As it absorbs more or less moisture from the air due to ambient humidity changes, yes. The same reason your wooden furniture will also significantly change in dimensions such that there's no point in being more than 1/16th of an inch precise in your measurements; your design and construction technique is far more important.
If you're looking for an efficiency-focused harness, I had a pretty good time using the Dirac agent. The line-based anchors were slightly buggy though (this was a couple months ago) and would sometimes add the same line of code multiple times or leave an anchor in the output.
Running locally is even worse for this, because if you're running 4 jobs at once they just run at 1/4 speed. Not literally, you can make up some of the difference with batching, but you have limited resources instead of spreading your requests out on an API provider's nodes.
To quote a message I wrote on a finance channel on telegram:
The TAM for "enterprise applications" at 28 T sounds both too much and too little: by the time the tech (and/or overall economy) allows it to reach that number, that number itself will look unimpressive, and this kind of scale seems to be reachable with ground-based more easily than with space based (at current energy prices, even that TAM is only about 2% of being Kardeshev 1).
Feels like Musk did vibe-economics for "how big is the global digital economy?", much like the claims about factories on the moon making data center satellites looks like he prompted grok with "if I tile the moon with solar powered factories and mass drivers to launch them, how many TW can it launch per year?"
The world's GDP is about 100T. That would mean more than 1/4 of every expenditure in the entire world would go into buying AI or by AI providers into their consumables.
While I essentially agree Musk is BSing, TAM doesn't imply "we can actually get this entire market". The TAM for the food sector is *all food*, not what one particular alcopop manufacturer can sell: https://en.wikipedia.org/wiki/Total_addressable_market
AI today can't do all desk jobs, I don't know how far we even are from that given the spiky nature of ML, but it smells like this IPO is using that as the justification for the claim.
The TAM of a particular alcopop manufacturer is "all beverages sold on the area it distributes". Misrepresenting it as "all food" would be borderline fraud.
The TAM for AI right now is the sum of all revenue from all AI companies. That seems to be something around $100B, or about 0.3% of the number on that document.
If they plan to grow that market, that's a different indicator.
He personally and his companies have been sued for fraud, so I wouldn't put it past him to be doing so here. That said:
> The TAM of a particular alcopop manufacturer is "all beverages sold on the area it distributes". Misrepresenting it as "all food" would be borderline fraud.
If you're confident of this, you may wish to edit the wikipedia page I linked to, which I was paraphrasing, that distinguishes between total available market (TAM), served (or serviceable) available market (SAM), and target market (TM). To quote it instead:
For example, the total UK consumer expenditure on food in 2014, which is the total addressable market of food, was £198 billion (including catering, alcoholic drinks, non-alcoholic drinks and other foods).[3] The serviceable available market for alcoholic drinks, which producers of alcoholic beverages target and serve, is £49 billion.[3] Since the market for alcoholic drinks is not a monopoly, the market share for a company producing alcoholic beverages can never reach 100% of SAM.[4]
reply