sdiupIGPWEfh

It's possible I've accidentally swallowed some labor union propaganda during my education, but that's a much too rosy interpretation of Taylorism. Line workers have always been up against pressure to be worked to the bone, before and after Taylor. As you've already pointed out, workers were getting away with soldiering pre-Taylor, and then weren't immediately post-Taylor. I recall something about subsequent labor strikes. The very working conditions he helped impose would have made the ideal scenario for him to identify just how much a shorter work day and short breaks were now needed. But then, at least in the US, we didn't get the Fair Labor Standards Act for nearly three decades later.

Also, Robert Owen had Frederick Taylor beat on the 8 hour workday front by like a century. Maybe Taylorism was, in practice, one of the last straws. Or maybe it was Taylor's scientific support that helped convince the legislature an 8 hour workday was needed. Personally, I have my interpretation based on the evidence, and I don't see it boding well for software engineering as a profession in our immediate future.

I kind of love this analogy, actually, because Taylorism and scientific management also had serious downsides for the employees involved. That kind of micromanagement leads directly to burnout, and the elimination of supposed "wasted movement", in search of repeated optimized motions, leads directly to RSI and workplace injury. But if the workers are just replaceable cogs in the machine, who cares? Certainly it boosted productivity.

> If you refuse to use LLMs on the job, you are stealing from your employer. Your boss has every right to fire you, and they'd be right to do so.

So now it's our chance to be abused to improve the bottom line.

Do you write the tests yourself? Or at least thoroughly read and understand the tests?

If I was told that, as a senior, writing code was now a waste of my time and I could only contribute by telling juniors what to do and reviewing their code, I think I'd have lost my mind years ago. Just like you say, at least any junior worth retaining will learn, maybe even excel beyond me in some particular niche or other, and establish trust and rapport. Claude, on the other hand, may behave like an expert on some matter one day and act like an idiot in the same area the next, so how can I give in to trusting it. Unless I'm doing something wrong and it's on me. The mandate is no more hand-written code. The claim is that if I'm any good, I should know how to squeeze consistent quality results from LLMs. The reality is that I find continuously arguing with and correcting a chat bot all to be mentally intolerable. I'd very earnestly like to understand if the problem is me and what I can do to change.

I will at least acknowledge that having full freedom to mix models according to which balance of strengths / speeds is appropriate to the task does help, both in terms ultimate quality and reduced review iteration. I won't go so far as to say it eliminates all headaches, but I'll take what wins I can get. All too regrettably, that's no longer an option at the day job, with managers wanting to move to a single standardized vendor.

> But there again, nobody is forcing you to use LLMs.

In the sense that I can quit my job at any time, sure. But my employer certainly requires LLM use, and even if I found another position that doesn't, I have no reason to believe that will last.

All I want is an explanation of how you're all reliably getting good enough (or better) results from LLMs. Surely us AI critics are doing something wrong, something much more specific than needing to "prompt harder", if we accept the framing that 90% of world class programmings are achieving that, while the rest of us are not. There simply has to be some specific, repeatable, and yes, teachable thing we're missing. So what is it?

> It never forgets to do something. And as far as I can tell, it does it best work all the time.

I wish that were true. Even Opus 5.5 still ignores instructions and finds things it missed on a daily basis.

I have, on a lark, reached 20 to 30 adversarial sessions a few times. For some tasks, the sessions will just never converge on anything that yet another session won't find fault with, recommending an alternative already ruled out by another session. Even if all the reasoning in between was documented, the new sessions will endlessly claim to find flaws in past reasoning.

Right, but in this case, exposure is a work requirement, which is a bit different. While an employer isn't going to require tobacco use, work involving asbestos exposure is still a thing (removal perhaps), and an employer knowingly or negligently requiring work activity without both risk disclosure and supply of proper safety equipment would face legal action.

Hence, if mandatory workplace use of AI inflicts harm, then I could imagine things playing out similarly. The damage slowly accumulates, may or may not be noticed in time, becomes difficult to pin down on a past employer, and when you finally realize you need to sue for damages, both the employer responsible and the system are rigged to drag out the process until you're hopefully old and dead before you have a chance to win your case.

...oh.

How can it be a fireable offense when it's actively encouraged and rewarded?

I was anticipating that having AI write code to pass tests (human and/or AI written tests) would be worthwhile, but in practice, I've found that even models such as Opus 4.6 Thinking, High Effort simply "cheats", or rather, fails to generalize much too often. It's occurred to me that perhaps I need some amount of randomness in the tests to keep the models honest, but it feels wrong. We'll see.

Anaphylactic shock, and possibly death, is a potential outcome from the worm breaking internally. Far too risky.

Steam Deck spotted, two minutes, seven seconds in. Seems to be getting a fair amount of use for puppeteering robots at Disney.

Isn't a rather good deal of color from feathers a result of "structural color", rather than pigmentation? I'd be curious if fossilized feathers could ever, in theory, preserve enough microscopic detail to guess at that.