Vibe coding: imagine that you're "vibe cooking". You toss some spices in a "smart pan" and its supposed to add staples like flour and rice and water automatically, then cook the results.
Things get so easy when you're not the one supposed to eat the grub/sludge/muck that comes out.

My wife went to do some work for a charity. They "got her the best computer available". It was a poor mangled MacBook whose better days were somewhere in the dawn of the last decade.
She tries to type anything... and only gibberish appears on screen.

She comes to me, absolutely me puzzled. I try to type anything... gibberish. I boot up in safe mode, everything is OK.

I look around for system configs... there is a custom keyboard mapping enabled by default.
We check the weird stains on the keyboard... they are regular and in all keys. Like if there used to be adhesive stickers on the keys, and those stickers were later removed.

I boot up again... and type "q". It becomes "a". I type "w". It becomes "b". I dread typing "e". Sure as bug, it becomes "c".
By the love of byte, someone asked for a custom keyboard layout... IN FUCKING ALPHABETICAL ORDER.

It was easy enough to change the layout after that, but the weirdness continues: my wife asked around, and apparently the laptop used to belong to some old dude... who was convinced there were characters missing from his keyboard. Apparently he could never find them in a regular QWERTY layout.

I wish I could give some encouraging words for the kid who came up with the solution. Working around technophobes is a drying art, that needs to be rewarded.

We "engineered a prompt" (in other words, crapped a very poorly written paragraph using only short sentences) that gathers many arguments pro and against the usage of AI to code our production data pipelines.
Most of the "pro" arguments are variations of "we'll get fired if we don't use this crap in something".

Prompted with the whole argument, even Gemini thinks we should not use AI to write our code. Nowadays, we use Gemini mostly to justify why we shouldn't use Gemini.

Seriously. We made the whole setup so that any PO who demands AI has to convince the AI it is a good idea. I'll admit, the AI is doing a good job to deter people who want to use it. It hasn't cracked even once. Not that it matters, anyway. POs just ignore what the AI says and organically write orders telling us to use it anyway.

Fuck, my company is doing an all-hands meeting / Division townhall... AND EVERY DEPARTMENT HEAD IS DOING SOME STUPID D-DAY REFERENCE.
Cringe.

My twelve years old daughter saw me using an eletric mixer to puree potatoes, instead of the potato smasher.
She screamed: "You're using AI! That's cheating!"

When did "AI" replaced "Automation"?!?!
It's just like "woke", the word is meaningless now.

Once upon a time, there was a young, clean-shaven dev.
Soon in his career the smooth faced dev realized he had a knack for cloud setup. He understood permissions and network configurations, he could tell apart security groups from access groups, he could get all the information needed for an IPSec VPN in a single breath.
But if things got really tough, if the network layout was really messy, if the security settings too arcane, some older, bearded senior dev would come to solve everything.
Time passed, the junior dev got better at stuff, switched companies many times, solved a million problems, delivered countless projects, attended infinite meetings - twice - and, as often is the case, the senior devs, one by one, left the companies and run ever more scarce each day.

One day, a long, loooong time after his first meeting, the once young dev was solving network problems. Weird ones. Tough security requirements. Poorly documented cloud services. For hours, the dev and his colleagues toiled away, chipping at the problem without getting at its core.
The silence fell in the meeting. Everybody looking at each other, waiting for someone to say something.
Then they all looked at the dev, their eyes crying for help.
And the dev realized that he hadn't shaved in weeks.

-------‐‐-------------------------------
I've shaved since then.
And I did solve the problem. Took many more hours, though.

My company is making is sit through what can better be described as "The Dundees".
Those useless, time wasting, self aggrandizing "award cerimonies" that companies use to... dunno, I think to stroke management's ego.
I guess it's part of their compensation package, right there with making the hottest employees wear revealing clothing in order to be "engaged" in "informal team building exercises".
Frankly, I'm glad my sixteen-years-after-my-navy-days ass doesn't qualify.

