"Here's the sad state of the cloud"
also, 10% off new guidebooks!
I just released new editions of the Cloud Resume Challenge Guidebooks for 2024. These are the most up-to-date and extensive companion resources yet for crushing the Cloud Resume Challenge, including a new chapter on how (and how not) to rely on generative AI for this kind of hands-on learning. Grab a 10% off copy for you or a friend looking to “learn like an engineer” this year.
Now, onto this week’s Tech Things - the good, the bad, and the ugly.
What’s after cloud?
Here is a tweet/X that was tweeted/posted:
I don’t know Darren personally, but I respect him (he is responsible for the excellent CNCF project k3s) and he is on the shortlist of people whose opinions challenge me to stop and think.
By “the cloud” Darren makes it clear he’s referring to the big hyperscalers, specifically AWS, and I for sure sense that the infrastructure hype has moved on from them. Part of it’s the natural ebb and flow of the industry, part of it’s the interest rate environment, part of it is that the cloud providers really aren’t dropping paradigm-shifting announcements every year like they were in the mid-2010s.
I think the “serverless” paradigm (true scale-to-zero, usage-based managed services) was the last time the cloud suggested a genuinely new way to build applications. The trouble with it, from the cloud providers’ perspective, is it simply wasn’t the optimal way to make money. Huge customers weren’t going to blow everything up to adopt AWS Lambda at the scale it would take to make a dent in EC2’s revenue, and the smaller companies who were “good” at serverless tended to be cost-conscious. When I started at A Cloud Guru, a proudly serverless startup with an empty EC2 dashboard, our monthly AWS bill was less than our catering bill. We eventually sold to Pluralsight for $2 billion and AWS wasn’t taking nearly a big enough piece of that pie to be super happy about pushing serverless to the next unicorn.
So instead the cloud providers have retrenched on a whole bunch of incrementally-improving ways to run ML training jobs and load-balance Kubernetes clusters. “Serverless” has become an overloaded marketing term that conveys little more than “a managed service.” And as the rate of change has settled down, actual best practices are emerging and people can get to work.
As the infrastructure thought leaders are moving on to standalone database services and DIY cloud-in-a-boxes, the mushy late middle of adopters are finally figuring out how to get some value out of the public cloud for the first time.
I think that means that AWS now officially qualifies as Boring Technology and a whole lot of people are going to have great careers in it for a long time to come.
Verdict: Good Tech Thing
Imagine you are an artist and you are absolutely terrified by AI art. You are terrified that you are being made obsolete by a technology that is flooding the world with a faster, cheaper, more mediocre version of your life’s work. You are angry and you are desperate.
Along comes a very nice researcher from the University of Chicago who says: “How would you like to stick it to those mean ol’ AI image models once and for all?”
You say yes, you would love to do that, that would feel so incredibly good.
So the researcher shows you this new tool called Nightshade and explains that it can “poison” your images so that large AI models will think they are something they are not — for example, instead of what you drew, which was two anthropomorphic meerkats having anatomically implausible sex, the model instead might see a bowl of grapefruit.
The researcher says if you run everything you draw through Nightshade, and if enough other artists all run all THEIR art through Nightshade, then eventually there will be billions of poisoned images on the internet and the AI training sets that are scraping them up will get so polluted and their models will become so confused that the AI companies will have to break down and pay artists to license their art, the unpoisoned versions of their art, because in the long run it will be cheaper as well as The Right Thing To Do.
The researcher says yes, probably the model trainers will figure out antidotes for Nightshade’s poison, but also Nightshade will continue to evolve more potent forms of poison, and this will be a long fight but don’t you want to be part of the fight?
And you are nodding and saying YES, hell yes, I am in this til the bitter end.
And then the researcher says in a fine-print voice that ohbytheway there is one small catch, which is that the images you poison with Nightshade might look kind of gross and be full of weird pixel artifacts. That won’t be a problem will it?
As an artist myself I’m not sure why anyone would take that deal at the moment. You are messing up the one asset you have right now, which is the ability to create images that people might like better than what AI spits out, in hopes that your jankified images will somewhere down the line, maybe, but quite possibly not, make things more annoying for an imaginary future AI researcher.
I do not think this is a good trade. I think a better thing to do would be to make the best art you possibly can, and register with “Do Not Train” lists like this one if you are concerned about your art being sucked up by AI without your consent. Not everybody respects those lists, but I think they’re at least as likely to reduce your exposure as relying on something like Nightshade.
Ultimately we all know that the AI copyright / consent arguments are not going to be addressed technically but legally. The people who don’t want to hear this are just the next iteration of the people who thought “smart contracts” were a solution to anything. Human problems are messy and rubbing AI on them is somehow making the humanness even messier.
