then tell Cursor to use them as examples. 6. Use @ file, @ folders, @ git to scope Cursor’s attention to the right parts of your codebase. 7. Keep design docs and checklists in .cursor/ so the agent has full context on what to do next. 8. If code is wrong, just write it yourself. Cursor learns faster from edits than explanations. 9. Use chat history to iterate on old prompts without starting over. 10. Choose models intentionally. Gemini for precision, Claude for breadth. 11. In new or unfamiliar stacks, paste in link to documentation. Make Cursor explain all errors and fixes line by line. 12.Let big projects index overnight and limit context scope to keep performance snappy. Structure and control wins (for now) Treat Cursor agent like a powerful junior — it can go far, fast, if you show it the way.