unlike a work computer, apparently AI subscriptions are not self-evidently worth having

Back in the 1980s it was highly debatable if there was value in desktop PCs beyond playing Solitare.

employees can just get money

There’s your problem, right there, money is dangerous, can’t let the rabble have money - they might get too much and then they’d have power, power to disrupt the people who have money…

And the RAM upgrade prices have been a consistent Apple profit center for over 20 years now.

Well, if the puddle jumper pilot is dumb enough to hop in the cockpit of the 747 with 500 souls on board and use AI to get tower clearance to taxi to the active and take-off, he - and everybody who might be paying him - deserve the jail terms they should be serving.

If you get an MRI that shows your child has a brain tumor, you don’t buy yourself a skull saw and scalpel to save a few bucks using AI to guide DIY removal surgery.

At some point, hopefully, the users will shape the social media platforms instead of the other way around.

what ways your business encourages safe use of the thing.

We have extensive regulatory quality safety procedural frameworks in place around pretty much everything we do, regardless of how we do it. So, we use AI as a tool to make us more efficient in the things we already know how to do, we don’t use AI to replace expertise we don’t have on hand, or let it “do our jobs for us.” Our job is to put the right things on the page, if AI does that for us, that makes us more efficient, but if AI does it wrong and we don’t catch it, that’s us not doing our jobs correctly, grounds for disciplinary action, firing, potential legal exposure, etc.

They are asking their employees to produce more

always

and understand less.

Usually not. At least from where I sit, we retain employees with experience / understanding, even at 2-3x the pay of alternate new hires who have all the degrees and certifications to wear the hat. AI is a tool to help leverage your existing understanding, not enable you to make “expert decisions” about things you know little or nothing about.

There are plenty of places that just hire the cheapest they can get, I took a job with one - briefly - 15 years back, before I found a better place. They didn’t impress me as actually efficient while I was there: headcount was high, labor cost was low for the headcount, but high for the actual productivity. They also didn’t survive COVID as a business.

How long after the chainsaw came out did they develop those safety courses?

How long after the ladder was invented and widely used before OSHA and similar started codifying “don’t be an idiot, people get dead that way.”?

Business just about always rushes into “the new thing” seeking “first mover advantage” which often involves a lot of coloring outside the regulatory lines before those lines get drawn - in some ways legitimately hoping to establish valuable use cases that shouldn’t be arbitrarily impeded by overly restrictive regulation - but in most ways just trying to grab cash for the next quarterly report.

Responsible businesses are telling their employees to “use it wisely” - Risky Businesses? They’re out to make money as fast as they can, and “clean up their act” before they get busted, but not before they maximize their cash grab.

In C++ land, I lived in Qt for 20 years. It did… most things, so if you “just” imported Qt (or Boost or massive API environment of your choice) you could usually do most things “just” importing one or two additional external libraries. I frequently would split a system into “micro-ish-services” with each service importing one or a few of these novel external libraries, partly to isolate them so unexpected interference at least wasn’t coming from within the process, also as damage control incase one behaved badly it could be excised at runtime without taking down the larger system.

Rust feels even more like a case for cooperating microservices, but it does seem to bulk them up fast - faster than Qt, and that’s saying something.

it’s unsafe to use if you don’t already completely understand the output

Yeah, and that has been true of “modern tools” since forever. Somebody picks up a chainsaw - looks easy…

Update: 4 hours of refinement later, I have 4 elevation drawings ready to submit… it would have taken 4 hours to select and engage an architect and meet with them to describe the project and collect their work - and that would have been spread over a week or more, instead of done in one evening.

I’m not sure if I’m reading your intent correctly or not, but the AI agents actually excel at “puddle jumper” tasks. Stupid stuff that you could write a one-off script for, but damn that’s a lot of hassle. This afternoon a colleague and I were putting together a powerpoint slide deck based on a folder full of disorganized garbage. Claude digested the garbage and wrote a python script that wrote the .pdf and I’d swear it took almost as long to open Office 365 Power Point as it did for Claude to write the script that wrote the 7 slide .ppt file.

They should…

In the 80s the hand written assembly was more reliable and performant than the C, at least on many of the compilers.

Even in 1990, I tried to launch a serious project in C++ on the IBM-PC, and the best available compiler was too buggy to use. It did fine for little demo apps, but by the time you wrote code for 2 weeks, you started hitting bugs - not in your code but in the compiler output… we had to fall back to C for the project. Even later, around 1994, we had two C compilers for 6811 work and one of them was garbage, I could hand write the assembly better and faster without even trying hard. The other one was pretty good, and by the late 1990s I stopped looking at C/C++ compilers’ assembly output because it was consistently better than I would write by hand.

One of my junior devs was having trouble with a bug in an internally developed tool, apparently for weeks before I saw her struggling with it over her shoulder - it was a 5 minute fix, I hope I made it clear to her: speak up when something’s wrong - this 5 minute fix has cost you many hours already because you never told me you were having a problem.

I’m in one of those others I mentioned (and I try not to reference my company online because of… reasons), and we’re getting strongly encouraged to “integrate AI in our daily workflows, where it makes sense” - not just coding, but coding is an obvious target. As a business we tend to change slowly, so this will be… interesting.

Safety critical (aerospace, medical, precious few other) industries have regulated quality, with moderate success. It’s far from perfect, farther from ideal, but it is providing some additional resource and schedule allocation to do the things that need doing to ensure the systems don’t screw up too badly, too often.

Insane, yet reliably employed in the field for 30+ years - first and current job for more than a decade.

Supposed to be ≠ is be.

Opus 4.6 is genuinely pretty decent at programming now if you give it a good backbone to build off of.

Soup from a Stone.

To an extent, yes. The more “broth base” I feed Claude, the better it does. If I just vaguely describe a program, I get a vague implementation of my description. If I have a big, feature rich example (or better, examples) of what I want the program to do, Claude can iterate until the program it make’s output actually matches the examples.