Kris Puckett

Their words

you have to demonstrate that you understand these systems and you're like you're actually building something with them. Whether it is like another shader plugin system, just do something with it to see and explore and be able to to talk to it and like talk through the process.

Their words

do your products have an agentic focus to them? Or is everything readable by an agent because there's going to be a lot more agents doing the work for people. So you have to like think through that as a designer.

Their words

the context window, the window that the AI can hold and think gets filled up very quickly with an MCP.

Their words

MCPs are they're not going anywhere. Anyone who says an MCP is dead is like, that's hype. Like that's crazy. But, they're going to change quite a bit and we're going to see a lot more companies leaning more and towards better semantic APIs and CLIs than they will an MCP

Their words

it's easy for us right now, I think, to have like, "Oh my gosh, we can just do all the things." And it's like, "Yeah, but is it worth doing? And do you feel proud of the output itself?

Their words

chat surfaces are only going to be more and more prominent.

Their words

So AI native has to be beyond just like I made a really cool shader or I made this like dither effect that every other person is making.

Their words

all it is is it's a set of repeatable specific instructions that augments what Claude code can do.

Their words

The more context you give it about your work and what you're trying to accomplish, the better the output will be.

Their words

if you don't know where to start, you just start typing.

Their words

One thing that LLMs I don't think do really well at is like creative exploration, but you can teach it to do divergent thinking very quickly.

Their words

at their core LLMs are just pattern matching and so they are fantastic at very quickly identifying patterns that even though the human brain is an unbelievable like the greatest pattern matcher in history, we still have these gaps. And so you can use AI to become this mirror that's like reflecting back at you some things that you might be missing about yourself and your work and where you're at.

Their words

I think a lot of times we don't think we've got repeatable patterns. We do, we're just not great at identifying them.

Their words

There are other massive unlocks to me for AI around workflows, around like how whether it's content design, things that you wouldn't see, organization, like analysis. So these Claude code skills, you can I mean you can create anything that's repeatable pattern. You can create a Claude code skill for.

Their words

LLMs are going to hallucinate like crazy, going to make stuff up.

Their words

I don't know becomes kind of a superpower cuz I'm assuming I could work with Claude code to figure it out

Their words

I don't need to have fleshed out plans. I don't need to actually have like the perfect spec docs or PRD docs to get Cloud Code to start working