rogerrogerr

It’s not about who’s developing the models, it’s the fact that free alternatives that are neck-and-neck are available at all. What’s the story for OpenAI & Anthropic’s valuation if they have to compete against free-weight models? Starts to feel like a commodity.

Holy shit, how’d I miss that? https://openai.com/index/gpt-5-lowers-protein-synthesis-cost...

Forget Iran, this place is what we’re going to wish we’d bombed.

The apparent bias may be explainable as it’s not remarkable for OpenAI or Anthropic to be slightly ahead. It _is_ remarkable for an open weights model to be better than the closed models from the trillion dollar (allegedly) companies.

Works for me on a major US cell network.

Pretty sure there are not 250 countries. How hard is it to validate the enormous number in the header image?

You can't interrogate the context of an LLM when you only have its output.

A lot of that sense was probably based on side channels like proper grammar, writing style, etc. That’s all gone now :(

I am implying that because it is true. SpaceX owns the $80B. Elon owns stock in SpaceX.

The IPO was a primary offering in which SpaceX issued new shares and received the proceeds. If Elon had personally sold shares, the prospectus would have identified him (or other selling shareholders) in a secondary offering. It doesn’t.

Musk's shares are locked up for a year.

I'm just correcting the false statements (probably an honest mistake) in the comment that said "this a long talked-about problem with the South Carolina factory, that the place does not follow aerospace standards and practices. The door plug failure was the highest profile QC miss out of that factory"

SC wasn't really involved in this. The fuselage came from Spirit in Kansas, and final assembly happened in Renton.

Nah, you can figure that out pretty easily yourself, and I get the sense your question is not in good faith. Earlier in this thread you called someone "profoundly lazy" for asking a less simple question.

There are many types of designation in the system; your quote above is _not_ about Production Certification ODA staff (which this article is about), and indeed there is no world where Production Certification ODA would be expected to find the MCAS issue your quote is about. Those would be Type Certification ODA folks.

You are confidently acting like you understand this process, while repeatedly confusing design certification (of the type) vs. airworthiness certification (of individual airplanes).

This is accurate, except “recertify” is the wrong verb. This is about signing off on individual airplanes.

Do you have wet clothes sitting in the dryer until 2am? Curious how you deal with that.

And I’m saying the electricity companies will have excess power, but will also have excess debt to service on infrastructure they now don’t need. They’ll get rate hikes approved to force the remaining ratepayers to pay down the capex.

Also, a bunch of that money they spent will be on transmission infrastructure that is 100% useless if you aren’t moving that amount of power along exactly that route.

No, we won't - there's significant capex on all that infra that will have to be paid down, and we won't have datacenters to help pay for it.

Re. toothbrush, I have to take the opportunity to give Philips credit for the Sonicare electric toothbrushes they made ~15 years ago. That thing just keeps trucking, it’s incredible. I go on 2 week trips and don’t even bring the charger, and have no worries about it dying.

It must have been crazy overspecced, I expected it to be a 5-year disposable piece of non-serviceable tech.

At least in the US, the dudes who pick you up in the ambulance do not have medical degrees. They have various certifications that are far less than med-school levels of debt.

There IS value being created, or people wouldn't use the tools.

While I do think there is value being created, I think this form of argument is not as watertight as it appears at first glance. Humans are very capable of behaving irrationally.

I think there is a lot of harmful, negative “work” being done by AI today. Creating fake videos for Facebook, running girlfriend bots, automated scams, etc. Even employees just trying to be at the top of their employers’ AI token leaderboards. (So last month, I know). There is legitimate value being created, but I don’t think it’s obvious that the positive value is swamping the negative value 10:1.

“There is value being created, or people wouldn’t buy the meth” - people do buy meth, quite enthusiastically, but any sane person would think allocating 23% of a nation’s electricity to a meth factory is a bad plan.

That’s what code review is still for.

You can have a separate LLM as a brutal enforcer of an architecture decision. Works pretty well in the right harness.

They’re very, very different projects.

I’m convinced the magic bullet is deterministic checks. Linters, static analyzers, etc. Whatever you can do to create deterministic gates that the LLM simply must overcome to reach a “done” state, do it. Has been making a huge difference for my team, but sister teams are so invested in writing the perfect Make No Mistakes prompt that they just can’t see it.

Basically I treat it like a junior dev. We don’t get junior devs to write code correctly by cajoling them just right, we add CI gates. It still works.

Any pointers how to get into that? I think I’d enjoy it. Was at a company that did manufacturing, and I was supporting it, but factories were in China and I was in the US.

Huh? Your own quote literally does not frame the center screen as "the car":

that does take over every screen of the car.

Eh, this isn’t really how businesses operate. How many businesses refuse to give devs large-spec machines? That’s very clear positive ROI.

I think it’s excessively charitable to assume businesses are uber-competent ROI-chasers. The expense people are eventually going to win on AI too, this blip of unrestricted AI budgets will be gone soon.

Does high school attendance matter for anything? Genuine question. Always seemed like pre-college schooling always wanted you to think everything was more important long-term than it really was.

He’s been remarkably consistent on the Mars thing for a looooong time.

The real answer? (bolded) It’s complicated.