I have been thinking about trying this approach. We really need something that's in-between, right? There's going to be code that is boilerplate implementation, and I don't need to drive that - like we have, in the before times, relied on macros and what not to do that work. But then the interesting logic, seeing the suggested implementation, copying or working from that. That's the sweet spot.
I've also been thinking about the partner programming craze phase our industry went through. I write the test, you write the code; well I write the test, LLM satisfies with code. (Then we have less of these weird LLM generated test cases that test _nothing_. We can also use the LLM to suggest tests to complete coverage.)
These would be slower to work this way, but the end result is:
1. humans still learn
2. you have a good grounding in how everything has been written
Perhaps I'm just daft or slow, but I'm not 10x-ed by LLM usage. It's precisely because the details matter. (And yes, I read the article and it's lamenting the loss of low level machine-level details.) I read the LLM output, it slows me down. I challenge the LLM output assumptions, that slows me down. I stop to step away and think, it slows me down. I cannot run a bunch of simultaneous prompts and supply this same level of mental context to all of them, because my brain works best when I treat it as single-threaded. My experience with this approach to using an LLM has been, so far, positive.
What I have noticed as I've migrated to newer models is that the assumptions made are incorrect less often. But if I don't pay close attention wrong assumptions can still be made. Cross cutting concerns are lost, or written in such a way as to be useless.
So, the loss of details, to me, would be my own laziness in not performing due diligence.
The example could be hyperbole, but Home Depot and Lowes both offer a wide selection of "smart toilets" and if we think "oh this is just a weird niche thing that will never catch on" let's remember how difficult it is to not purchase a Smart TV when television shopping these days.
Most appliances have "smart" features now, it's getting hard to avoid. But I've never enabled them. My dishwasher and my clothes washer have both bluetooth and wifi features but I never set them up. I just use the controls on the device. I have no interest in getting information about my appliances when I am not physcially in front of them.
Yes, but the controls on the device cost the manufacturer a couple of cents that they will be happy to not pay as soon as they can get away with it. First it'll be (in some cases already is) the advanced or rarely used functionality. Eventually you'll be expected to connect the damned thing before it can be used at all.
It is kinda funny that “smart” is the branding we’ve converged on for these devices that have unnecessary computer features. Most engineers would, I guess, call it pretty stupid to add a computer to the BoM of your toilet.
Auto-flush, seat warmer, scheduled seat warmer, cleaning cycle, scent dispenser, "you haven't been going as often" sensing, auto high/low flush depending on time on toilet...
Just off the top of my head, and I'm no toilet engineer. I'm sure there are lots of possibilities with just a sensor or two.
Gastroenterologists generally categorize feces into ~7 categories. They can be indicators of a variety of dietary deficencies, including dehydration or insufficient fiber intake. Others can be indicators of a variety of medical issues, up to an including colon cancer. Ubiquitous smart toilets scanning and categorizing our feces would move the needle of overall public health and well-being.
The whole point of the label "smart" is to fill in vague positive thoughts for the customary "..." that comes right before "profit" in the standard /. business plan. And the real problem is that it works.
Depending on where I park in the city, I have to install a different parking app. And create an account, and pay with a credit card via the app. There was a time, however brief, when some of the parking meters just accepted credit cards, or get this, coins.
This is what happens when a large number of people claim they "don't seem to feel the pain" and comply with the data broker's wishes. We all get stuck in the data brokers web, or we have to break the law to park on a public street.
100% this, and it is annoying. I pay for street parking (i'm ok with it, it's a tax base for the city I live in), but i'm not Ok with having to pay .50 per transaction. And i used to carry a stack of quarters in my car. And then you add all the vendors with their proprietary apps in this space, the tracking of the data of what car parked where, when, etc.
I'm suprised they were allowed to remove the coins from the meters, but they were. So people who don't have a premiere phone, because the apps don't work on your old crappy phone, or people who don't have a credit card to feed into the app on the phone, are SOL.
I've also been thinking about the partner programming craze phase our industry went through. I write the test, you write the code; well I write the test, LLM satisfies with code. (Then we have less of these weird LLM generated test cases that test _nothing_. We can also use the LLM to suggest tests to complete coverage.)
These would be slower to work this way, but the end result is:
1. humans still learn 2. you have a good grounding in how everything has been written
reply