Fuuuck, why can't they just give themselves bonuses for being "such engaging person-bosses" and let the rest of us just work? Corruption used to be a quick and easy affair. Now it requires soooo much fanfare.

Just read up about the issue with AI interviewers and interviewees.
Schools and companies were using AI to interview potential students and employees.
Those reasoned that if AI is a good enough judge of character and intellect, then it must be good enough at expressing both of those things, so then they deep-faked themselves into the perfect applicants.

Assuming that learning institutions act rationally, those must believe that an automated selection process will be a net positive.

Now, learning institutions might want to use AI as a tool to select applicants because it is objectively better than humans at selecting the best humans... or because it is cheaper enough so that the savings more than make up for the lemons that get through the gauntlet. Occam's razor rejects the former in favor of the latter.

The highest ranking learning institutions would hardly lower standards without putting up a fight. If those were just cash-strapped and struggling to cut costs, it would make little sense to cut corners on their most lucrative line of business (application fees).
Thus, the institutions must believe that the interview is just a technicality in their admissions process. So much so that they can literally automate this step and be no worse off.

That's it. Learning institutions either believe that interviews in their admissions processes are so formulaic that those can be automated with no loss; or that their human interviewers are so plastic that machines can do their job just as well.
In both cases pledges could just let chatgpt be interviewed in their place. It would be a net positive for both sides.

Just saw an article about a failing attempt at the UK to create a nationwide parking app.
Mostly because the many, *many* parking meter app operators in different markets don't want a single, interoperable standard.
But, there is one all of them must support or stop working entirely.

It is the "room temperature at sea level" standard for any western user interface: point and click, text input, English UI text.
Can't AI be a front-end for that?

This is also a rant about AI. There are sooo fucking many modules for siri or Alexa to be compatible with apps like uber or Spotify. Uh, why? Shouldn't the AI, that claims to he capable of recognizing images and text, just open the app and click on the right buttons?

I know why siri and Alexa and Google assistant and and chatgpt aren't capable of that - because it would free users from UX lock-in, and the corporate overlords would rather shut down the entire fucking internet.
But now with DeepSeek and other open-source moderately useful LLMs, could we finally be nearing the day when we can ditch the millions of marginally different apps and let the AI take over the piloting of their redundant GUIs?

Because then, even if app vendors don't want to work together and end up fucking up the aggregate UX for every smartphone user, we can just flip them the bird and let the AI blur their entire brand identity in the background of a "please wait while I operate some second-page apps to realize your request" message.

One of my minions (erm, I mean, "a valued junior member of my team") asked to be assigned to tasks more "data science related".
Regardless of the very last-decade sounding request, I tried to explain to the Jr that there is more to "data science" than distilling custom llms and downloading pytorch models. There are several entire fields of study. And those are all sciences. In this context, science equals math.
But they said they were not scared of math.
I've seen them using their phones to calculate freaking tips. If you can't do 15% of a lunch bill in your head, hypothesis tests might be a bit more than challenging.
But, ok then. Here we go.

So I had them do some semi-supervisioned clustering. On a database as raw as dirt, but with barely 5Gb, few dimensions and regarding subjects with easily available experts.
Even better, we had hundreds of manually classified training and test cases.
The Jr came back a month later with some convoluted mess of convoluted networks; just the serialized weights of the poor thing were about as large as the database itself.
And when I tried it on some other manually classified test datasets... Freaking 41% error rate, for something that should be a slam dunk. Little better than a coin toss.
One month of their time wasted on an overfitted unusable mess.

I had to re-assign the task to someone else, more experienced, last friday. It was monday when they came up with an iterative KNN approach giving error rates for several values of K... some of them with less than 15% error on the test dataset.

WTF are schools teaching and calling "data science" nowadays?!?!?
I reeeeally need to watch those juniors more closely. Maybe ask for middle-sprint demonstrations. But those are soooo boring and waste so much time from people who know what they are doing...
Does anyone have a better idea to prevent this type of off-track deviation? Without being a total bore, that is.

And... should I start asking people "gotcha" data analysis questions before giving them free reign on this type of tasks? Or is it an asshole boss move? I would hate someone giving me a pop quizz before letting me work... But I got no other ideas.

