Ask HN: What's a Dirty Secret No One Wants to Admit in the Tech Industry?

2 points by karakoram ↗ HN
What’s a Dirty Secret No One Wants To Admit In The Tech/Software Industry?

Industry as a whole or within software engineering.

28 comments

[ 0.25 ms ] story [ 12.3 ms ] thread
Most of the work is meaningless, lots of folks are chasing total comp. Many metrics measured are orthogonal to success. Complexity for complexity's sake. "Just one more framework|technology|abstraction bro."

In Defense of Simple Architectures (2022) - https://news.ycombinator.com/item?id=39440179 - February 2024

It's all bash scripts
120 hours of leetcode prep to get a job centering divs with claude
120 hours of leetcode prep to get a job writing design documents
The business customers don't understand the concept of a dangling else.
The average programmer is an illusion. Most programmers working in software, and I’m including not just the tech industry specifically, but all companies that employ or contact for programming are not very good. The relatively few who are outstanding skew the average. You know the joke about Bill Gates walks into a bar?

Most programmers were already doing what vibe programmers do with AI now, just more manually.

1. True security cannot rely on code correctness but only on compartmentalization. Qubes OS does this right.

2. Your phone is an entire computer, https://news.ycombinator.com/item?id=47367568

3. Intel ME / AMD PSP is a computer inside your computer that obeys someone else, not you.

That most people could care less about tech and are just in it for the grift.
The same programmers that shame others for using AI tools like Codex, Claude are also secretly using it themselves.

Whether at work or personal projects or anything else.

Nobody wants to admit it, but they are using AI.

This is a huge cope
It is real.

In the end you and I and everyone else will be using AI for work and personal use.

No amount of shaming others using AI would stop it.

The bigger cope will be engineers losing their jobs by not adapting to the change that is happening in front of them.

The only shaming I’ve seen are for people using AI badly or not at all.

I also think we should differentiate between programming to get a program you can do something else with and programming to understand the machine and give the programmer a logical challenge to keep their skills sharp.

I will never use LLM's. Never. Your cope for your own faults is very sad.
You might not, but your children and grandchildren will use LLMs and AI.

These technologies aren’t just going to go away just because you choose not to use them, or if you change careers, retire or whatever.

That's a different point from your original comment.
Two supporting arguments can be true at the same time.
Virtually no one in tech seems interested in doing what Jobs suggested despite the level of success he delivered.

Telling an average developer to work backward from the customer will generally be met with a foul mood. Especially when that constraint suggests that the currently preferred technical ideology or tribe is no longer suitable.

The whole thing is one big ego trip most of the time.

A compounding factor is when requirements are filtered through B2B layers. At which point the end-user is not the customer and it turns into a game of telephone.

Apple sidesteps this issue with vertical integration.

Automation isn't having robots assemble things, it's laying off accountants who read numbers from spreadsheets, type them into their calculator, and enter the number back into the spreadsheet, and replacing them with a spreadsheet that was created by someone who actually knows what they're doing.
People who work in tech aren't as interested in the tech as they are in the money.
Most production code nobody actually understands in full. People just know their part, and the rest runs on inertia and fear of touching it.
AI is not making engineers more productive. It makes them lazy and think less critically. Also leads to soft-mindedness.
That AI and search engines use a lot of cheap labor. On systems like Mechanical Turk, or hiring third world quality raters to check results. It is particularly obvious.
The placebo effect of useless dashbords.
Computer security is a solved problem, and has been since the 1980s.

Unfortunately it requires us to abandon the ambient authority model of application development and our operating systems that support it. When you tell an application to open a file, it can actually do almost anything. There are operating systems that enforce your choices, not many, they're very niche, for now.

We're at the point where the power grid would be if nobody ever used fuses or circuit breakers. All power could rush to any fault. The same is true with any application a user might run under Linux, Windows, etc.

To save a lot of grief: I'm not suggesting we use Windows UAC, AppArmor, or any other administratively driven security measures. Think power boxes that replace file dialog boxes, and the give file capabilities to the application per the users will, in a secure, transparent and easy to understand manner.

Most of software should be solved by now. And we shouldn't be reinventing wheel each time we make anything. But that would be just too boring.
> Most of software should be solved by now.

what does solved even mean? If I don't own the wheel, I'm not reinventing it. Paying rent for Microsoft's or Google's wheels don't solve my software needs.