theAurenVale

this maps to what im seeing in visual AI too. rendering quality is insane now, nobody disputes that. but most AI generated images look generic and flat because theres no direction behind them

same thing with writing imo. the output quality is technically fine but if you didnt wrestle with the ideas yourself the result reads like noone actually thought about it

writing forces you to confront where your thinking is vague. directing a photo shoot does the same thing actualy, the moment you have to commit to a specific angle or framing you discover what you realy want to say. skip that step and you get competent emptiness

this is the thing that keeps me up at night about AI tools across the board. the moment your tool starts optimizing for someone elses goals instead of yours the entire value propostion collapses. doesnt matter how good the output is if you cant trust the intent behind it. we already see this with AI image generators where certain styles get pushed becuase of partnerships or training data bias, you just dont notice it as easily as an ad in a PR

constraints are what make demo scene art so good imo. when you have 64k to work with every single pixel has to earn its place. compare that to AI image gen where you can produce alot of variations at zero cost and somehow everything ends up looking less intresting. theres something about working within tight limits that forces real creative decisions instead of just iterating until somthing looks ok

this is already hapening in the visual space too. go look at AI generated product photos or headshots from two years ago vs now, everything converges toward the same clean, competent, completely forgettable look. the dark forest isnt just text, its images, its video, its anywhere the cost of producing "good enough" drops to zero and nobody has to make an actual creative decision anymore. the irony is that real direction and real taste become more valuble when everything else is noise, but most people cant tell the difference until they see it side by side

you can already see this in AI generated images tbh. compare the outputs from early midjourney v5 to whats coming out of newer models trained on synthetic data, theres this wierd homogeneity creeping in. everything looks techincally clean but increasingly samey. the unique visual accidents and imperfections that made early outputs intresting are getting smoothed away

the real bottleneck isnt writing code its knowing what to write. every productivity article about AI tools measures output volume but nobody seems to be asking whether the things being built faster are actualy the right things. ive been building a side project and honestly the hardest part is still deciding what matters, the AI just helps me iterate on that decision faster

im bored of talking about AI in the abstract yeah. but im not bored of building with it becuase the gap between what the tools can do and what people actualy ship with them is still enormous. most conversations about AI are about capabilites. barely anyone talks about taste or direction, which is where the real bottleneck is imo

tbh the real productivity gain for me is iteration count. before id try maybe 2-3 approaches before shipping something. now i can test 10 diffrent implementations in the same window. code quality doesnt always go up but i understand the problem space alot better by the time im done