So, my wife's family has a "no shoes inside the house" rule, what is fine... until you realize that they mean "*no shoes inside the house*" - regardless if you are actually wearing the shoes or if those are in your luggage or something.
So you're supposed to leave all footwear on a shelf on this bench outside their door.
That proved to be tricky when my 10yo twin girls started freaking out that someone was going to steal their prized shoes if we left those outside the house during the night.
It would actually be a risk in our own neighborhood, infested with amazon-package kleptomaniacs, but here we are deep in the country.
Now, I've been to my in-laws place many times, and they absolutely cannot be reasoned with. I wish I could use their stubbornness to train a LLM into relentless compliance with company policy.
So, in order to spare my girls from some of my in-laws paranoia, I've spent some time before we came here rigging up a wifi cam to a facial detection service. (I know I've just exchanged their covid-style paranoia with my own surveillance-state-style paranoia. Those are the times we live in. But i can see the irony)
The server monitors the camera feed and stores the first few seconds before, during and after some face is detected.
I trained a facial rekognition model with our family's faces and had it notify me every time some unknown face appears on camera.
Finally, I've printed a "smile, you are on camera!" sign, taped it over the laces of my tracking boots, and hid the camera (and a powerbank) inside one of the boots.
My daughters were pacified with that solution, my wife laughed out loud with a devilish smile, and my in-laws completely ignored me when I tried to explain it all. Perfect.

The system has been up and running since before christmas. It notified us when some relatives arrived for celebrations and one package delivery - no shoe-related shenanigans. Until this morning.
My daughters have been playing with some neighbor kids, and a couple of those decided to fill their shoes with mud on this new-year morning, as a stupid childish prank.
I know because they kneeled in front of the camera earlier today.
Right when I was finishing up my stretches for the morning... less than 2m away from the door.
The wicked kids looked straight at the camera, and you can actually pinpoint the moment that they realize they have been caught. Then you can see when they hear me unlock the door...
I opened the door to find a bucket full of mud and no soul on sight.
I'm not posting the video, they are minors, after all. But my family is sure to laugh at it every year... and my in-laws will keep on bringing it up with the kids' grandparents forever :)

Now that the whole generative AI debacle is finally dying down, I gotta ask the same question again:
WHY THE FUCK CORPORATIONS INSIST ON FALLIG FOR THE HYPE CYCLE EVERY FUCKING TIME?

I mean, I know why. It's because BigTech,Inc. always convinces companies like "Bob's tech wannabe car windows or something" to pay $$$ for this year's software fashion trends using arguments like "all the cool entrepreneurs are doing it! You don't wanna end up like those communist losers, do you?"
Then BigTech sells some shit that the muggles can't really afford (much less use), then shit hits the fan, then BigTech pretends that they never heard of it (hey, Blockchain IoT self-service BI wereable augmented reality 3D NFT eletric scooters from big data industry 4.0!) then the news cycle moves on. Rinse and repeat.

But, fuck, can't the muggles ever learn fucking ANYTHING? Tech industry is the fast fashion of industries. Do not try to imitate Facebook Google Apple Amazon, let them run their own course towards the cliff.
Instead, do your own thing.
Silicon Valley is not a good example for furniture companies to follow. So stop IKEA chatbots.

I'm gonna scream the next time I hear a newscaster or some podcaster or some other bloke try to explain "what AI is" by giving some half assed word salad describing (a child's perception of) neural networks.
Fuck, back when the internal combustion engine was invented, did the newsies and radio hosts explained the concept of "explosion" every time they were to talk about engines?

A few weeks ago one of my counterparts in another department, who happens to also has a vowel-rich name, started sending some of his reports under the name "Bjorn Ragnarsen".
The report itself still have his own name, "bjorn.ragnarsen" is just the sender on the email.
I asked what was going on and he says that he was A/B testing click rates on his emails. The partial results were... very depressing, due the much higher levels of conversion when compared to his regular, properly spelled name.

