Is Slack CPU and memory usage actually a problem? Sure, developers complain about it but they almost always have powerful PCs

They have one powerful computer and multiple apps which think that each of them has the powerful computer for themselves.

I mean why set the bar any lower for Altman? Shouldn’t AI be able to produce an executable that minimizes RAM and storage? If not, why not?

If it can’t produce an executable that minimizes RAM and storage for something like Slack, are we supposed to believe it can for something more traditionally power-hungry (browsers), or storage hungry (Call of Duty, iPhone apps)?

Shouldn’t AI be on the forefront for reducing resource usage in general (not just digital resources), or is just a giant exercise in induced demand / Braess’ paradox?

So far most of the datacenters are built in very convenient places and people will start to build them in inconvenient places like Sahara or Mongolia way before they will building them in space

Maybe. But for SpaceX, it’s more aligned with what they’re trying to do to just learn to manufacture them at scale and lob them into space. And one of the benefits there is the uniformity of it - they can treat them all the same, rather than dealing with a bunch in different geographies with different power issues, governmental issues, etc. That’s been one of the major issues with rolling out solar. In the US, there are >20,000 AHJs, each with different rules and processes. A huge constellation of satellites seems easier to reason about and build systems to maintain en masse, because it’s more uniform.

I’m not saying this is a good idea. I’ve got a lot of SpaceX stock, and I wasn’t really happy to hear the news, this is mostly me trying to understand why they might think this is a good idea, and brainstorming out loud, with a dash of coping. Seems most here think that it’s just stupid, but then, most commenters thought Starlink was stupid, iirc, and that turned out to be wildly wrong. But it might also just be stupid this time.

Building nuclear-powered and solar powered datacenters in places with low population density will still be cheaper. Do you think Mongolian government won't allow China to build datacenters if the price is right?

It might be easier in China but that doesn't help Elon or Americans.

Solar powered datacenters on Earth don't make sense to me. The GPUs are so expensive you want to run them 24/7 and power cycling them stresses the components a lot so increases failure rate. Once it boots up you need to keep the datacenter powered, you can't shut it down at night. Maybe for CPU datacenters solar power can make sense sometimes, but not for AI at the moment.

Nuclear is super hard and expensive to build. It probably really is easier to put servers in space than build nuclear.

The only purely military thing is rockets and everything space related, there's just no way private businesses would've poured so much money into it

Computers and internet being storage, processing and communication systems are clearly useful for civilian purposes

Twitter's back-end is written in Scala, but they used "better Java" style so an average developer should have no problems making changes

Anyway, what kind of features Twitter (or any social network for that matter) needs after it existed for so many years? Hacker News haven't changed a bit a it does what it does perfectly well

> Twitter's back-end is written in Scala, but they used "better Java" style so an average developer should have no problems making changes

You sound like someone completely oblivious to software development practices who somehow felt compelled to post opinions on software engineering.

Your choice of language is irrelevant if your goal is to maintain software. What matters is systems architecture and institutional knowledge of how things are designed to work. If you fire your staff, you lose institutional knowledge. Your choice of programming language does not bring it back.

“Your choice of language is irrelevant if your goal is to maintain software.”

It may not be the most important choice, but it’s not irrelevant. And whether the staff he fired had useful institutional knowledge is an open question. Didn’t he fire a lot of non-technical, recent hires and people likely to leave eventually due to his muskism? I’m not convinced that his initial firings are the wpest move he made. Sadly, being overconfident, he assumed the same model could be applied to government, a mistake that will take a long time to fix if it is even fixable given America’s overall trajectory and the fate of the dollar.

I generally agree with you but I think you were a little strong in your view that the OP was "oblivious." I only say this because an enormous percentage of companies hiring software engineers specifically with requirements of X years with Y language and W years with framework with silly name Z. I think they are also misguided in that but I think it is is too prevalent to say they are all oblivious but honestly that may actually be more of an apt description.

Language is relevant, Haskell or Scala written in functional style uses quite foreign terms and paradigms. Like, wtf is Kleisli arrow?

From what we gathered on the kitchen side he fired the most infrequent committers. Which statistically speaking would not affect the institutional knowledge much.

Senior ICs tend to commit relatively unfrequently compared to junior ICs who keep pumping tickets.

Not in my experience, as long as we are talking about ICs.

Or alternatively (assuming that's true) he fired the people who thought about what they commit and kept those whose commit logs look like: "push feature WiP", "fix", "more fixes", "push", "maybe this works?"...

Reportedly a portion of them were thinking so hard they did not commit anything at all.

Ironically, those may have been the staff with the most institutional knowledge. Seeing people argue, here of all places, that loc or commit frequency == institutional knowledge is … unexpected. New hires committing “whitespace cleanup” != institutional knowledge.

Someone had to actually write all that code and it inevitably shows up in the stats. People who work on the code most tend to know it the most. Although people in non-coding roles sometimes prefer to deny it.

Sure there had so be some frequent but low impact committers. But implying that people with lowest amount of code contribution must have more impact is ridiculous.

I mean, a staff engineer who stopped committing couple years ago? Yeah could be burnout, or could be some major contribution that's not in the stats. OTOH an IC on their second year in position who hadn't pushed a single line? Nah the institutional knowledge is safe without.

> From what we gathered on the kitchen side he fired the most infrequent committers. Which statistically speaking would not affect the institutional knowledge much.

This take is, quite bluntly, stupid and clueless. Do you think each commit reflects the volume of institutional knowledge of any individual? Unbelievable.

Twitter had thousands of coders, each certainly with varying cadence and style of committing. But variation goes only so much and taken in aggregate yes, the amount of commits/diff size is correlated to contributor prominence. It's kinda hilarious the "my -2000 lines" types deny the obvious.

Hey man, just wanted to let you know, I had to downvote a bunch of your comments in this thread, not because I disagree with you, but because your commenting style is unnecessarily hostile and abusive. You can politely disagree with someone without calling their take "stupid and clueless", or any of the other mean-spirited things you've said elsewhere in the thread.

Or there's a mole in Ukrainian intelligence who sold him out

It was the default method of contacting the dealers on Russian darknet when everything was a just a message board (hell, it was available without TOR) and not a proper marketplace

Russian Gosuslugi and Ukrainian Diya coming to US!

I don't get their strategy. While they are an true tech company - the marginal cost of servicing next client is essentially zero, their network effect is limited to a single city so it will be easy for potential competitors to enter the market.

Their products were feature-complete a long time ago. Why are they still in a growth stage and not making profit which should be easy for essentially brokers?