amluto

18 ms·

361.
by amluto 5mo ago
> In some plastics the monomer is toxic, but the polymerized form is safe. This is common. Isocyanates are a common example — isocyanate monomers are nasty, and the very light ones are very nasty. They’re used to make polyurethane, pol
362.
by amluto 5mo ago
> insanely toxic Some quick searching suggests that it’s toxic but not even close to “insanely toxic”. And it’s not persistent in the environment. If you want a genuinely nasty chemical, check out methyl isocyanate, which is some two or
363.
by amluto 5mo ago
The pricing is weird. If you want to buy à la carte storage, it appears to be $40/TB/month for business and $30/TB/month for enterprise: https://knowledge.workspace.google.com/admin/storage/buy-
364.
by amluto 5mo ago
Google Drive has gotten inexplicably cheap for Workspace users, too.
365.
by amluto 5mo ago
At least there is a Files app these days, and many iOS apps interoperate a little bit with “Files”.
366.
by amluto 5mo ago
I don’t see why sub-nanosecond sync is useful for Spanner-style ordering. Your average server is more than 1 light-ns wide! Your average cable from server to TOR switch is several light-ns long!
367.
by amluto 5mo ago
About 25 years ago I did some work on Xilinx FPGAs. The hardware was nice. The ability to get them free (as a business development strategy) as a research user was nice [0]. The quality of the software stack was awful . The FPGAs were g
368.
by amluto 5mo ago
This is also Uber we’re talking about. The company that famously developed a massively engineered ledger to track every event across the entire company, globally consistently, forever, in a single database. This definitely adds enormous v
369.
by amluto 5mo ago
When you use DeepSeek’s first-party API, you are giving them your token stream . This has some training value, but it also has incredible amounts of, well, business intelligence value. When you tell AWS your secrets or your customer data,
370.
by amluto 5mo ago
As far as I can tell, Micron and SK Hynix are using EUV lithography and may be constrained by availability of the equipment, whereas CXMT does not have EUV machines. There were reports that EUV lithography is needed for high yields, but CX
371.
by amluto 5mo ago
Wait a moment. How exactly is Starship supposed to usefully deliver a pile of FPV drones to a hostile area? Starship itself is highly engineered to survive reentry and can even land anywhere that an appropriately flat surface is available.
372.
by amluto 5mo ago
I’m reasonably sure that one can use AI as a fancy spellcheck plus grammar check — this of it as a copy editor, not a co-author. Even professional authors have beta readers and copy editors.
373.
by amluto 5mo ago
I don’t think plutonium is the right comparison. Plutonium is expensive, and nuclear bombs are neither cheap nor particularly useful for doing things like attacking 10k different targets in some foreign country. I’m imagining a launcher in
374.
by amluto 5mo ago
Am I the only one who read this and contemplated that it would be entirely straightforward for a Chinese AI inference provider to efficiently scan their input and output token streams to identify cryptographic keys, access tokens and intere
375.
by amluto 5mo ago
It’s an awkward comparison, but F9 can deliver a payload to orbit at a slightly lower price per kg than a Tomahawk missile can deliver it to a target. Starship would be MUCH cheaper if the economics works out the way that SpaceX would like
376.
by amluto 5mo ago
I would also believe that a robot could inspect and replace tiles a lot faster than humans.
377.
by amluto 5mo ago
I read the first link and it said: > When low-frequency sound becomes strong enough to be heard or otherwise felt, it can cause annoyance, discomfort, and sleep disruption like any other normal noise pollution. So which is it? Sure, I d
378.
by amluto 5mo ago
You can also get gadgets called “permeate pumps” that consume no electricity and help a bit.
379.
by amluto 5mo ago
Datacenters in space might help with energy (that’s a big if, but let’s suppose it works out). How does it help with chips? The chips are still made on Earth, but once you launch them, you add some radiation exposure and you also add a
380.
by amluto 5mo ago
The deal with Anthropic, even if good, doesn’t seem repeatable. Right now, quite a few companies are discovering that the can turn inference capacity into revenue. Anthropic also can turn inference capacity into happy customers and mindshar
381.
by amluto 5mo ago
The use of EBITDA for Starlink is also interesting. For something like terrestrial fiber, I can imagine thinking that there’s a lot of depreciation on the books, but that most of the equipment keeps working after the depreciation period or
382.
by amluto 5mo ago
The shared part has almost nothing to do with this, IMO.
383.
by amluto 5mo ago
Kind of. A vscode workspace can trivially execute code on the machine that runs the server end of vscode. (This is how building works -- there is no sandbox unless the workspace config explicitly uses some kind of sandbox.) So the workspa
384.
by amluto 5mo ago
The security measure that the developer didn't use was completely refusing to use vscode. vscode has no security model. It's not like swiss cheese where there are holes and some of the go all the way through. vscode is all hole
385.
by amluto 5mo ago
> But I'm not confident that even that guarantees adversarial robustness: for instance, there may be some clever way to efficiently implement a nonzero polynomial that vanishes at every small integer (say, every integer with absolut
386.
by amluto 5mo ago
It seems to me that one could maliciously construct a failure. For example: phi(1.0) * 2^32 = 0 So: 1.0 + … + 1.0 (2^32 terms added) Will turn into zero in the embedding. (I bet other, dramatically smaller zeros could be
387.
by amluto 5mo ago
“Far better tooling” means that you don’t need to do git help reset and try to remember each of the nonsensical choices, for example.
388.
by amluto 5mo ago
git rebase, for all its warts, was always better than mq. A failed mq-driven rewrite was destructive! (And it kind of had to be — if you were trying to edit revision 17, there was no number available for the original revision 17 because t
389.
by amluto 5mo ago
The branching model being keeping a record is fine, but needing to make up a name for your branch before committing (unless you like rewriting) was not awesome, and the names being a global namespace was unpleasant. And then, when you pull
390.
by amluto 5mo ago
I also did this. Both in hindsight and at the time, I thought Mercurial had far better tooling. But it was not all amazing: Mercurial’s branching model was very poor, and its sequentially numbered revision system was and remains a very bad

More ›