Replaceable? No. Reparable? Depends on the extent, but even that's hard: the wings are full of hardware, and if you have to spend a year and invent a process for dismantling everything to get at the damage, it becomes financial infeasible. Even if you pull it off, you have new inspection requirements, operational limitations, etc.: it's not the same revenue generating plane after something like this.

There is a lot at stake. Emirates operates these with over 500 passengers. If that manifest burns to death on takeoff because a wing folds there will be hell to pay.

I think the end state of all this looks like game cartridges. If people could buy the weights of a frontier quality model in a high density, high speed ROM that ran locally plugged into an PCI-E or M.2 slot, all you would need then is a reasonably fast tensor processor with a little RAM for context. This is possible, and there is even a company (Taalas) with an early product. They have an online demo that is crazy fast.

You would buy one and use it for some time, probably years, and then discard it when the value of some newer version makes sense to you.

Not far from the truth.

The big implication from this is that making these inference engines isn't difficult and they can't be protected. That's the doom of all of AI money: the product is getting commoditized; you'll buy a Claude/ChatGPT/whatever in a box from Amazon in a few years.

I had one breakthrough DMT experience where I saw 'the machine elves' (I just saw what I describe as fast-moving fractals that I 'felt' were beckoning to me); but, we have matching experiences w/the other primary psychedelics: I only had relatively minor on-top visual distortions with even the largest doses of LSD (1500+ mcg) or mushrooms.

That said, everyone is different. I know that some of my friends absolutely lost their fucking minds on a few tabs of LSD and, purportedly, experienced wild hallucinations that I have to trust were real to them but haven't ever experienced myself.

Not surprising. Tesla is its 5th generation (AI5) processor, currently manufactured by Samsung and TSMC. I suppose they imagine there are others that will want to use these for their own purposes. Musk is creating his own supply of chips for SpaceX at his TX Terafab. Having terrestrial customers to absorb some of the supply and provide revenue as that ramps up the obvious thing to do.

Perhaps not. There are Chinese fabs that can't make HBM, but can make DDR5. Corsair is now (as of about 2 weeks ago) selling modules with CXMT RAM. That's the first time one of the "major" RAM module manufacturers have turned to Chinese chips.

That may be the answer to the DDR5 shortage over the next year or so.

We need to beat China at green energy by bulldozing our undeveloped land for solar panels.

Great. New fab in the West. It's just power stuff; silicon carbide FET and whatnot. Still, if concern for "sovereignty" is what it takes for bedroom community Europe to get off the dime, then good for them. Better than being industrial vassals of the US, China and Russia, which is where they've been heading. Feel free to make more "sovereign" fabs.

LOL

"(Genetic based) IQ test are not very accurate" the company says. No kidding?! You aren't able to measure the IQ of an unborn person? What a shock.

From outside, an IR imager should be able to see the heat from a leak. But who knows; one would imagine if it were that easy they'd have done it 7 years ago, when this crap Russian module's chronic leak issues began. My inner cynic thinks this is probably political; competent people aren't being permitted to deal with it.

They'll deal with it when some hatch or docking port blows out and sucks a couple people into space wearing NASA tee shirts. I'm rather certain about that much.

The IPO push is tied to their profit forecasts, and they expect to be profitable in Q2. They're revenue is growing rapidly and it's likely true they'll actually turn their first profit, so don't bet against it.

Sure. That could happen again. And again. It's even likely with these LLM operators. A few years from now the necessary hardware be lower cost and lower power. The models they've built will be cloned and surpassed by multiple competitors.

But in the immediate future investors will fill Anthropic's pockets. One benefit in all this is that there will be more scrutiny of the spending, which will create friction in the both the Buy All The Silicon and Datacenters Everywhere departments: the investors of both Anthropic and OpenAI will want to milk the value of tokens while spending as little as possible and avoiding risk.

What a funny thing to say about something that is literally all text. Match up the code itself with the commit message and the ticket that caused it to happen - we work in the most documented business there is.

If you don't force/write good commit messages then you get what you deserve.

If you don't force/use good issue tracking then you get what you deserve.

In general, AI now composes my commit messages. Then I delete 2/3 of it. Sometimes I'll touch it up a bit. So it is helping our process...

For every line of code in our repo I know who wrote it, when they wrote it, what they said about writing it, and why they started to write it in the first place. If you don't know those things then you (or your organization) are doing it wrong.