Undefined behaviour is still deterministic.

100% this! If I had mod points, you'd get them!

Most of my time at a large corporation was spent dealing with the bureaucracy to get something small done ... Something that should only take about 5-10 minutes of my time would be spent chasing down "the right person" to get me "access" or to "authorize" my request to do something that was only ever internal, even if it was a task from the CTO himself :|

I'll never forget: I made a small static web page that ingested some JSON from our internal DB in order to display that data in a way that was easier to "see" (charts/graphs for the non-numbers people), was never going to be seen or used outside of a handful of people who needed this data ... made the page in about 10 minutes on my lunch break (literally just some HTML/CSS/JS) to help out some of the managers.

The problem was that I needed to host that static page somewhere for the team to see ... should be easy to piggy-back on any one of the 1000's of web servers we have, right!?! ..... Took me 10 minutes to make the page but took a solid month trying to find a "cost center" that could "host" my page because no-one wanted to authorize it since it wasn't "for them" .... fml .... since I had access to one of our file store buckets, I ended up just sharing it out through that, so when you would go to company.site/files/static/page.html it would just load like a normal page

The #1 question I got asked: "what took so long" ............... smh

Given that you call programming "progging" (i.e. British slang for prowling), I'm not surprised that you think AI helping you out daily is "beyond any single humans knowability [sic]" .. you might be surprised what a real Software Engineer (or any educated/passionate/learned) person is capable of knowing and or learning.

I know many seasoned professionals who are experts in their domain and know quite a lot of details that an AI doesn't know (because it's either literal rocket science, quantum dynamics, proprietary information, or even just basic maths).

Depending on the API, it takes me about 10 minutes to fully understand what it does and doesn't do .. in its entirety. No AI involved, just simply reading some basics of the manual. After those 10 minutes, I can reliably implement said API without issue, any time, and generally faster than an "AI" could because now I have my "boilerplate" code that I can just reuse when needed (e.g. a properly extrapolated, abstracted, architected and tested interface).

If it takes you months to properly learn an API, "progging" might not be the right fit for you .. and there's no shame in that.

If you're just slapping together websites, then yeah, "AI is coming for your jerb" .. that's been the case since Web 2.0 and "web devs" are so numerous, "they took yer jerb" happened well before "AI" ever came on the scene.

If you're doing anything else, it -might- be a helpful tool, or it might not, but that's for you to decide, and if you're making those decisions, then you're likely not being replaced by "AI" any time soon ... bad economy, shitty management and loss of revenue (due to the aforementioned) might take your job, but "AI" is the least of your worries.

That aside .. what the post is trying to get at is simply the "human" side of things that is required for software engineering. Typing out code (what an LLM can do exceedingly quick) isn't the major time sync for a seasoned engineer. It's the "human" side of things that requires learning, adapting, being honest and creative .. those are things "AI" can't do.

And when an "AI" -can- reliably do that (without hallucination, predictably and without the need for an expert in the field to correct >50% of the output), then yes, a bipedal robot can absolutely be built using the same "AI" model that can then tango much better than you, and for a much longer period of time (even at your best, you still have to stop for a bathroom break).

That is the "AI" that I personally want because then I can direct it to properly clean my house, cook my food, take out my refuse, take care of my yard, and much much more, when I don't want to (or can't) do those things ... leaving me time to learn to tango for my own pleasure with the help of an AI robot who's better than you.

Or hell, even a "robot band" that I could tell to do certain things while I jam on my guitar, would be a major step forward. I enjoy having human input, but they're not always available. Having a few robo-instrumentalist at my disposal that I could riff off of would be a great way to work out some of my blue/jazz riffs that I've got .... as it stands, I cannot tell a robo-drummer to syncopate a little less while putting more emphasis on the kick and tell a robo-bassist to switch from the key of E to the key of G and have them not only "know" how to do that, but to just "go with my rythm" and change/adapt.

Will we ever get there??? Maybe. But it would require massive changes in how computers work (i.e. not deterministic silicon based and the languages to go with them) and would also require massive amounts of government help since none of that R&D would be profitable in any way for a long time (which is why the AI hype continues ... it's not profitable and no government sees actual valuable use cases in the current iterations that haven't already been done before).

