Add some melody, and you have a LinkinPark song.

@devux-bookmark all parties move out of India. You wouldn't believe how rich are clients here on the outside.

@devux-bookmark all parties move out of India. You wouldn't believe how rich are clients here on the outside.

@darksideofyay 2020 was 5 years ago, and software ages like milk in the sun - especially consumer-facing software.
Mostly because everybody is always buying new phones, and the networks keep changing, and new regulations take place, and then you get budget cuts to backend services, or 3rd-party providers go bust and have to be replaced...
Perfectly good software is actively maintained. If another company bought your HC provider, they most likely gave the boot to all devs of the 2020 app.

There is also "enshittyfication", it's like "shrinkflation" for food products (when companies reduce portions instead if raising prices). Software companies cut backend services or maintenance cycles or fire senior devs and hire greenhorns. Thus you get shittier software for the same (or not-as-high) price.

Finally, there is bait-and-switch (A.K.A. dumping). Offer great and cheap service, corner the market, then either raise prices of fuck up service - no client is able to leave, anyway.

Have you noticed that you didn't describe any of those as "poor" or "cash-strapped"?

Insurance is in the business of taking your money and giving in return the minimum possible service.
An effective way of doing it would be deliberately filling their app with dark patterns and bugs, and then training their call center operators to keep pointing insured clients to the app.
I wonder if one checks the source code for those apps they might find bunches of compatibility layers commented out by the PO.

Use a VPN, then you can be somewhere where it's 6 AM

@kiki crap, it was in the fifties? I thought they would have standardized by the bloody WWII

Imagine that an US unit is searching some Baghdad basement in the height of the second Saddam war. They have a finding, and report 50.000 9mm rounds as freaking old-timey 9 inch naval ordinance - just because they see the 9 and the 9 inch ordinance is the first in the reporting system results.
They have about a second to report it before they have to move, it's understandable that the first result in the system might get clicked.
Then...
The coalition force might have to reallocate an entire bloody carrier battle group. A fucking billion dollars evaporate just because someone reported the wrong size of ammo.
No fucking way the US armed forces will use an entire fucking measurement system different from their coalition forces.

@Demolishun chemical engineers are effective, cheap, and easily available. They are the good-at-math kids whose mothers reeealy wanted them to be doctors and didn't let them play with computers.
That being said, ok, they are a bit... analog when evaluating exact solutions. They need a more... through review.

"In X years, LLMs will replace almost all engineers,"
Not exactly. We will basically lock management into the matrix while we work.
Imagine that LLMs in the future also have live deep fake video capabilities. They already have batch video and audio, we're not that far from it.
We need only setup some langchain guidelines to have the LLM answer stupid status updates based on observability metrics and jira tickets, have it record the meeting and add markers to the few seconds of actually relevant discussions, and say stuff like, "Hey, <marketing jerk #8>, how are <family/sports/dogs/weather/games> today?".

Done. AI can do useless meetings like a champ. And the muggles and the suits won't even notice, even if they start to book the same "engineer" in six or more simultaneous meetings. They really know nothing of tech things like "doing actual work".

The only tech we still need are holograms so that we can give the suits the open offices full of "meat" that they love so much.

@kiki may sudo have mercy on your UID

There is a whole subject named "approximation algorithms" that deals with this exact question.
Because those big fancy NP-Hard problems can not be efficiently solved by divide-and-conquer... optimally.
That is to say, you can get a "pretty close result" using some method that is not guaranteed to find the best of the best results, often by breaking the problem into subproblems.
How "close" is that result? Approximation algorithms theory answers that question as: "within log(N)/4 of the best possible result for N" or something like this.

Ex: find the *longest* path in a graph, without any cycles in the path (in other words, find the diameter of a given geaph)

So, yes, in practice, we often can break any problem into subproblems. But if we ever find yourself into a discussion with a complexity theory nerd... be careful, always put the "in practice" caveat in your comments.

Keep saying things like this and a suit will say you have management potential.

Have you ever heard of game theory? Nash equilibrium? The worst possible scenario for a side is that it destroys it's weapons cache, but some other side doesn't.
A side can only control its own actions, so they will not only *not* destroy their weapons, they might also ramp up production.
So the doomsday mutually assured destruction is an equilibrium we all get fucked up on.
Nash won a Nobel prize and went insane thinking about this.

I've managed to get a beach head with some people by pointing out how the referee is completely wrong by calling off their team's goal due to a forward pass, or some other sports thing. It works better when the commentator agrees with the referee, because then the voices in the black mirror box are all wrong.
Still I am not able to have them make the jump from "TV might, sometimes, be wrong about sports" to "TV might, sometimes, be wrong about the political interpretations of events or even about the actual events"
But I think I'm getting there. The actual events have been helping, absurdity is the norm nowadays and it helps breaking the denial shell.