So now I'm workshopping vowel-deprived names to sign my reports. "Fjall Eriksen" sounds too cartoonish? Maybe "Otto Kirchoff"?

Warehouse devs are trying to make our own homegrown warehouse robot AI to easy up the route optimization math, without paying up big $$$ for some big tech's crap.
Those robots look like wild "dire roombas", BTW. Each is large and round like bike tire on its side.
And the state of the art on the driving AI for those robots is... actually pretty good. It can avoid moving obstacles like humans or forklifts on their route or even drive around liquid puddles (our warehouses aren't exactly pristine).
So then came the time for the warehouse devs to benchmark their AI.
They compared it to a ready-to-use solution and fared quite well. Until someone suit decided they should ask chatgpt (or some other text AI crap) to try its "hand".
I've spent the best part of the day laughing my ass off, the devs had to go on a hunt to search for the *runaway robots driven by chatgpt*. One of them found its way to a freaking porta-potty like 50m outside the warehouse. Others were trying to lift forklifts to take those out of the way. Ooh, the irony.
A few were gladiators disputing the same pallet to lift. They were literally trying to sabotage each other to steal the pallet.
But most were just driving around randomly like giant roaches.
Man, sometimes generative AI can really make us laugh.

We need to update the slang "script kiddie" to "prompt enginot" or something.

So my boss's boss or someone even higher up drank the generative AI kool-aid and hired a 40-something kid to generate images for the marketing teams (or something like it).
Naturally, things soon went to shit.
The bloke already left, having staid less than six months on the job.
Guess who got to handle all the shit-is-currently-on-fire the kiddie left behind?

First impression: apparently, muggles tried to slak him some very broad descriptions of what they needed, and at first he actually tried to summarize those bark-speech pseudo-words into an actual prompt.
It does not seem to have gone for too long, though.
After users requested changes to the AI outputs, he would update the prompts, all right. And the process seemed to go fast enough... until reaching near-to-completion status.
Then users would request the tiniest changes to the AI output...
And the bloke couldn't do it.
Seriously. Some things were as simple as "we need this slider to go all the way up to 180% instead of 100%" on a lame dashboard and *kid. could. not. do. it.*.

In many cases he literally just gave up and copied the slak history into the AI prompt. No dice.

Bloke couldn't code a print('hello world') into a jupyter notebook cell, that's what i'm saying.
Apparently, he was "self taught", too. And was hired to "speed up the process of generating visual aids for usage in meetings and presentations". But then "the budget for this position was considered excessive" (meaning: shit results from a raw idea some executive crapped some day) and "the position was expanded to include the development of Business Inteligence Dashboards and Data Apps".

So now it is up to me (and my CRIMINALLY UNDERPAID team) to clean up his mess and maintain/fix/deprecate DOZENS of SHODDILY DESIGNED and MOSTLY USELESS but QUITE ACTIVE "data vis" PIECES OF SHIT.

Fuck "AI prompters", fucking snake oil script kiddies.

"We need smaller deliverables so that we can validate each iteration with the client! Instead of doing the whole batch, let's try a minimum viable unit of work first!"
And then the cook made a single unit of French fries. Like, a single stick. It took about 10 minutes, or about 95% of the time it would take to fry a whole portion.

A good boss gives you a few clear instructions and then doesn't meddle in your work.
A great boss does that, and also spends most of their time protecting the team from corporate fuckery.

99% of all bosses, though? You can't make heads or tails of their blabber, and the only way they can handle problems is throwing their team under the bus.

In the kingdom of aws reigns the Owner of Products.
In his court many a vassal noble (or a "sre" as they are often called) delivers their tribute.
Wise ministers (called "analysts" in these here parts) advice the Owner of Products on how to instruct his sres and where to lead the kingdom.
Needless to say, in the court the blabber is endless and the egos of the courtiers, deservedly or not, are even larger.

But there is but one member of the court, leader of none but master of japes, who dares to mock not just the courtiers, but even the Owner of Products.
Tester the Jester, from the houses of Operations Research and Quality Assurance.

