Spent my morning setting up Cursor automations and Claude routines.
The demos make agentic coding look effortless. The configuration absolutely is not.
But watching an agent successfully pick up and complete work on its own for the first time?
Extremely satisfying.
I've been using Matt Pocock's AI coding skills and the biggest improvement isn't better code it's feeling more involved while writing less of it.
The structured workflow keeps me on top of the decisions while the agent handles more implementation.
Process > bigger prompts.
There is a toddler mode the 15 minutes before bed time where they run around like they have 20 projects deadlines they have to complete before going to sleep and it’s hilarious.
Reports say 90% of AI projects are failing.
The reason? Most teams treat AI like traditional deterministic software. But AI is probabilistic.
Instead of coding for fixed inputs/outputs, you’re wrangling a system with infinite possible behaviors (chatbots).
Much harder.
Instead of debates, we should just have all the presidential candidates play One Night Ultimate Werewolf together to see which ones are the jerks no one would want to play with.
@Coutsey Yes, Vasilevskiy is better but Fleury especially in his last 4 years here was performing at what most would call "elite" status. He is a less sound, athletic-based goalie so we remember the soft goals he lets in.
I love the Pens but us Pittsburgh fans can be very entitled.