An old coworker of mine got his first job out of school with IBM.
IBM hadn't done many layoffs ever at that point and apparently didn't have system for it. He showed up on layoff day and they laid everyone off with a very generous baseline layoff package, including substantial education benefits. So he just went back to school on their dime for several years ;)
Most early stage companies are now doing the same thing as well. It's basically a re-invention of the "build fast and test" model that was the norm amongst startups in the late 2000s and early 2010s.
Yes it led to some degree of tech debt, but it also made it easier to experiment, validate, and identify good and bad workflows.
At least in my network, we don't think AI will replace all workers and we strongly believe AI will lead to a significant amount of tech debt, but we do also recognize a lot of work in tech today is busywork and will be automated away in the hands of actual engineers with domain expertise.
How is it nonsense? Vision doesn't matter - customer feedback matters.
You start off with a hypothesis (X will solve Y's problem by doing ...), you build a prototype, and then you start testing with multiple Ys. Based on that feedback, you then tweak your initial hypothesis or you scrap it and pivot.
The whole point of engineering is to build tools that solve a specific class of problems for the buyer.
Most people are not Steve Jobs, and do not have such a vision. Or the ability and capital to see such a thing through to the end. Steve Jobs also had a number of visions that didn't work out so well too - he learned and interated from them.
Thus, for most folks releasing a product and getting to market/revenue is more important than anything else. Then iterate from there.
YMMV of course. But the older I get, the more I realize "just get it done" is far more important than almost any other metric there is. There is a ton of navel gazing in tech that provides negative value. If I had released some of the things I worked on in the past vs. carefully designing and polishing them, I might still be working on them today. Competing products have maybe 50% of the "quality" of even my prototypes of 10 years ago - but they exist in the market and are used every day by customers to generate income.
Chaika was not a copy of a Packard. (They certainly admired the Packard bodywork, but Soviet industry was in no way ready to clone a Packard sedan)
Tu-144 was not a copy of the Concorde. (Convergent evolution is not the same as copying a design!)
The Soviets did clone a lot of DEC gear but I don't think SM-1, specifically, was a DEC clone. (In this lastmost case, the Soviets were left cloning computer equipment because it was forbidden to export to COMECON states)
Sorry, SM-4 not SM-1, was a full emulation of 11/40, with UNIBUS, and all. There were DEC copyright strings latent in some system files. It was a pretty good copy, but quite unreliable, and the reason was quite pedestrian---the connectors! It was a good lesson on how the entire technology chain needs to be high quality for the final product to work well.
Another example I forgot: the first Soviet nuke was directly copied from the stolen Fat Man design. Of course later they did novel stuff, especially the fusion designs of Sacharov et al.
It is well known that KGB got hold of the Concorde blueprints, so yeah, not a direct copy but certainly a lot of influence in that design. Again. the details like engine performance made the difference: apparently Tu144 had to continuously use afterburners to stay supersonic. It was also quite unreliable---I've heard that towards its end of life it was just flying cargo and airmail.
The Concorde and the Tupolev both relied on afterburners, because they operated under similar design constraints -- the "western" jet engines in the Concorde were not that much better than what Soviet design bureaus could produce.
The Concorde was much smaller, and lacked one of the major innovations of the Tu-144 -- forward flap canards to improve handling on a larger jet.
Probably for the better. The Tupolev killed a lot of its passengers, and it was almost immediately withdrawn from service after the first few incidents. The Concorde, a simpler and smaller design, served for decades.
The Americans "hold my beer" and then later "you know what, fuck this". Classic example of bad choice, good choice. Overall the arguable made out the best with this. Boeing instead focus on 747 and commercial planes airlines actually wanted and damn near became a global monopoly.
This. It's fundamentally a social problem. The moment that reader mode becomes the default, they'll start gradually extending it with "useful" additions until it's just as bloated and painful again, and then we'll have some rebrand of the concept of reader mode, and the cycle starts anew.
"Why can't we have a functional version of the site for the blind, and the normal one for everyone else?"
Sure, but that's a big if: "Oh, just this one small thing for interactivity would be nice ... and this other thing ..." just like with how the early web expanded functionality.
beyond meat was a super cynical bet that ordinary non-vegetarian consumers would no longer be able to afford meat, so they would turn to meat substitutes even if they were more costly than meat had been in the psat
now they are publicly listed, and their cynical premise has not born fruit
TFA was published Jan 30, and ORCL's recent peak was last September. The stock continued to slide until the recent minimum Feb 5 and in the month since then has rallied 12%. Any possible moment to respond to the story is long gone.
You don't have to actually do #3. What most companies do is just get a UL certification (to reassure consumers) and put the label "no user-serviceable parts inside" on the case (to meet UL mandates for safety)
That's more than enough to avoid civil liability for user stupidity
Locking shit down is something you do for other reasons entirely
humans without credentials are bad at basic algebra in a word problem, ergo the large language model must be substantially equivalent to a human without a credential
thanks but no thanks
i am often glad my field of endeavour does not require special professional credentials but the advent of "vibe coding" and, just, generally, unethical behavior industry-wide, makes me wonder whether it wouldn't be better to have professional education and licensing
And that many mathematicians got monty-hall wrong, despite it being intuitive for many kids.
And being at the top of your field (regardless of the PHD) does not make you immune to falling for YES / EYES.
> humans without credentials are bad at basic algebra in a word problem, ergo the large language model must be substantially equivalent to a human without a credential
I'm not saying this - i'm saying the claim that 'AI's get this question wrong ergo they cannot be a senior software engineer' is wrong when senior software engineers will get analogous questions wrong. If you apply the same bar to software engineers, you get 'senior software engineers get this question wrong so they can't be senior software engineers' which is obviously wrong.
amusingly Motif and CDE were derived from HP attempts to copy Windows 2.x and the betas of Windows 3.0
not windows 3.1 -- windows 3.1 was popular! Windows before 3.1 was distinctly unpopular. It had basically no installed base. The only Windows 2.x applications I know of actually shipped an embedded Windows copy on the floppy disk.
HP was carefully tracking all the much less popular stuff Microsoft was doing in the late 80s because they thought this "WIMP" paradigm had staying powers, even if Microsoft was not exactly selling a lot of units
the common element between VMS (the subject of this post) and Windows NT, is Dave cutler.
Cutler lived in an extremely overcomplicated world of VMS kernel primitives, and given the chance to let his freak flag fly, he really overcomplicated his past work for Windows NT
In case you ever wonder why your 1 gb/s ssd has ~100 mb/s throughput on windows. there are often quite literally hundreds of layers of filters on even the simplest i/o
but it is super flexible! just slower than iced treacle. aren't you glad you had an object oriented I/O subsystem supporting microkernel services and aspect-oriented programming? i bet you use those features way more often than you read or write files from disk
These clowns live in a dreamworld created by their PAs and cronies
reply