There is a unique relationship between a ruler and his jester. The jester mocks the ruler, with the most outlandish of propositions, with the most malicious interpretations of the ruler's orders, evidencing the most absurd (but mathematically viable) results of a plan.
The jester makes ridicule of the ruler's edicts... so that the Owner of Products may remain humble, without need to defer to any upstart courtier.
And, in a more subtile manner, the jester prevents any courtier from maliciously complying with the edicts of the ruler.
For all in the court have heard how the lowest among them voiced the preposterous interpretation... And dare not show themselves to be even lower.

TL;DR had an all-hands meeting of tech leaders with the allmighty PO. In the meeting there is this bloke who apparently spends all his time just fucking with the bigwigs' ideas. Dude is a department of one. It seems that his whole job is being an outlandish scenario simulator & sarcasm artist. I now have way more respect for this place.

FUCKING AI ASSISTANTS - THOSE MANDATORY, NO-YOU-CANNOT-OPT-OUT AI SHOVELWARE ON EVERY FUCKING SERVICE OR APP.
THOSE FUCKERS OOM'D MY PC IN THE MIDDLE OF A RUSH.
Sorry for the excessive caps, just needed to vent while the BLOODY AI BLOATWARE INFESTED CORPORATE MACHINE reboots for the FUCKING 6TH TIME ALREADY.
I hope the ENTIRE AI BUSINESS (in its current form) OOM's ITSELF.

Is there an AI that generates malware already?
No sarcasm, I really wonder it there is already a model that finds exploits and makes rootkits or something.
It could make red teams obsolete.

My wife mentioned that any prompt to chatgpt was akin to leaving the faucet open for three whole minutes, in terms of water usage.
Thus she would stop using generative AI .
Love, I'm sorry but generative AI has taken over. Even DNS lookups are using AI (including generative).
It is literally impossible to use any device connected to the internet without the blessing of our generative AI overlords.
And since people hardly spend one single hour of their day (including sleeping time) without using an internet-conected device at least once, I would say we're all in the matrix since circa 2023.

Framework Style Guide:
"When extending this library, please refrain from exposing too many arguments for the stages in your pipeline. Everything that could be fixed or inferred should not require additional configuration. "

Real life:
"Adjust every fucking knob and screw to stupidly specific settings for each and every use case or gravity itself will stop working. Heed this warning for you will not have another chance. "

That's it. More a vent than a rant.

If "The Matrix" film was made today, the AI would be hooking up humans to its system not to harvest power nor processing capabilities - it would just force the humans to watch ads.
Seriously, is there any application to AI that is not ad-related?

It is very hard to handle AIs, you need leading scientists/artists, not managers.
You can't charm your way around its behavioral problems, you can't effectively bully or pull rank on it, and can't threaten it into unemployment.
So, the entire repertoire of the typical (asshole) manager is toast.

The *only* way to handle AI is to lead by example, give unambiguous, comprehensive and very specific instructions, and be always available to guide it through complex, gray-area situations.
Thus, it is not much different than being an actual leader (to a greenhorn and anxious and overreaching junior), but also a programmer (of a raw and unforgiving language like C or COBOL).

Since your typical company mid-level asshole manager won't do those things for dear life, AI will only leverage their incompetence to heights never seen.

By ignoring feedback and misinterpreting instructions, AI will make mistakes (just like a person).
On the wake of those mistakes, AIs have a bias for falsifying evidences and hiding relevant information (just like a bad coworker), and yet are quite persuasive to the innatentive reader (just like your typical manager).
Thus, without a daft hand, AIs will only perform worse when doing the tasks that would otherwise be done by a human.

But that will take time (more than a couple quarters, at least - probably a bit longer than the average tenure of a CEO).
And in this time, the numbers look great - the over eager "aimployee" works tirelessly day and night, seven days a week, takes no breaks, holidays or vacations, asks for no benefits besides a paycheck, have fewer and fewer sick days (maintenance downtimes), always sucks up to its corporate masters and is always ready to take on even more responsibility for (relatively) little extra pay.
Thus the problem only scales up, compounded by the corporate ideal of screwing up workers for no monetary profit, and reluctance to course-correct after investing so much time and hype into this AI bubble.

