When in doubt, decouple.

The easier they make it look, the harder they worked to get there.

Something about trying to build computers to work like the human brain when we don’t even really understand the human brain feels wrong.

My "Vanilla Roam" course is now available for free as a YouTube playlist. 58 bite-sized lessons covering how to get the most out of Roam Research using only its native features. #roam #toolsforthought youtube.com/playlist?list=…

Generalize, don’t memorize.

Spoke with a friend in #AI the other day. Said there’s three kinds of people out there: - those who are working on AI - those who are digging in and learning about AI - those who have no idea what’s about to hit them

I realized tonight that prompt engineering is really just about learning how to have effective conversations with an LLM. Then it struck me how many people I’ve come across in my life who could use a prompt engineering guide for in-person interactions.

This feels like the kind of thing strategy consultants will have an easy time saying and a much harder time actualizing. What are companies going to do when their competitors lay off 20%-30% of their workforce and cut prices to consumers?

The amount of time you spend at work managing edge cases and exceptions is inversely proportional to the speed with AI is coming for your job.

Advice from a Product person to all non-Product people: everyone asking for their feature to be prioritized thinks their feature will be simple to build. Added pro tip: it’s rarely simple.

I still find it mind boggling that we can map this out so precisely so far in advance.

Clever

"If you desire glory, you may envy Napoleon, but Napoleon envied Caesar, Caesar envied Alexander, and Alexander, I dare say, envied Hercules, who never existed." - Bertrand Russel - The Conquest of Happiness newsletter.pillarsofimpact.com/p/the-comparis…