Unpopular opinion but, taste is not the moat in AI. Designers keep treating it like an irreplaceable skill. But taste is pattern recognition (“intuition”) of great work.
The gap between “AI slop” and “great design” is closing every month.
This is for what’s worth.
6/ New LLM eval idea: The McMaster-Carr Test.
Ordering the right part is so specific and nuanced that it requires a level of detail most models still struggle with. If an agent can navigate those SKUs and find the exact bolt for the job, it’s legit.
McMaster-Carr: the website that software engineers geek about and hardware engineers worship. It’s a multi-billion dollar business that wins by doing the opposite of every modern e-commerce trend. No ads. No tracking. No "recommended for you."
Here is the business logic behind the ultimate industrial moat. 🧵
5/ Total Cost of AcquisitionMcMaster is rarely the cheapest, and they don't care. They focus on "total cost of acquisition" rather than unit price.
If a machine is down, the $20 shipping on a $5 part is irrelevant—the only metric that matters is that the part is there by 10:00 AM. Operational mastery > Marketing hype.
In the past two weeks, we've had Cursor events in...
New York, Singapore, Ethiopia, Tallinn, Cairo, Milan, Oslo, Warsaw, Bucharest, Ottawa, Seattle, and more 🤯
This community is incredible!
built this magic canvas at the @cursor_ai Singapore hackathon and got 1st place 🥇.
it’s a multiplayer canvas where users can draw and work on using hand gestures and voice commands.
Built with @elevenlabs@GoogleDeepMind@NanoBanana@fal and @supabase
congrats @3thumb for winning the cursor hackathon in singapore out of 500+ participants!!
his demo of drawing with hand gestures blew the entire crowd away!
@SherryYanJiang Thanks for putting this together. Reminded me of the hackathons I used to attend in the bay area 10 years ago. Glad to see the community building again.