An increasingly large amount of people judge models based on outputs instead of critically questioning their ability to prompt
This used to be clear at some point but inputs very much influence outputs
The whole Kolibri debate is so classic European. Cheer for smth that’s months behind and would barely be worth a look if it came from china or the US but wowow they shipped and llm
Reusability in Software is a bygone principle of times when time was scarce and engineering hours were constraint
No more no excuses
Would never send any of these agents to do payment related stuff for me bc they miss proper guardrails
So easy to prompt inject - billions will be lost bc you trusted some vibecoder with your most sensitive data
It’s so funny to see the left experience cancel culture for the first time
Like yeah hun look at the monster you birthed
The fact that Iranians managed to use Claude for anything illegal despite it being censored to death atp is an accomplishment in itself
If you want to find the real opportunities in consumer ai talk to people
Most people still think of Chatgpt as just a better google search
Listen and learn not preach and lecture
Am I the only one that’s completely turned off by the notion of an agent scheduling "meetings" with my friends in my calendar
Like the last thing I’m gonna ever give my agent is control over my time
Would be kinda funny to compete with meta and pay them $ for muse and $$ for ads
Reselling tokens for tokens we really live in a video game

