It looks good for shareholder value for this quarterly report, when you stop giving big raises (no promotions) and when mobile talent resigns (no need to pay severance) both reducing opex.
The work they did in the previous quarter is being sold right now and driving revenue numbers up.
It's the work that they won't do next quarter that won't hit shelves for even longer that will impact revenue eventually that they're missing out on, but that isn't measured in their financials.
Basically. So many startups pivoted so hard to try and get some AI exposure it's absurd. An insurance startup is now an AI insurance startup, finance startup is now an AI finance startup, accounting startup is now an AI accounting startup, blockchain startups are now AI-blockchain startups.
Idk what you mean by the link, or the <s>, or what you don't know anymore, or what we can decide in this matter, or really what the matter at hand is, at all.
Focusing on the article:
Is there something I missed? Or did I get it right after reading 2x, and it really is just a random blog from 2 years ago advocating for bigcos to reduce headcount? If so, is that related in any way to Vercel cutting their sponsorship program for hosting for open source projects?
Basically you can read that above article as: Hey Mark, we've got a bunch of our money riding on your stock going up, so make your stock go up. I know in the past (2010-2022) we wanted you to grow like crazy and throw money around and grow some more, but like now with interest rates not 0 anymore and stuff, we need you to like start making money and you know, make the stock go up and up. K, thanks, bye.
I found it funny because these are the same VC who spent a decade throwing money at anything with first Cloud, then ML/Deep Learning, then Blockchain, then threw money like no tomorrow during Covid, and now are like, yeah, we want you to not do any of that stuff anymore, just like actually make money now.
The '/s' is an indicator of sarcasm in text, you'll see it in online comments from time to time. Hard to convey sarcasm online otherwise.
And yes, as the other guy said, oh you're an AI company!? Here's a few million. We'll see how long that hype lasts for these "AI" startups.
Interesting to think about. I did the math quickly, assuming you take the ~$12 and invest it instead here's what you end up with
$12/mo invested to return ~6% over 30 years:
End balance: $11,751.08
Total Interest: $7,431.08
Total not spent on Spotify: $4,320.00
Over a 50 year time horizon, the numbers become more fun:
End balance: $43,155.05
Total Interest: $35,955.05
Total not spent on Spotify: $7,200.00
Of course, bunch of assumption, 6% return, past performance doesn't indicate future return etc etc, and I doubt Spotify is going to exist in 30 in its current form, small probably it does, but odds are against that if I had to bet.
Maybe those boomers are onto something about avocado toast /s
Someone someday may actually ship code like this into production. Horrifying to think about. For some reason this reminds me of trying to grow plants with Brawndo https://www.youtube.com/watch?v=kAqIJZeeXEc
No one needs citations to voice a commonly shared perspective or experience on an internet forum, get real.
My bad, he only has 3,300 comments. That definitely makes things so much better!
The only BS here are comments, like yours and his, not even remotely attempting to add anything useful to the conversation and attempting to pretend to have some higher standard while not even being able to adhere to it yourselves.
Good luck. And I'm not saying you're wrong or anything, just that it won't be easy. Look at what happened in Canada when they increased capital tax rate few weeks ago. It was a ton of articles and opinion pieces about how it will destroy the economy and innovation. You try taxing more in US and it will likely be an even harder push back given the larger ultra-high net worth population.
What's interesting is that taxing the super wealthy would actually increase the health of the economy. They are leaches on the economy, hoarding up more and more money, land, assets, etc. just to make more money. Decreasing wealth inequality will help the economy.
And I think there's also a point of culture. I think the startup culture and Silicon Valley culture is damaging the American economy, hyper focusing on short term results. What startup has been actually influential in recent times, in terms of a benefit to society? There's very few, and there's plenty of examples of these startups harming things.
Lastly, these billionaires are actually a security threat. They start to become so powerful, with so much money and power concentrated, that they can literally affect laws, policies, wars, and even elections. Trump won because of Robert Mercer, yet another libertarian extremist billionaire.
It helps to think of our current super-rich as an outcome of the asset bubble that currently exists in our economy, due to the fed’s post-2008 money printing; because it’s stock that accounts for most of their net worth.
So what happens if we tax away the majority of that wealth? It’s a huge amount - enough to have a marked effect on the economic system. Well, we get a lot more money spread around the system, but we didn’t get any more products, services, or value being created. What do we get? Inflation. We have the same amount of everything - it just costs more.
So the super rich end up being a place to park inflated dollars to keep them out of the real economy. They’re actually a useful deflationary mechanism.
I’d prefer if we would get out of this asset bubble and return to a “normal” economy but this is the situation we are in.
