had a great meeting with Chancellor Scholz this morning and a fun talk at TUM this afternoon, and finally enjoying a quiet moment in a beautiful city

Unpopular opinion: Using a monolithic codebase can be more efficient than microservices.

Evolution of the bicycle.

There’ll be a fuzzy & heavily litigated line between training an AI and copyright infringement. Is it a fuzzy interpolated search restating data from Yelp, Quora, Stack Overflow? Or remixing a thousand copyrighted artists, coders, authors? Or is it learning, thinking, creating?

There's _always_ an end customer. If your client is in the B2B space, you must talk to both them and _their_ customers to understand what's important to build next. You are not serving your client if they fail because _their_ customers are unhappy.

1. Small engineering teams move much faster than big teams. Obvious, but it still needs to be said. Elon has factored this in. 2. If you are willing to throw out large chunks of a system, you can rebuild the core parts in a fraction of the time with great people.

Advertisers should support:

Also, to head off the enviable, I'm not talking about a quick sketch just prior to implementation, and no, it's not cheaper to design on paper. I've seen users approve the paper and hate the implementation too many times. That initial paper phase is a complete waste of time. 3/3

You've been speaking English your entire life but did you know that the words you speak are all created from 26 individual letters? Here are the 26 letters that you MUST know:

In 1663, the partial fossilised skeleton of a woolly rhinoceros was discovered in Germany. This is the “Magdeburg Unicorn”, one of the worst fossil reconstructions in human history.