Verdict: Unhelpful Tech Thing
I’m doing these little roundups every week or so; subscribe here to get them in your inbox.
Google is still the team to beat
People really like telling you why they left Google. Ever notice that?
A new set of “Why I left Google” posts are making the rounds this week in the wake of Google’s latest layoff saga. Look at all the upvotes. Read the comments. The ones from 2012 and the ones from last week—they’re oddly similar, almost like shared liturgy.
Google has seen better days. Bureaucratic, slow-moving, weighed down by thousands of lazy, elitist employees who don’t care and aren’t trying. They lost the ability to ship great products a decade ago. I don’t respect anyone who works there.
It’s long been true in startup-y tech circles that working at Google is sort of low-status but having worked at Google can be higher status, and the transition between these two planes of existence is a ritual “Why I left Google” blog post where you make it clear that Google was not enlightened enough for you.
In most of the rest of the world, including where I live near Charlotte, North Carolina, Google is not thought of this way. Google is still a name bearing tremendous talismanic power. “She works at Google” is shorthand for “She is 99th-percentile smart and probably rich.” The Google logo on your health insurance card gets a quick double take and a look of newfound respect from the receptionist in the doctor’s office. A guy who saw The Internship, that Vince Vaughn movie-slash-Google commercial from the early 2010s, asks you if you really play Quidditch on the clock.
Mostly, your employer stands in as a meme representing your unattainable intellectual powers. Trivia night? Don’t go up against this guy. He works at Google.
When I told people around town that I was leaving Google to work for myself, they did not treat me like I’m transitioning to a higher plane. They treated me uncomprehendingly. I thought he was smart?
The thing is, the tech people who love to hate on Google and the non-tech people who still think of it as the world’s bastion of brilliance are more alike than they realize.
You want to know why commenters revel in the schadenfreude under those “Why I left Google” posts? It’s because Google is still, after all these years, the team to beat. (Apple and Microsoft are bigger by market cap, but Microsoft-bashing is passé and Apple’s internal dealings have always been too shrouded in mystery to make for interesting conversation.) A decade and a half of Y Combinator startups have taken their shots and not one of them has come close to building a Google Search-shaped business model.
Meanwhile, it’s fun to punch up at “Googlers” who are pulling down ungodly amounts of money to work at a leisurely pace— especially when they have a reputation for snobbery. It’s totally healthy to hate Google. They are the New York Yankees of tech. Everybody else should loathe them, in a good-natured, non-toxic, sports-villain kind of way.
It’s also fun, not to mention a massive flex, to let everyone know that you used to be a star player for the New York Yankees but walked away because it just wasn’t stimulating enough for you. That story never gets old.
Nah, actually, that’s too mean. I think the real reason people write “Why I left Google” posts is that for most of us, working at Google is the biggest thing we will ever do. Not necessarily “biggest” in the sense of “did the most good for the world”, but biggest in the sense of … just big. Enormous, whooshing bigness all around you.
Have you ever changed a flat tire on the shoulder of the freeway? When you get out of your car and stand by the side of the road, you realize you are in a world scaled for 30,000-pound tractor trailers, not for human beings. The white stripes between the lanes that usually flash by like little punctuation marks are actually ten feet long and thirty feet apart. Even the road signs look absurdly out of scale, like those textbook-sized large-print editions of bestsellers the library stocks for elderly people. And then one of those tractor trailers hurtles by at seventy miles per hour, inches from your fragile body, and it feels like God himself is screaming in your face.
That’s exactly what working at Google feels like at first. You realize very quickly you are inhabiting an environment that was not designed for humans to experience with their naked minds. Even the smallest problems—which customers are we allowed to send this email to? What countries will have special legal requirements around the deployment of this tiny feature?—are so mind-numbingly complex that they make you want to shrink down into a fetal ball on the side of the freeway and just let the traffic shriek past.
Oh yeah, and also half the traffic is driving on the wrong side of the road, there are people pedaling along the shoulder on unicycles, occasionally some of the cars just disappear for no reason, etc. “Thriving in ambiguity” is one of Google’s official corporate values. “Sloppiness at scale” was how I described it to friends on the outside. A coworker of mine used to say that Google operates like a Michelin-star restaurant that is secretly a gigantic Chuck E Cheese.
Over time, hopefully, you build up a cyborg exoskeleton of automation and human alliances that allows you to navigate this environment at a safe speed. (The safe speed is pretty slow.) You even get some things done. And the small problems you solve, like the white lines on the freeway, may look insignificant from the outside— but you know that they were really ten feet long and thirty feet apart. You also know that next time you cross the road you might get squashed like a squirrel.
If you survived a trip down Google’s freeway, wouldn’t you want to tell people about it? How could you not? You have to get some sort of closure on it or your brain might actually explode.
Verdict: I forgot what I was talking about.