Declan Kramper (@dkbuildsco) header
Declan Kramper (@dkbuildsco) profile picture

figuring it out

Tinkering
Joined December 2021
so power to you. the one that has it all.
Maybe there are infinite parallel universes. The only one you can be is who you are right here.
4
3,231
self improving software or big brother
1
29
Makers, not takers
1
20
first make it work, then make it better
1
25
Declan Kramper retweeted
every product i use should just be a headless mcp, that feeds into an agent, where i never have to think about anything underneath. except for our product, of course. in our product, customers love the ui and love us very much. we're special that way.
149
407
107
7,054
324,958
yesterday I sat down at probably the last new coffee shop on a 4 year journey that has amounted to visiting 53 different Chicago coffee shops many having been visited many times some having closed or rebranded i like coffee shops. you can be anyone at a coffee shop. sometimes tho I spend too much time deciding which coffee shop to go to. So I made this little website as an excuse to try designing fast + slow generative experiences that would help me make that decision declankramper.com/coffee it uses the new OpenAI decisions api for the fast experience picking up coffee shops as you type based on keywords ---- and then a standard LLM call for the slow experience when you’re input has grown more complex, it’s time to press enter ---- some other decisions along the way that were fun - every time you hit enter, its a new message in the thread. but a conversation didn’t feel right, so instead claude haiku will classify each message into keywords and only display your latest after each message - on mobile the balls wanted to move, so they move with your phone - and shaking it is the magic 8 ball but for which coffee shop to go to ---- creds to Stefan for the OG balls design teapaw.org/heystefan_/status/2101…
when a designer gets access to Jev
1
2
120
when making a product decision, it can first feel comforting to hedge i see why agents like hedging so much but your job is to decide else its confusion/fatigue for user and mediocre product gotta get to the point where keeping hedges are actually uncomfortable because
1
18
if the screen starting to scare you, it’s definitely been scaring them
1
25
so good so fast posthog thinks its rage-clicks
1
21
Queens gambit, Johnny manziel, flu game
2
42
been thinking about this. i like the "pre-built" part. because if the argument of vertical (agent) companies against labs/incumbents rides on the vertical/embedded company knowing the work better, then you shouldn't need to necessarily generate everything from scratch because the work/customer/problem is understood even in future when generation probably happens indistinguishably from static, there's still a lag between discovery of new problem and understanding of it so then maybe that lag is handled by generation, then once closed, the understanding is generalized back into static for the others
this is the shapeshiftui pattern at enterprise scale. one input → Jev classifies the intent → picks the right pre-built interactive UI (form, calendar, approval card, etc.) instead of a static screen or full generation. shipped the same idea a few weeks ago: shapeshiftui.vercel.app nice to see it shipping inside Workday. credit would be cool though :))
1
19
visuals visuals visuals
1
16
Been thinking some thoughts regarding AI adoption and what those products would look like/do Relatedly have been seeing more companies lately (@tryramp's Glass, @YakLabsAI's Bonsai, @GetEnergy_) start to tackle similar problem of general AI adoption in businesses today via new product surfaces
1
29
The job of a footnote is to bring in tangentially relevant, yet optional, context that adds an extra level of depth / credibility / parallel thought to the current line of thinking should the reader please to inquire. Going back and forth between the location of a footnote, and the current line of thinking, can break the connection to said line of thinking. So I added a fun way to see that extra depth to my writing where you already are, should you want to, with a touch of the human-ness that comes from having said parallel thought.
1
31
I used to villainize attention. Think it was selfish. Narcissistic. Evil traits. Still can be. But I now think that’s largely cope. Got a lot of it when I was younger and think it led to some maladaptive behaviors and thoughts. Like the above. Yes there’s genuinely unwanted attention, and ill-intentioned attention, but there’s a lot of good that comes from it too. We are social creatures anyways. It’s okay to want it. (some even may say it’s all you need)
2
22
DJing is like an ideal form of product management Feedback is immediate and unfiltered The obsession is focused solely on the customer experience How to make them feel a certain way It requires judgement on what’s good, given factors like domain (venue), timing, and customer empathy Not taking every request as-stated, and having an opinionated view Best moments are often when they get what they don’t know they wanted Forced to make fast decisions And downstream of all that, if the product is done well, is business impact (tips)
1
22
no Elon hate accepted
1
14
my job is to ask good questions and decide what good means
1
20
yeah paper is nice. chat made me this video to show how i use my old iphone + a skill to make it easier to provide that paper as input.
Lately I’ve been learning math with ChatGPT (Astra) as my tutor, using this document scanning camera and a vibecoded tool that pastes the captured image right into the chat. This way, I just work problems on paper and ChatGPT gives feedback. It’s really nice.
1
27
hi, yes, one placebo please
1
18