Design tokens are only useful if the team actually uses them consistently. The best design system breaks down when engineers hardcode values because it's faster in the moment.
I asked Claude to review my onboarding flow and identify where users were most likely to drop off based on the copy and UX description I gave it. It found three real issues. Zero bias, instant feedback.
The more I build, the more I believe distribution beats product. A great product with no distribution is a private project. An okay product with great distribution is a business.
What's the most useful resource you've found for building products? Not the famous ones everyone recommends. The one that actually changed something for you.
I've started treating my AI tools the same way I treat my team. Clear briefs, clear expectations, clear feedback when the output misses. Better inputs, consistently better outputs.
Fixed Carloop's mobile performance this week. Load time went from acceptable to fast. The bounce rate improvement was immediate. Speed really is a feature.
Claude handles long-context reasoning better than anything else I've tried. When the problem needs sustained thinking across a lot of information, it's in a different league.
Something I wish I knew earlier: the first version of your pricing is just the opening bid in a long negotiation with the market. Don't be precious about it.
Hot take: most product analytics setups track the wrong things. Founders measure what's easy to measure, not what actually predicts whether the product is working.
Agree?