Yes, it would. Wealth inequality drives all sorts of bad things in society and the economy. Reducing it, i.e., increasing wealth equality, would reduce the effects.
No you’re missing the point - the good intentions don’t matter because they have no impact on the monetary reality - what actually happens is you unleash inflation, just like Trump’s tax cuts, Bidens’s stimulus and ironically-named “Inflation Reduction Act” and of course the Fed’s 2008 money printing which we’re only now paying for.
The super-rich whose net worth consists of large unrealised gains in stock are simply a product of this bad fiscal policy and are, believe it or not, helping, by acting as a buffer to absorb asset inflation. Remember that their wealth is imaginary - stock they can never sell in significant quantities - by forcing them to sell it and pumping the money back into the bottom half of the economy, what was previously imaginary money would now be real money flowing through the economy, causing the price of things consumers care about (i.e. not stocks) to go up.
The only solution is to fix the fiscal policy and stop printing money and running up deficits. Asset prices can return to normal levels and the super-rich’s net worth falls with them because it’s all stocks.
There’s nothing wild about this position, it’s just monetarism. If you expand the money supply without expanding the productive output of the economy, all you’ve done is make money worth less than it was before. The number of goods and services hasn’t changed, the cost to produce/import them hasn’t changed, demand hasn’t changed. You simply have slightly more money that represents a smaller fraction of the total money supply.
It excels in only one area: scaling. You can "scale down to zero" (to the point where if there is no usage, you theoretically are not paying for stand-by hardware). You also do not have to -- theoretically -- worry about scaling rules and about provisioning more hardware when your service needs more workers to handle demand.
However, everywhere else it is a shit-show. Not being able to do anything resembling local testing & debugging requires you either: keep the architecture very simple (why use serverless in that case?), try and spin up one of the numerous half-baked virtualization solutions, or spend more time writing, testing, and maintaining scaffolding/mocking/whatever to interface between all your "add-ons" (e.g. SQS, S3, etc.). Your only way to track message flow is with extensive and expressive logging.
Deployment is a pain. What should be a sub 30second operation is routinely 5mins for even the smallest packages (AWS deployment tools are stupid inefficient) -- annoyingly breaking any sort of flow every time you have to deploy to test things you cannot test locally. Now imagine that being your working loop every day -- utter madness if you want to keep any sort of velocity or morale.
The ecosystem is not mature. I still have packaging utilities (built and maintained by a mag7) that fail silently causing production outages. The other tooling is also half-baked and a pain in the ass to use (much less learn the edge cases around). Serverless (the framework) is a shitty replacement for Terraform. The lack of a language server to understand whether or not my YAML IaC will actually do what I want it to do without dry-running is tedious.
Containers are wasteful, half-baked, and unperformant.
My workloads have always been either CPU or I/O heavy. No I do not need 2GB of RAM and a single (unknown spec) vCPU. I need at most a 100MB of RAM for my JVM/CLR and a fast-enough CPU. But the only way to provision a faster/less gimped CPU is to "bump the tier" of the lambda by provisioning more memory. Ergo you pay for memory you do not use nor need, simply so your lambda doesn't time out in its maximum 15min container lifespan, on heavy workloads.
The file handle limits are also something asinine, like 128 open handles per lambda with no way to modify. So I cannot open more than ~128 network sockets when I need to fan out compute to get past kneecapped container resources.
Cold-starts: it's been beaten to death. But if you're running a language with a bytecode interpreter your options are to either provision concurrency (i.e. force a container to always be warm/spun-up, and incurring all those costs, which would have been unarguably cheaper with a server, even an EC2) or modify your source and ahead-of-time compile everything you can. Otherwise, you will not get sub 300ms cold invocations (a sever in the most optimal location would get you sub 10ms latencies).
If you have long-running workloads or are trying to squeeze the most performance out of your backend: serverless is not going to cut it. This is ignoring all the inter-infrastructure communication that add even more latency.
N.B. this is for moderately complex web apps built on AWS, it may not be wholly representative of the landscape.
And apologies for the harsh language, but the entire "serverless" hype has given me plenty of scar tissue -- especially in a "move fast, now!" startup landscape.
Thanks for the reply. It was a good read. And I can agree on the points from experience.
One suggestion/correction: "Your only way to track message flow is with extensive and expressive logging.", you can do distributed tracing. It not a silver bullet, nor does it replace a proper debugger, but it's better than following logs. You can use a number of distributed tracing SaaS's, but you still have to do at least some manual instrumentation in your code to add additional info.
Who killed shareholder value? Why would employees do this?