For any tool, you want to measure the productivity gains, not the usage of the tool itself. Are these companies really that bad at measuring the work that gets done? You don't care how many times the hammer was used. You care that the house got built and the time it took to build it at what level of quality.
... because they noticed first-hand that AI fails often and that hallucinations can cause problems for them.
It's very easy for a manager to say "just use AI". It's A LOT harder for the employee who will be personally held responsible for AI-generated work that contained mistakes.
I left a career of 20 years, and good pay, and part of it is AI being slammed down your throat. It is not work, which may be iteresting, on doing your own AI, its just enforcing that you are an end-user of it.
Fuck that, I am AI vegan at the moment, and happy. I may build a machine and find out how to fix the architecual mistakes in this new scene, but thats very different than being an end user of it.
We were burning through all hours of the day trying to get the customer deliverables out, but mostly burning donuts in the customer feedback parking lot and wasting an absolutely unthinkable amount of time in the desperately awful[1] ecosystem. I'd been measuring weekly sleep totals in single digits, for months, and I know everyone else was in the same boat. We're barely hanging together and it's getting ugly, we're missing deadlines right and left, hundreds of bespoke customers and just two or three actual engineers on customizations.
So there is a big all hands on Teams, with the President and everything. I get called out specifically by the big man. This was the exact moment I knew I needed to skiddaddle, pronto.
"How long have you been working on Client XX"
"Well, five months, but we're mostly going in circles. There's also the nonroman font problem, and diacritics."
"OK, how long would it take to do . . just a vanilla, just the simplest possible customization?"
"I dunno, no one asks for that. There's no quantifiable measure of the complexity of a print design. A few weeks?"
"See, what I am thinking, we can bring that down to a few days. Maybe even a few hours."
The room holds its breath.
"Has anyone here heard of . . . vibe coding?"
So this virtual space had already been one of the most silent virtual spaces I have ever experienced, and after that one single question[2], it felt like a full five seconds of complete silence passed. No one else made a sound for the rest of the meeting - the Big Boss actually asked if anyone could hear him.
After the perfunctory "well, nice meeting", in response there was not one "gbye!" or "have a good one!" or any sounds at all, except for the staccato blip of a hundred people leaving Teams.
I could hear the job applications being sent over the wires.
The thing is, that Big Boss, I actually liked him, a lot. He was by far the most knowledgeable person I have ever met at that level - but he was also under a nearly unbearable amount of pressure himself. I still feel kinda bad I couldn't fix all his problems . . but when it comes to DITA, at a certain point you have to realize that the ecosystem is making all of its own problems.
[1] The one time Claude (Pro Max) complained - like, actually complained about a technology task - was with DITA-OT XSL. I think DITA gave it some kind of psychological crisis. Same here, buddy.
[2] To which the answer was yes, a resounding yes. We heard of it alright, and use it, too, a lot. We're using every tool we can scrabble together, as fast and as hard as we can . . but with a tool specialist ratio of 1/100, there's no optimization in the world to fix that. You're so much better off training up customers to maintain their own stuff, which you SHOULD be doing in the first place regardless instead of sucking all their content as fast as possible to trap them in your garden. Which, I only realized later, was the actual business model.
For any tool, you want to measure the productivity gains, not the usage of the tool itself. Are these companies really that bad at measuring the work that gets done? You don't care how many times the hammer was used. You care that the house got built and the time it took to build it at what level of quality.
... because they noticed first-hand that AI fails often and that hallucinations can cause problems for them.
It's very easy for a manager to say "just use AI". It's A LOT harder for the employee who will be personally held responsible for AI-generated work that contained mistakes.
I left a career of 20 years, and good pay, and part of it is AI being slammed down your throat. It is not work, which may be iteresting, on doing your own AI, its just enforcing that you are an end-user of it.
Fuck that, I am AI vegan at the moment, and happy. I may build a machine and find out how to fix the architecual mistakes in this new scene, but thats very different than being an end user of it.
We were burning through all hours of the day trying to get the customer deliverables out, but mostly burning donuts in the customer feedback parking lot and wasting an absolutely unthinkable amount of time in the desperately awful[1] ecosystem. I'd been measuring weekly sleep totals in single digits, for months, and I know everyone else was in the same boat. We're barely hanging together and it's getting ugly, we're missing deadlines right and left, hundreds of bespoke customers and just two or three actual engineers on customizations.
So there is a big all hands on Teams, with the President and everything. I get called out specifically by the big man. This was the exact moment I knew I needed to skiddaddle, pronto.
"How long have you been working on Client XX"
"Well, five months, but we're mostly going in circles. There's also the nonroman font problem, and diacritics."
"OK, how long would it take to do . . just a vanilla, just the simplest possible customization?"
"I dunno, no one asks for that. There's no quantifiable measure of the complexity of a print design. A few weeks?"
"See, what I am thinking, we can bring that down to a few days. Maybe even a few hours."
The room holds its breath.
"Has anyone here heard of . . . vibe coding?"
So this virtual space had already been one of the most silent virtual spaces I have ever experienced, and after that one single question[2], it felt like a full five seconds of complete silence passed. No one else made a sound for the rest of the meeting - the Big Boss actually asked if anyone could hear him.
After the perfunctory "well, nice meeting", in response there was not one "gbye!" or "have a good one!" or any sounds at all, except for the staccato blip of a hundred people leaving Teams.
I could hear the job applications being sent over the wires.
The thing is, that Big Boss, I actually liked him, a lot. He was by far the most knowledgeable person I have ever met at that level - but he was also under a nearly unbearable amount of pressure himself. I still feel kinda bad I couldn't fix all his problems . . but when it comes to DITA, at a certain point you have to realize that the ecosystem is making all of its own problems.
[1] The one time Claude (Pro Max) complained - like, actually complained about a technology task - was with DITA-OT XSL. I think DITA gave it some kind of psychological crisis. Same here, buddy.
[2] To which the answer was yes, a resounding yes. We heard of it alright, and use it, too, a lot. We're using every tool we can scrabble together, as fast and as hard as we can . . but with a tool specialist ratio of 1/100, there's no optimization in the world to fix that. You're so much better off training up customers to maintain their own stuff, which you SHOULD be doing in the first place regardless instead of sucking all their content as fast as possible to trap them in your garden. Which, I only realized later, was the actual business model.
https://archive.md/2tLkm
Because “ittel take thur jerbs”