... can't service their customers because they don't have capacity and if they think about raising prices to increase profit that's another sign that they're going broke.
As long as you have a memory that only lasts 30 seconds it all makes sense! Study it out!
It's very simple. Tech companies have overspent on data centers so they're going broke but the data centers aren't actually being built so Nvidia isn't really making money and the revenue AI companies are reporting are totally fake; they're actually going broke because nobody wants it but they...
Drove a couple of hours to see a space telescope—this is the Nancy Grace Roman space telescope launching no earlier than September
Your comment got me thinking, and I got excited because hey, BE-3 issue, but BO's got tons of data on them with New Shephard.
And then I looked it up. New Glenn's BE-3U is pretty different than the BE-3PM they fly on New Shephard. Damn!
Still, I'm rooting for them!
Sometimes the point of a robotics endeavour isn't to do anything useful. Just something incredibly hard. Sorta like running a half marathon.
So, as a runner that also builds pointless robots... this all seems pretty normal.
I graduated HS in 1998. The school library did not have a computer until maybe 1996.
I didn't see a digital card catalog unless I was at the university.
This makes 1000x more sense than putting them in space.
Ugh, New Glen had a 2nd stage failure, lost the payload. Vulcan's had two SRB issues on their initial launches and now USSF is considering an option to let them launch payloads that don't need the SRBs.
Really hoping they get these rockets working right!
Don't feel bad of Opus cleaner than you. It's on par with the best devs I've ever worked with, at least on first-pass type stuff.
I couldn't make it churn out the junk I've seen people write over the years.
Good point, though that's more like 10 years back in AI tech. The general public doesn't see translation as an AI thing, just computer magic that works in everyday life.
Getting out to mow the lawn (half done!) has done wonders for my mental health today.
Also, EGO mower sat outside all winter, I popped in charged batteries and took off.
That didn't take long!
Random thought but I feel like a lot of AI hate is akin to doing someting like 3 * 0.1 on a processor and seeing that it doesn't equal 0.3 and pitching a fit.
Some obvious stuff it gets wrong but overall? Bangers.
I never saw GPT-2 but GPT-3 was flat out racist as hell. Sometimes "too dangerous to release" isn't about tech capabilities. It's about their failings and inability to align them properly.
Mythos is in a different ballpark for sure.
G = 9.8
if z_accel < G * 0.75:
robot_idle()
That's what took down bsky the other day. It wasn't vibe coding it was Jerry's vibes!
Yeah, that was my 9am thought. I'd like to mow the lawn and clean it up but it's a sponge right now. Can't mow and, honestly, out of fucks to give on another level about everything. Weather don't help at all.
Tao's broken barriers in maths with it. Knuth did some recent work that blew his mind. Dude in AU made an mRNA cancer vaccine for his dog. At least one human cancer drug is in trials that came from AI. Plus Ag which is benefiting from it in robotics.
I could go on.
That take, LLMs write code poorly, is laughable. Humans write poor code all the time. Most of my career has been spent hammering on code written by sub-standard programers.
LLMs whip their ass hands down in quality. They're up there with the best of us.
Fair, but you can just keep asking the tool to explain stuff until you get it. And by asking questions it might actually see a goofy mistake.
There are so many things I jump into, written by humans, that make my eyes twitch. I want to fix them for my own sanity, make the code not suck, but nobody's paying for it so I trudge through the bad stuff, fix what's needed.
With an LLM you can just clean it up in 5 minutes.
I'm basically an artisan fart huffer on the matter. I hate sloppy code. I hate inefficient code. I've been in the game 25+ years and I've seen humans put out utter garbage. The latest LLMs are on par with the best people I've ever worked with. Their first pass at something beats mine hands down.
When it comes to larger projects there's a whole process to validate stuff. George is capable of doing it for his "toy" tool I'm sure. When it comes to bigger stuff though we don't just ask the developer, no matter how they good they are, if it works. They aren't the answer to that.
Look I was doing most of my coding in 'vim' until 2017 and then I got cozy with VSCode. 'Cuz it has a vim extension.
I don't jump on new stuff but this AI coding agent stuff has gone bonkers since Opus 4.5 hit. I was mid on it before then but holy shit. I'm saying holy shit every day.
I did some extra-hours work tonight for a client just because I wanted to push Opus 4.7 a bit, see where it went.
Super fun problem to push on too. Mix of Ag, trucking, computer vision, and robotics. So up my alley!
To be fair in the good old days we had functional alcoholics instead of dysfunctional ones.
Washington was hammered out of his gourd frequently but he's still kicking Hegseth's ass.
Not excusing the implementation but I've coded for those things before. They aren't very friendly and things like loyalty or gift cards require dropping the encryption that they all insist on performing so you can get the raw account number.
I might revert to 4.6 after this weekend unless I start seeing 4.7 do good things.
True. I meant an immutable one. I hate PoW for currency but for something like this you just put the compute barrier high enough it doesn't make economical sense to fork the chain with a ton of compute and pretend you didn't sell a load of cabbage to Meijer. And if you do..
yeah, something like food supply tracking would benefit from it. It'd be nice to have a public ledger of where stuff came from. Just make it mandatory that nobody gets paid until a shipment is logged into the blockchain.