Neddy Seagoon posted:

Meanwhile; Inside Twitter Headquarters...

Ignis posted:

motherfucker seriously believes designers have to code

https://twitter.com/elonmusk/status/1633016157568651264

Telsa Cola posted:

Has a large social media site ever been actually sued successfully for selective enforcement?

zedprime posted:

No one has been in a hurry to try because it's a process with no winners if you try enforcing anything but the most robotic take on it.

It's a level of legal involvement you want your chief counsel making sure you have the legal bandwidth in case it goes to court and ideally what your chances are in court and finance or operations able to account for the penalty if it goes south.

The very most you can do if you want volume of exceptions is take these chief level decisions and give them to a minimum wage or off shored moderator slamming buttons in the mod console trying to wrack up activity bonuses.

gschmidl posted:

I really hope someone snaps and beats this poo poo rear end motherfucker into a fine paste.

Breetai posted:

What makes me waver between laughing out loud and cursing profusely is that stupid loving code review bullshit that he insisted upon. "Show me a sample of code that you wrote that is spectacular and that will wow me/show your worth to the organisation" is so unbelievably wrongheaded that it's hard to describe exactly how terrible it is, but I'm going to try. Apologies for a rambling infodump but I'm going to try to explain it in a way that non-full-time code monkeys can understand it

So I'm a programmer on a Maintenance team for the scheduled reports that my organisation creates. Our job is to keep the programs running when there are changes to databases or our physical infrastructure, and to fix any identified problems with the code.

One such task I was given was when a program kept on outputting different results with the same input. You could run it twice on the same data, even on the same day, and the results were grossly inconsistent. None of the values that were actually critical for the report for were inconsistent, but every other value was - this was taken as a sign that the report's accuracy was suspect.

I had to look at the program, which I had never seen before, figure out where the inconsistencies were coming from, and fix it. This involved basically unpicking thousands of lines of code that someone else wrote to work out what did what, and running it multiple times in stages to try to work out what was happening. Eventually I found the cause.

Basically, the data table that was feeding into the report contained transaction data for a large number of customers. This included an ID# for the customer on our system, 3 columns with the data values that we cared about and which were critical to the report, and a bunch of other columns that we didn't really care about the contents of. Now several customers had multiple entries in this data, but we didn't want them to appear several times in the output report. As the ID# for the customer and the values we cared about were the same across every entry for each individual customer, the guy who had written the program included a step that basically told the program to only retain rows that contained the first time a customer appeared in this data table.

The issue was that the program was written in a way that when it created that data table it sorted the data by the customer ID# and the 3 data values that we cared about, and that was it. As our system runs on multiple processors, it divides up tasks according to what the load on the system is, and this means that different chunks of information are potentially handled in different order each time a program is run. When the data was being sorted it sorted by the 4 explicitly defined sorting variables just fine, but any rows that had the same values for those sorting variables but different values for other variables could potentially be returned consecutively but in any order. So when the program subsequently took the step to only retain rows that contained the first time a customer appeared in this data table, the values for the 5th variable onward were essentially selected nondeterministically. Hence the inconsistencies. Basically the way that the results were returned for each customer looked something like this, with the first 4 columns being the variables that they were sorted on:



The above is dummy data that isn't anything like the info we were looking at, but the point is this is how they were differently ordered despite being sorted by the first 4 variables. In the first run it would retain the data from Service #1, but in the second run it would retain the data from Service #4.

It took me ~8 hours to investigate, test, check and double check a solution, and submit my changes for review. A lot of this time was spend waiting for the program to finish running so I did other work concurrently, but essentially the whole thing took a full day of work to solve. The solution that I found which adhered to the principal of making the minimum change necessary to result in the desired outcome was to add an additional variable to the sorting step - this variable was a transaction number unique to individual transactions, which meant that the data would be sorted in the same way each time, which meant that taking the first row that a customer appeared in would always return the same data given that the input data was the same. The actual code that I wrote for this full day's work was exactly this:

code:

 , trnsnmbr */CHG#10 additional sorting variable added to resolve nondeterministic selection of data /*

That's it. The comment explaining what the change did is like 8 times longer than the actual change itself. But the report was fixed, would return consistent results when run on the same data every time which means that it met our quality standards, and could be used again in the knowledge that the info in it was sound.


Now imagine being asked to provide a code sample for a half-witted egomaniac of a boss to justify what you've been doing with yourself, and all they see, devoid of all context and understanding of what you've actually been working on, is " , trnsnmbr".


Granted it's a bit of an extreme example, but I think it illustrates the difference between required effort, impact, and the actual volume of code produced. And consider that this kind of thing - small changes that solve issues with minimal impact - is generally the purview of teams responsible for maintaining and upkeeping existing code. Particularly consider it in the context of Twitter's current performance. I'm not saying that he actually personally reviewed the code that he was provided - that was clearly just fodder for his fanbase who thinks that Elon is a real-life Iron Man who is capable of instantly assessing a coder's worth from a snippet of contextless code sample. But if his selection of who to cut was in reality based in any way on metrics that included volume of code produced, then maintenance teams would tend to be disproportionately represented, wouldn't they?

Roblo posted:

🤣🤣

bobjr posted:

https://twitter.com/igb/status/1632942860495626241?s=20

Breetai posted:

" , trnsnmbr"

Breetai posted:

That's it. The comment explaining what the change did is like 8 times longer than the actual change itself. But the report was fixed, would return consistent results when run on the same data every time which means that it met our quality standards, and could be used again in the knowledge that the info in it was sound.

Ignis posted:

motherfucker seriously believes designers have to code

https://twitter.com/elonmusk/status/1633016157568651264

Breetai posted:

Paracausal posted:

https://twitter.com/dom_lucre/status/1631764763385405443?t=t0VBcwMMiSJ950tnJUvJFg&s=19

Breetai posted:

Paracausal posted:

This thread is proof enough that anyone who paid for Twitter blue should be on some kind of registry, blue tick pushes an insane theory, other blue ticks drooling at the mouth to agree, algo prioritises it. Congratulations Elno you have the truest genious platform
https://twitter.com/dom_lucre/status/1631764763385405443?t=t0VBcwMMiSJ950tnJUvJFg&s=19

thin blue whine posted:

The guy didn't even insult Elon or slight him at all, and Elon just openly, in front of God, the Meme Community, and everyone viciously poo poo talked this dude for daring to speak to him.

Funky See Funky Do posted:

How can centrism have a cornerstone?

quote:

He thinks for Mr Musk it's about money. He says cleaning and catering staff were all sacked - and that Mr Musk even tried to sell the office plants to employees.

Animal-Mother posted:

this is the most unprofessional thing ive ever seen and ive seen heavy equipment operated by drunks

Tai posted:

What am I reading....christ

Scratch Monkey posted:

😂😂

delightful posted:

Elon loving sucks but holy loving poo poo his sycophants are soooooo much worse. Jesus Christ, humanity is doomed. Please let them all end up on Mars and get them the gently caress off this planet.

Kale posted:

Now Elon's beef seems to be that the Jan 6 commission was bad and faked and spreading disinformation and propaganda regarding it's activities to help repair the image of fascists as best he can. Kind of makes you wonder just how much money he's going to be shoveling at awful GOP candidates once everything has completely shaken up by 2024. It'll probably be taking up 100% of his time and twitter will have transitioned into a full blown right wing propaganda social media site the likes of which the world has never seen, even in Truth Social and Rumble. Like the things is there's tons of options already if you're into this poo poo.