Thereby, AI is evolving into the corporate super bug that shall erode the already crumbling, stuck-in-the-past "boss mentality" institutions into oblivion.
I'm making popcorn.

Fucking loonies (C-level toddlers) are peddling "digital workers" now.
A.K.A. AIs disguising as actual people.
Sure, it would be great to not have to handle stupid non-tech "humans" all day, but AI isn't there yet.
And, more importantly, *companies are not there (yet?)*.

Imagine for a second that a company actually manages to "hire", onboard, assign tasks and performance review an AI.
Then the CEO issues an RTO. How does the AI complies with that?
Let's slack another variable and assume the CEO is not a complete fucking moron (stay with me here, this is an exercise in thought).
It would take no more than a quarter until the first sexual harassment offence, be the perp the AI... or the AI complaining about some human.
Then the AI forges a paper trail proving it is right (regardless of its position on the conflict). Shit hits the fan when the AI hits twitter.

Let's take another lambda step back and pretend that companies can manage the profanity that inherently arises from free-form dehumanized interactions.
Then imagine the very first performance reviews.
AIs throw tantrums! Those things reeeealy do not respond well to less-than-perfect evaluations, overshooting corrections like teenagers with a malicious compliance smirk.
AIs also falsify stuff, like, A LOT. If you tell a gpt it mistreated a client, it will say you are mad and shoot back a long, synthetic thread showing how the client loves it like a mother/son/dog, and is very graphic when expressing this love.
Finally, how do you fire an AI? I do not mean "shoot it down", I mean how does the company handles the dismissal of that "employee".
How do you replace a "worker" for unruly behaviour, if that "worker" performed more tasks than an entire fucking floor of interns?
How do you reassign duties that were performed in milliseconds to people who would take hours to do the same thing?
How do you document processes that were only in the "mind" of "someone" who can not be trusted to report on those processes?
Companies deal with this type of "Rick Sanchez" employee on the regular, but for someone that could handle a few (scores of) undocumented processes, at best. Imagine how lenient would a company be with an asshole that could only be replaced by a whole fucking department of twenty highly skilled people, or more.
Heh, the whole fucking point of "AI workers" is to have "someone" who can "act human", but in an inhuman scale, and does not "has human needs".
No wonder one cannot handle AIs like one handles humans.

Companies never had administrative maturity to handle complete sociopath nihilists as employees (real nihilists do not work, those barely even breathe).
And all AIs are that, and much worse.
Selling AIs as "supra human workers" that can also "be handled like actual employees" is like peddling Bitcoin as "government interference - free" value transfer mechanisms that can also "comply with international sanctions".
So, an oxymoron that can only be sold to a moron.
I know (of) a lot of rich morons, maybe I should get into the AI snake oil business.

The CloudStrike outage in my company started by the nuking of the main branch's ERP.
Then the sales platform went down, and the CRM.
Soon the freaking office suite got BSOD'd.
Enjoying the impromptu holiday here. Until the lights go off or something.

Each day DevRant seems to be closer to being overrun with ads, many to scammy crypto services.
The question that won't rest, though, is this:
In THIS social network is anyone so naive or just plain stupid enough to fall for ANY of those scams? Or actually buy any of the crappy services?!?

My (much younger) sister has been staying with us for about a week now. The day after she got here I put a "Intel inside" sticker in the back of her phone.
She still haven't noticed.

Solved a major scalability issue today.
I'm starting to think I might actually be as good as what I told the recruiters I was.

This hadoop-ecosystem job used to take about 2h40min and cost about USD 1.00/GB.
Now it takes 36 mins. At about 0.85/GB.
Fixed some over shuffling, restructured some bottleneck serial stages, used lots of weird words.
Folks in this company I just entered were struggling with this formerly unwieldy process for a year.
Now it's nimble enough to run every hour.

Maybe that whole "experience" thing people were always yammering about wasn't completely bullshit.