Realistically email hosting is probably a safe-ish business. The required software is already more or less commodified(and the sort of people who pay extra for Proton rather than just using gmail or whatever their ISP points them to are the ones who specifically Do Not want you churning the UI all the time), just not desperately entertaining to administer; and the major challenge(even if you just ignore the amount of admin required to stay on top of security risks when internet facing and hope to get lucky) is achieving the necessary scale and stubbornness to get decent odds of delivery on the various major recipients of email who, at least initially, shoot the spammer and ask questions later or never if novel senders start popping up.

You could already DIY-mailsever in a fairly short span of time well before it became viable to ask a bot to do it; people just prefer not to because babysitting postfix just so that you can get rejected by the various services people you are tryign to reach use is not terribly rewarding.

It's not necessarily a glamorous or high margin sort of business; adequately competent hosted email isn't a wildly differentiated product; but people tend to not enjoy changing email addresses/repointing DNS/migrating mailboxes; so it's not a super cutthroat commodified market where people switch every quarter for the absolute lowest price. If you don't do anything particularly obnoxious or try to push price hikes there's a decent amount of inertia.

However, and this is what makes this guy's ranting seem so unhinged, it's precisely not the sort of business where you'd expect some sort of "AI transformation"; or a bunch of panting about whether the stack has velocity. Proton's config is a bit more fiddly than average, because it is supposed to provide better than 'just trust me bro' privacy assurances; but they do operate the fax machine of the internet age. We call it "email". And their customers are paying them for mailserver stay up; not the agentic everything app.

I'm not sure if this guy is butthurt because he didn't manage to score a nobel during his science time; or because he's not looking at an altman IPO; but the fact that he is screaming like a lunatic about “ruthlessly smash everything that stands in the way of AI velocity" when he runs something that is, as a feature, supposed to be a deeply boring utility suggests profound temperamental unsuitability.

It seems ironic, and a little telling, for the CEO of a company whose whole purpose is "mailserver stay up; securely" to be screaming at people about " Don't be the professional fax machine operator in the internet age".

Dude, you aren't some kind of randroid superman hero of industry. You run a kind of boring utility company. Which is fine; civilization runs on boring utilities quietly handled by competent people. Problem is that you don't seem to be fine with it.

Agreed. If you don't know what you're doing it's hard to even ask for the right thing or know what's possible. For example: I'm currently developing a system that uses a local AI to help debug internal issues. The initial incarnations of it were very slow to start up each command; clearly, it wasn't doing proper KV caching and was having to prefill every turn; it required patches on the llama.cpp fork and our API wrapper, and also adjusting the system prompt to keep variable content at the end. But if you have no idea about KV caching issues, how do you even know something is wrong with the backend, let alone what to tell the AI to fix? You'll end up deciding you need to upgrade the GPU instead.

(That said: "lovingly handcrafted" code is a hilarious phrase that should give any developer who's ever worked on a big corporate software project the giggles, and also, these software devs are clearly people angry at knowing they're going to be let go looking for revenge by badmouthing their boss to the press)

Farewell, mon capitaine. See you out there.

Google's fundamentals are web ads on its search.

AI search is more relevant and popular than ever, whereas web search results are less and less so. And AI search is proving to be less monetizable than web search.

Ergo, fundamentals have changed. More monetizable search is slowly getting deprecated, whereas less monetizable one is replacing it everywhere.

Millions of users don't have proper connectivity, they are stuck behind CGNAT and cannot self host or p2p. They are dependent on those walled gardens hosted by someone else.

The continued use of legacy IPv4 and slow rollout of IPv6 breaks that notion completely. Now you have a two tier system of users with full connectivity, and many millions of second tier users with partial connectivity stuck behind CGNAT.

These second tier users can't self host, can't p2p, and become dependent on additional third parties for things that first-tier users take for granted. Many of these third parties are abusive in various ways.

Back in the day you could host a low bandwidth chat server such as IRC, and high bandwidth activities like file sharing or video calls were direct p2p. Now many users cannot p2p so you need a high bandwidth and low latency server to relay all of that traffic. Only a few large players have the resources to do this.

Because there are so many second-tier users now this abusive system has become the norm, and many users aren't even aware of other options.

Escape from Pluto is much easier, and the Pluto-Charon distance is 1/20th of the Earth-Moon distance.

...Nothing on them but cranky bastards contributing to climate change while moving between ticky-tacky, 40+ hours of indentured servitude, and temples to Mammon.

I largely agree with most of Klee's points, but there's a serious bias issue with his conclusion: "The internet" is just the road we use to get somewhere. Where we go is up to us most of the time. And in this case, the "old internet" is still mostly there - Case in point, I still have a crappy static personal website up that hasn't been updated since the 1990s.

Problem is, nobody wants to read my ramblings from 30 years ago any more than they did 30 years ago. People want FAANGs, even while they may openly resent those same platforms. Sure, Amazon is evil... But the alternative is leaving the house and going to a store, eew! NetFlix kills everything good after two seasons, but two seasons on-demand beats a 30-second stick-figure animation that takes six weeks per "episode" and depends on the unpaid hobbyist creator not getting bored with the project. TikTok is measurably destroying the attention spans of our youth, but we can't risk missing the next big "hold my beer and watch this" moment.

The problem isn't "the internet". The problem is us as a collective whole. When there's demand for trash, sanitation engineers flip to supply side.

The primary reason companies advertise with Google is to drive search results to their site.

If AI summaries become good enough that nobody needs to click the links anymore - What are advertisers paying for? "Brand recognition" only takes things so far, at some point they need real customers to drop in and throw money at them or they won't have anything to advertise.

My company ships data dumps like this to multiple vendors on a monthly (sometimes weekly) basis.

My side of that equation takes security very seriously. The recipients... Well... When someone offers to send you credentials out-of-channel, the right response is never "Just email it, this isn't a shared account" followed by getting cranky at being told that ain't happening.

Normal industry: Your factory has an expensive milling machine and a cheap stamping machine. Thankfully, customers actually seem to prefer stamped parts for most cases and you can produce those quickly; so you keep stamping out parts and rolling in cash.

Tech industry: As above but your entire c-suite believes that the milling machine with turn into mecha-jesus and end history; so you are bulldozing everything you can to build more milling machines.

Your best engineers will be AuDHD, and AuDHD folk don't do well in these sorts of synthetic tests.

I can code in 20+ languages for over 12 different OS', write code that's good enough for CERN to run for particle accelerator data collection for a decade, have accelerated systems by an order of magnitude by understanding the poem space better than their previous engineers, but I am abysmal at exam style settings.

The software industry is riddled with SEs who write abysmal code but can pass a coding test easily. They're the folk that you have always needed to reject but didn't because they ticked a box.

I missed the part that said it involved leftists, fats, or lesbians.

There's an established way to answer that question. It's called market research. You talk to a lot of people and ask them what products would make their lives better. You try to understand their problems. That's what any competent product manager spends a lot of their time doing.

I wonder how much market research Ms. Mayer's company did? How many people told them, "I really want a product that tells me what to wear for Halloween so I don't have to think about it?" I'm going to guess somewhere between none at all and a number that's close enough to none to not make a difference.

That's part of the problem with the AI companies. (Not the whole problem, but one part.) They aren't just out of touch with the general public. They're out of touch with standard business practices. Instead of asking us what we want, they try to tell us what we ought to want. And they waste billions of dollars creating products that people didn't ask for and don't solve many real problems.