But where does taste come from?
38 points by dustyweb
38 points by dustyweb
We destroy institutions while wanting their positive effects. This quote captures it well:
“In a sort of ghastly simplicity we remove the organ and demand the function. We make men without chests and expect of them virtue and enterprise. We laugh at honour and are shocked to find traitors in our midst. We castrate and bid the geldings be fruitful.” - C.S. Lewis
Also, it is absolutely rich to expect practitioners to also be supreme arbiters of taste after decades of industry trying so hard to make them interchangeable.
it comes from putting something in your mouth and sometimes spitting it out. People who lack taste come in at least two varieties: either they have not put enough things into their mouth and so have not tasted the sublime or the putrid, or they have internalized, falsely, a belief that nothing can ever be spat out, and in doing so have become conditioned into accepting the mundane or the putrid as somehow inevitable, slowly learning to unhear their mouths.
I'd even say that using generative AI tools results in becoming nose-blind to their bad smells.
yes. The sublime being a statistically unlikely outlier, it is scarcely if ever the output of an LLM, and so LLMs are raising a generation that, having never experienced the sublime, believe it not to exist.
So you'd better have exceptional taste yourself, and here's the thing: ultimately, taste in terms of practice has to come from doing the practice yourself,
I don't think you have to be a great chef to have a great palette, you just have to put a great many things in your mouth and talk about them with other people. Being a chef gives the natural advantage that you have more agency over your exploration and can explore more flavors, but surely it is possible to be a gourmand without being a cook, although admittedly, likely easier if you are a cook.
So the best way to retain taste is probably to use these tools as sparingly as possible.
absolutely. Although I use LLMs at work, where I am encouraged to use LLMs, I refuse to use them outside of work.
Props, first of all, to the author for taking the feedback well.
I think this sentence is understated. Props to the author for taking the feedback well, for honestly reflecting on not only their output but also on the process for producing that output, and for posting that reflection publically in a vulnerable way.
Reflection is super important: understanding why you do something, or why you think a certain way, goes a long way towards being able to make a decision about whether you want to think that way, which, if the answer is no, in turn opens the way to improvement. Getting to the "improvement" point is hard and takes multiple steps, and honest (personal, private) reflection is a necessary first step.
I'm the opposite of spiritual and didn't experience much value from meditation when I briefly tried it. But reflection is good. Acting upon the conclusions is my next challenge. :P
Although I agree when it comes to api design, or sensible internal code structure... The obvious counter is...
Do architects have to build buildings and lay bricks by hand before they are skilled architects? I doubt it. They just have to practice designing things that are aesthetic.
The quality of AI code keeps hill climbing - I see no reason someone couldn't make beautiful things with them.
Do architects have to build buildings and lay bricks by hand before they are skilled architects? I doubt it. They just have to practice designing things that are aesthetic.
That's not really true - architects are, first and foremost, project managers and designers of functional spaces. From the BLS:
Architects develop final construction plans [...] These plans include drawings of the structural system; air-conditioning, heating, and ventilating systems; electrical systems; communications systems; and plumbing. [...A]rchitects must follow state and local building codes, zoning laws, fire regulations, and other ordinances, such as those requiring reasonable access for people with disabilities. As construction continues, architects may visit building sites to ensure that contractors follow the design, adhere to the schedule, use the specified materials, and meet work-quality standards. The job is not complete until all construction is finished, required tests are conducted, and construction costs are paid. Architects may also help clients get construction bids, select contractors, and negotiate construction contracts.
Architects who have an insufficient grasp of these aspects of the job are often criticized as working in "renderite"; that is, they design aesthetically pleasing buildings that are infeasible to build because of cost, physical limitations of building materials, or relevant codes and laws. Or at least that's what my uncle says about architects he doesn't like. (He's an engineer.)
I worry that a lot of people who spend most or all of their time vibecoding are becoming renderite software developers.
Do non renderite Architects learn by doing? My guess is they are in and around the doing, but not usually doing it all. This might be equivalent to iterating on code at a low level with AI, but not necessarily typing every line of code.
My feeling is that experienced engineers get more out of AI than vibe coders - but they don't need to place every brick (or line of code) by hand to make a good product.
This might be equivalent to iterating on code at a low level with AI, but not necessarily typing every line of code.
I think we don't have an exact equivalent to actually digging ditches and pouring concrete because software is essentially free to actually produce (though not to operate). I would consider "being in and around the doing" (a good phrase, thank you!) to be more like optimizing, debugging, and operating software - and, indeed, my experience is that engineers who are on-call for their or others' software tend to gain a lot of useful taste in error management and observability from that experience.