What's jarring for me about Isaacman's official Twitter is the way he describes his role: "Serving President Donald J. Trump." He does not see/present himself as a public servant. He presents his role as simply an instrument through which Trump's will may be imposed.

think of how much mileage we can get out of a swapped version of this image though

This is why we don't hang out down at that end of the scale, though

And for reference, all of your stuff about SSFL noted that cancer cases were mostly constrained to within 5mi of the facility. For reference:

Similarly, Falcons 9 and Heavy are basically the only functioning medium/heavy lift launch vehicles in the US right now. Atlas V is all sold out, Vulcan is in engine hell, Starship is Starship, and New Glenn, well...

This is a hilariously misleading depiction of what alpha particles even are. Alpha particles are literally just helium ions. Straight-up normal helium, just without the electrons around it. They're potentially damaging because they're fast, not because they're big.

literally the abstract

unsourced, quote-free AI slop article from a probably vibecoded website that runs AI generated clickbait youtube videos about how AI is destroying the world. Using LLMs to generate emails is singlehandedly draining the Colorado but using it to write fake space articles is fine, I guess.

Pretty good thirteenth test flight of Starship today! Apparently they're planning to attempt a catch recovery of the Ship on the next flight. Hopefully this can be the beginning of the end of the test flight program for this damn thing, despite the booster crash.

This started on basically day one of this administration. Usually when people talk about "posting coal" it's not this literal:

The best part is that this is basically just a very exaggerated version of exactly what's been going on in japan for decades

Really do think that the choice by frontier labs to promote AI image and video generation first was a massive PR error. Huge swaths of the population now associate AI with garbage slop images like on the left rather than the stuff on the right.

This closing paragraph kind of gives it away—the author clearly thinks AI is a technology doomed to go nowhere. He's shifting the meaning of LeCun's words here: letting LeCun state his (controversial) belief that LLMs aren't the path to AGI, and then using that to imply LeCun thinks it's impossible.

what, you mean the employees they hired and the contracts they got from NASA and the military to launch science, cargo, and our most sensitive military payloads? Let's please get real here. SpaceX is, right now, the US' only ride to space for anything bigger than a mini fridge.

GPA is frequently gamed in the US by both private AND public schools, see: UCSD math report. For reference here, MATH002 at UCSD is elementary-and-middle-school math skills remediation. senate.ucsd.edu/media/740347...

When you give it the same instruction and tell it NOT to think about elephants, meanwhile, it comes up with words corresponding to frustration and confusion (and also thinks about elephants and other animals).

Interestingly, it'll ping words related to things mentioned in the prompt even if the task you give it and the outputs it generates has nothing to do with it. In this example, the model is thinking of other elephant-adjacent words like "cute" even when the task has nothing to do with elephants.

They made a tool with open-weight model Qwen where you can mess with its J-space in real time and see how the answers change. In this example, by swapping "Yellowstone" for "Yosemite" in the J-space, I got the model to talk about an entirely different waterfall without changing the prompt at all.

250 years ago, our nation began as a promise that we’ve never quite lived up to. But for 250 years, we’ve fought to grow ever closer to living the ideals we profess. We fight this fight not because it is easy, but because it is hard. Happy 250.

This NASA will truly propose any mission other than Mars Sample Return lmao

this is very true however it is also equally true that Falcon 9, the rocket SpaceX actually launches for customers is one of the most reliable launch systems ever built, and thank god for that since it's basically the only functioning way the US has to get stuff into space right now.

they very much are not

I'd be infinitely more willing to give NASA the benefit of the doubt and assume that the all-male Artemis III crew was a quirk of mission scheduling if we weren't constantly being inundated by headlines like this:

Consider the following:

And as for the other point maybe we're just in different orbits but I see a lot of people on this website responding to any AI development with some variation on "nobody wants this/nobody will buy this/it can only make slop" It's becoming less prevalent but to cherry pick some recent examples:

I just cannot get these two graphs out of my head. Obviously this is constrained to just software and these are imperfect measurements of capability and utility, however I think it's very apparent that the impact here is real, and that the capability is high.

I highly recommend reading the entire article, but this graph is staggering. As Anthropic notes, code/person is an imperfect metric, but the quality of code Claude writes is both good and improving. Anthropic says it was a bit behind humans at the end of last year, and is about at parity today.

Easiest grift in the world is announcing a new Big Dumb Boat and getting the entire world to spend three weeks freaking out about it