Consider this: quantum computing still uses regular silicon to achieve some of its results. It's been in development for multiple decades by multiple governments with direct funding sources that are guaranteed (even if no profit is made). At best some of the languages we have for it are nothing more than extensions on currently languages (e.g. an "if" statement is still an "if" but with the ability to branch faster). It's had billions (if not trillions) of dollars sunk into it, and the best we can do with it in its current state is achieve results that a mechanical calculator could do 500 years ago (which is damn impressive given everything) .... we are still decades away from having the technology of quantum computers overcoming thermodynamics to avert quantum decoherence ... and you think "AI" will do better than a human can based on about 5-10 years of database inserts and matrix math??

In the mean time .. you enjoy your tango for what it is, and I'll enjoy my blues/jazz riffs for what they are: personal passion.

While your statements might be true, I would argue that those numbers are only looking at one part of the picture; strictly speaking, income brackets.

A middle class person moving to a higher income bracket in current times still can't afford what their 1970's counterpart could.

Using the years from that Pew Research article, an upper income household making $250,000 in 2022 would have been making almost $350,000 in 1971 (according to inflation). That's a significant difference.

Additional to that, the average house in 1971 cost about $24,000. $24,000 in 2022 dollars is about $174,000.

You would have been hard pressed to find a house for $174k anywhere in the US in 2022. The average house price in 2022 was $500k+. So even with those few middle class "moving" into the upper class, the cost of things has largely decimated any potential real earnings/savings.

Not trying to argue, naysay, or even say that what the Pew article would consider "upper class" today are actually mid-low after their buying power is taken into account; more just pointing out that no-one today, except the "real upper class" (i.e. the 1%) actually have the same real opportunities as their 1970 counterparts, all things being equal.

Yes .. it really is that different :|

Just as a comparison, at my local grocery store (not in a "food desert"), the cheapest 5lb bag of rice is about $3.50. 1lb of spaghetti is $1. A single head of broccoli is about $1.50 and a single bunch of bananas is also about $1.50 (which is about 5 bananas).

That's already $7.50 for about 1/2 a shopping bag :/

And that's for the basic store brand of those and not even counting other things you'd typically want with that (like sauce, spices, butter, meat, etc.)

... many Americans go across the border to get certain medical/dental procedures done because the passport/flight/hotel/procedure is so much cheaper than getting it done locally .. I think we're fast approaching that same idea for groceries :/

Hopefully not before I cash out my stocks and pay off my mortgage /s

(can /s stand for sardonic wit??)

Says you Li! I played my wedding photo backwards on my record player and my wife sung the entire Sgt Pepper album in the voice of Kermit the Frog while my kids actually did their chores and my dog finished my 42 side projects for me!!

Screw your 36PB of data!

Gotta hype that AI :| .. 280,000 hours is 134 years for 1 person working 40 hour weeks, non stop. If that were a team of 20, that'd be about 6 years ... Just to UNDERSTAND the code! Where the hell did they get that metric from? And why is that code so horrible that it would take 134 years for 1 person just to comprehend!?

Totally agree! And I also wish there were a case sensitive version as well ... but unfortunately that's a job of the filesystem itself :/ NTFS is 31 years old. It's old enough to have died in a war. It lived through the dotcom crash. It knows the origin of Pokemon. It saw Melissa and Stuxnet have its grandchild. It's older than formalized versions of C++ and C. It saw the birth of Java, C#, JavaScript and so many other languages. It's so old it actually knows what a book is. If it ain't broke, don't fix it ... spoken like a true [insert old person joke here] ye olde NTFS.

I remember splicing RGB cables into my coax so that my old CRT could play content off my PC that had a TV tuner wtih S-Vid out.

What's old is new again.

I'm not asking for "free" ... but when I go to WalMart or Target, I don't have to buy Tampons along with my potato chips (no matter the brand or flavor).

Ala-cart is lost on the for-profit entertainment system.

Having worked on hardware/electronics before as a the "coder", I can completely agree that it's hard to WFH when you have industrial tech or proprietary hardware that you can't take home.

To that, what I would do would be to build out simulators or emulators that would "react and respond" just like the hardware did, either by just displaying a simple prompt or even making some simple UI that showed what the hardware could/would be doing.

I did this for a few reasons; 1. it made testing my code before deploying to a million dollar machine much easier, 2. it allowed other dev's to get in the code and make fixes/updates as well especially if there was only one prototype, 3. it allowed the company to train people in a sim before going direct to the hardware, and one of the biggest reasons was that I could then WFH a bigger chunk of time because I didn't need access to the hardware to do the majority of my job.

I know that's not always an option for many due to various reasons, and totally not disagreeing with you, just saying that sometimes WFH is even possible when you have to deal with hardware.