Hey Tyler, thanks! I am testing it today and went through the flow. I love the deep thinking in the beginning of the process. Where it starts to break apart for me is when the coding starts to happen. Inevitably, Claude is making mistakes and the app isn't working the way it's supposed to. And then when I fix that and add a new piece of functionality, I don't see the DNA being updated. Maybe I need to be explicit?
So cool that you're trying this out! So the way this is *designed* to work is that when you start to iterate on different parts of your product, for example you say "add a swipable card row of selectable items", claude will dissect the prompt into card (schema), swipe (module), and select item (genome) -- and add each of those things to the respective DNA layers, then refabricate the code so that each new piece can be influenced by the rest of the context (layers).
As new pieces of your product are added, Claude *should* be ensuring that your DNA and product are evolving together, and if drift is detected Claude should be calling that out. Lets say in this scenario there is an existing toggle left/right module that exists on cards elsewhere in the app. Claude should detect that and ask you whether you really want to add an additional module (interaction) or use the existing one. If you choose to move forward with the new interaction, should ask for rationale and then consider adding that rationale to the appropriate layer (and refabricate the code).
I hope that helps clarify the intended design of the system, and maybe there are pieces of this explanation that you could add as context to help Claude achieve the desired result.
Nope, your project should continue to operate within the conditions/context of the skills one it's run. The only time I've had to 'remind' Claude to update the DNA layers first and to then build code based on that was when I switched models mid-project.
I wonder what is your perception when other roles get to this point, where we have a good set of these DNAs, as we wouldnhave a design system and its components/rules, and can create on the same foundation. Like the basketball player, imagine that now the players are "cyborgs" or like Neo, they download the skills.
Does that still remain a designer role? Trying to answer your first question in the article.
Regarding the topic of the article, your argument about product design needing a decisive shift rather than a gradual integration in the AI era is incredibly sharp and spot on. I'm reflecting on how this shift could redefine the 'craft' itself; perhaps moving designers toward a more profund role in shaping ethical AI outputs and ensuring true human-centeredness, which is something LLMs alone cannot fully achieve.
Hey Tyler, thanks! I am testing it today and went through the flow. I love the deep thinking in the beginning of the process. Where it starts to break apart for me is when the coding starts to happen. Inevitably, Claude is making mistakes and the app isn't working the way it's supposed to. And then when I fix that and add a new piece of functionality, I don't see the DNA being updated. Maybe I need to be explicit?
So cool that you're trying this out! So the way this is *designed* to work is that when you start to iterate on different parts of your product, for example you say "add a swipable card row of selectable items", claude will dissect the prompt into card (schema), swipe (module), and select item (genome) -- and add each of those things to the respective DNA layers, then refabricate the code so that each new piece can be influenced by the rest of the context (layers).
As new pieces of your product are added, Claude *should* be ensuring that your DNA and product are evolving together, and if drift is detected Claude should be calling that out. Lets say in this scenario there is an existing toggle left/right module that exists on cards elsewhere in the app. Claude should detect that and ask you whether you really want to add an additional module (interaction) or use the existing one. If you choose to move forward with the new interaction, should ask for rationale and then consider adding that rationale to the appropriate layer (and refabricate the code).
I hope that helps clarify the intended design of the system, and maybe there are pieces of this explanation that you could add as context to help Claude achieve the desired result.
Thank you! Yes, that helps. I’m debugging something more technical at the moment, but when I add the next feature, I’ll try this.
Should I re-invoke design DNA each time, like /design-dna [what I want to do]?
Nope, your project should continue to operate within the conditions/context of the skills one it's run. The only time I've had to 'remind' Claude to update the DNA layers first and to then build code based on that was when I switched models mid-project.
Tu, this is great.
I wonder what is your perception when other roles get to this point, where we have a good set of these DNAs, as we wouldnhave a design system and its components/rules, and can create on the same foundation. Like the basketball player, imagine that now the players are "cyborgs" or like Neo, they download the skills.
Does that still remain a designer role? Trying to answer your first question in the article.
Ty! Auto correct mistake, sorry.
Regarding the topic of the article, your argument about product design needing a decisive shift rather than a gradual integration in the AI era is incredibly sharp and spot on. I'm reflecting on how this shift could redefine the 'craft' itself; perhaps moving designers toward a more profund role in shaping ethical AI outputs and ensuring true human-centeredness, which is something LLMs alone cannot fully achieve.