0xdesigner pfp
0xdesigner
@0xdesigner
i almost flunked out in my first year of college. i thought, maybe i was partying too much? so i dialed it in and spent every night in the campus library. it didn't matter how hard i studied, or how much amphetamines i pumped into my body, i couldn't improve my grades. i chalked it up to being dumb. i just wasn't smart enough. the summer after my freshman year, i discovered a research group that conducted aptitude tests. it's basically a brain test for measuring natural talents instead of knowledge. over 2 full days, they gave me a bunch of weird puzzles to solve. the results showed i was in the bottom 10% of clerical speed--the ability to process written information. they told me to avoid becoming a lawyer at all costs. but the results also showed i was off the charts in spatial thinking. i should be an architect, they said. the next academic year, i started drawing pictures during lectures instead of the usual bulleted notes within the notebook lines. i was retaining information visually rather than verbally or linguistically. and for the first time in my life, i became a 4.0 student. it was a wild transformation. i didn't get anything less than an A- my junior and senior years. so why am i telling you this story? this isn't a recommendation for aptitudes testing (although i do highly recommend). my biggest challenge with vibe coding (bet you didn't see that coming) is living within walls of text. folder structures are 2-dimensional worlds. debugging threads are too linear. it's confining. i can't think clearly or see relationships between components. it's very taxing to read agent responses for too long, and things always go sideways when i quickly fatigue and blindly accept changes. using figma mockups doesn't address backend architecture. mermaid diagrams are a little more helpful, but it's not cutting it for some reason. everyone keeps saying designers are going to be the winners of the codegen era, and yet they're being squeezed into engineering ways of thinking. i don't think the solution is a departure from the agentic chat. chat is a great interaction pattern. but i do think the agent's output needs to be more visual or spatial than files with code and a folder structure. then people with brains like mine can really feel in control.
23 replies
9 recasts
90 reactions

Sangohan.base.eth 🟦 pfp
Sangohan.base.eth 🟦
@sangohan
Thanks for your super interesting feedback. I was actually just talking about this yesterday with some friends who were telling me I was intelligent. I asked them: what is intelligence? They still have the same biases, and think there’s only one kind of intelligence. Whereas, if you dig a little, you quickly discover the different types of intelligence that exist. What LLM are you using? So, Claude seems to already be set up by default to be open to visual explanations. Sometimes I ask it for .md docs, and when I open them I’m surprised to see it’s built well-formatted graphics, diagrams, and more. Maybe by fine-tuning your settings, you could get more output leaning in that direction?
0 reply
0 recast
1 reaction