Until they delay two-week tasks for freaking 2 months while promising every day that "it will be done by next week"

Ok, it could have been someone from anywhere, but my most pungent memory of this situation was regarding a brazillian team member.
He was the one who gave me the tip to switch from coffee to mate*, though. So I think we're even.

*that longer & harder to type brazillian version of mate

@BordedDev @antigermgerm I didn't know we had a queen (or princess). I thought we were an autonomous collective...

You cheat on both every time you eat a bad clam or drink suspicious water.

Old, about-to-end processes almost never get updated. Usually, it is the hamfisted brickheads that stick to those stupid illogical wastes that get fired rather than having anything changed.
Then, some poor bloke has to maintain the unsolvable mess and does it for a heartbeat or two until leaving the company.
Then everything stops.
Then, an overpaid tool builds up a whole new mess.
It's the death cycle of letting business majors create operational processes.

@jestdotty that worsens the likelihood of offspring from the person that was lied to, not the liar.

Darwin has some explanations.. mostly about how those who lie better have a higher likelihood of having offspring.
Then, the odds are that this offspring will learn those behaviors from their parents.

The moth that doesn't like to lie pretending to be a leaf gets eaten by a bird. The bird that does not like to lie dancing to pretend to be healthier than it actually is, doesn't get to mate.
It's the Circle of Lies!

@retoor about your question, if they should know a lot about the data... I have only the most cowardly yet the most realistic response: "it depends on the problem, the data itself and the technique used in analysis".
In this case, the first two solve to "yes, they should know as much as possible about the data".

That being said, maybe your "employee development" argument may be combined wit @jestdotty 's trial-by-fire approach... if the presentation of potentially disgusting results is made in such a way as not to make the juniors feel too much embarrassed. But maybe it would indeed be good for them to get some calluses in order to thicken the skin.
I just need something to avoid them losing too much time on a doomed approach...
Maybe I should indeed ask for a midsprint 1:1 demo and suck up the boredom. But ask in private, and only for the people I can not trust blindly... yet.

Did someone distilled just the parts of chatgpt that were trained on 4chan and nothing else?
I thought it would be better at using punctuation...

@tosensei 's response is the case for biological beings.
But you can see the same thing on machines. If a car moves continuously stopping only to refuel, it won't last long.
Even formula one cars, that are always the best built machines with the best parts, best fuel, best calibration and always on the best road conditions, require maintenance on the middle of a race.
The engines would just melt down, otherwise.

@TheBeardedOne I've seen and used the word "muggles", since what we do is basically "magic" for them.

Also, @jestdotty hypothesis needs to be published. I have some supporting data...

A paradox at the heart of capitalism is the fact that every capital owner wants to spend as little as possible, but somehow expects everybody else to have money to pay as much as possible for their products.

In other words, how can you sell anything if nobody can buy?

All those "robots replace all humans" scenarios fall into that paradox. The owners of the robots themselves would become luddites the moment those machines steal their consumers.
Just look at how the real state owning class / the rent seekers are reacting to how remote work destroyed their business.

Thus, we can have "robots replace a whole group of humans," but not enough to hinder the consumption of goods or the value of the capital itself.

Just imagine how the meat, restaurant, wellness and health industries would react if everybody had a home cooking appliance capable of, fully autonomously, turn cheap vegetables into healthy but actually delicious meals.

So, don't expect robots to actually change things.

Regardless of the technology that powers it, the quality of its products, or even the truth about its processes and suppliers, DeepSeek disproved the hypothesis that "every tech company that sells AI buys directly from Nvidia".
That is a careful statement - DeepSeek might even be using smuggled Nvidia GPUs. But two things are certain: 1) they are not out there kissing Nvidia's ring (at least not in public). And 2), besides that, they still have paying clients.
That means that Nvidia is not the only way to AI salvation (according to market analysts, apparently).
Thus, market analysts correct Nvidia's price to reflect the end of its "omnipresence".

Forgot about the borderline-unreadable documentations (fuckers can't have a cohesive naming system for shit) and their freaking obsession with protocol buffers and auto-generated docs & examples.
Google cloud exists because procurement depts will rather stay with the same providers that come up with gmail for their cloud service offerings.
Instead of, you know, getting the best provider for each service purchase.

So, yeah. I use it. Company mandated.
Abandon all hope.

@antigermanist basically that, but exagerated to unreasonable levels. Also, they crazy