Hardest Part of Software | Prime Reacts 

😂😅

💯

You mean like every knowledge worker in the world today??

True. My mom gave up by the end of highschool, and would just start calling me "the executive" in spanish, because I was always on my giant table and PC.

my mom never questioned it (started to use my pc at age of 6 for at least 4 hours a day)... now she is proud xD

Heck yea! Don’t let ‘em be lazy smooth brains + you have a record of all the backflips they are asking for.

EVERY time I don't get a CYA mail I regret it. It might be a week later in which case I'm lucky because I keep handwritten notes and quote them but if it's six months or a year later I have to kick myself. They always say something like there's no need to write it down, no need to CYA, effing liars.

@@programmer1356 If they refuse to write it down I often play fool and don't do that task. And when they ask I speak something like this: "Oh crap! Man, I completely forgot to do that task! I was very busy that day. That's why I asked you politely to write that task down or send me an email. Sorry! It's my fault! Never happen again! It is what it is..." And they learn fast)

@@vitiok78 Nice. I'll try a slight mod on that next time. I'll mail them "I'm sorry, I can't find your mail where you told me what to do. Please forward it or reply with your requirements."

This is good advice.

Couldn't agree more

That comic's pretty good.

Spot on…

Exactly, if someone can make a way to have a human language text have zero ambiguity and be totally and verifiably exhaustive in its conditions we could just write a parser for it and it would be the code

Repeated ad nauseam and also completely wrong and misguided. All you're describing is that a collection of code and instructions to compile it produce a deterministic program output. That's all it means to specify something precisely. The point isn't to specify something precisely, the point is to produce something useful that solves a problem. Here's one version of the problem: will people with real-world problems to solve have better luck hiring programmers, spending hours in meeting explaining the problem they need solved, waiting at least a couple weeks to get results back, and then going through months and months of iterations with 2 week gaps between feedback, or will they have a better experience chatting directly with ChatGPT, getting immediately-runnable programs back, and then talking directly to ChatGPT with critiques and corrections and iterating directly with the chat bot? For a whole lot of problems, the current non-gimped version of ChatGPT is already at the point where you'd have a better experience engaging ChatGPT directly than 90% of software shops you could hire out. It's not commercially available, but that experience literally already exists in the research preview. No, that doesn't describe every programming shop, but it describes quite a lot of them and a huge percentage of programming work that is done today.

Normally the Nick Burns type of people piss me off. But in this case, you're hilarious

your second insight is horribly accurate and I have never considered that. But, god its so true.

Quite the comedic interpolation sometimes

Depending on the size of the country and how far its away from civilization, you might want to have a car just to get away in an emergency case.

@@nosleep7026 Cars are actually a terrible evacuation method. Traffic jams happen very quickly with just small surges, so imagine when you need to empty an entire city. The best way to empty a city is again, trains.

@@nosleep7026 Also, I never said cars are never useful and you shouldn't have one, just that trains are useful in small towns

What I do is write a PoC, the less code the better. Forget about abstraction, architecture, DRY, tests, pipelines, linters, compiler optimizations. Just make something that barely works some of the time and get the requirements from there. Once you have your requirements then you can rewrite the whole thing properly.

True. It is the job of an engineer to protect the naive and innocent from their stupidity.

"maintain lane position" you do know that the only way an AI can possibly know where the lane position is is the white lines that now no longer exist? That only leaves it with "slow down" " pay careful attention" and I sure do hope it never "proceed to drive" right off a cliff...

The problem then will be the hackable exploits in such systems that don't exist with traditional cars or horses.

@@andrewdunbar828 Of course! But we will deal with them like we always do with new threats. Even with those exploits and car AI bugs we will quite significantly reduce the number of deaths on the road. Isn't it the goal after all?

It is hard to take seriously an argument that hinges on the word "if", which by itself implies that the solution might never come.

@@Salantor Fair point.. I admit it’s an optimistic take, but to dismiss the “ifs” seems to greatly underestimate the rate of progress we are making in AI and related fields. Hence my choice of wording “does not age well” to imply that Prime is making a similar mistake. But feel free to not take my opinion seriously.

Making an argument about a hypothetical scenario by conjuring a hypothetical scenario. Also, humans can't program for all possible scenarios. If we can do 1:1 human learning with a model, then the end result is the same.

and would you trust the car company not to switch the priorities with some update, or because your bank had some issues last month so your subscription to the expensive a was payed late, and the systen hasent got the memo yet that its back the other way around

also , and i know i do this in the my own text , but i really think we should stop calling this thing , AI , as it seriously lacks everything i expect of something worthy of the I in AI,... intelligence , can instigate an action or conversation without requiring input, a dog can go to its doghouse and ly in it without having to have been told to do so. llm's are the worst in not deserving that I , if you have the entire works on logic and how it works fed in to you and have any level of intelligence you would be able to spot contradictions in the rest of the stuff that was fed into you and not simply copy them into your answer when asked to name a few

i dont see how a product manager could give the proper feedback, hell even if it was a programmer i dont see it. And by that point you might as well do it yourself

Assume multiple product managers telling the AI to set different priorities and inputting contradicting facts and requirements. Will probably lead to the first AI trying to k!ll itself.

I can totally see AI replacing big part of programmers in "feature factory" types of companies. This sort of work is depressing anyway. It might be more difficult for new coders to enter tha market but besides that, it would be for good.

Tons of Americans love trains. It's just that most of the USA's territory is so sparsely populated compared to Europe or East Asia. AND the country is really bad at doing large public works.

:)

2