Does Rust not have SIMD intrinsics that you could use directly instead of relying on auto-vectorization?

It does, thus (1)

Companies are heavily incentivized to do so so my guess would be yes. I also don't write the code by hand almost at all.

If warranted by the law then yes, you have to abide to it.

Are there any places that require winter tires by law that don’t have cold winters? I don’t know of anywhere in the US that even requires winter tires, although I think some states require you to have winter tires OR chains.

Sometimes the winters may be cold, some other times they may be exceptionally warm, and sometimes they're a mix of both. The law remains the same regardless what the winter was or is like. Countries that have exceptionally warm winters do not really have a winter conditions so I'd guess their law wouldn't have a requirement for winter tires or chains

Yes of course, anyone who lives somewhere with winter knows that some winters are colder than others. In the late fall when it is time to swap the tires, you don’t know in advance whether the winter will be harsh or mild, but people put snow tires on anyway in places where harsh winters are common.

It would only make sense to require them every winter if the large majority of winters were quite cold. As such, nowhere in the US requires winter tires, but someone else noted that Quebec does. This makes sense, since even a mild Quebec winter will spend most of the time getting below freezing.

I would be very surprised if there were a place that required winter tires by law where the winter does not always get and stay at or below freezing for significant periods of time almost every winter.

There are mountain passes in Oregon with posted signs during winter that say "winter tires or chains required". I'm not sure if you can actually be cited by an officer for not having them, but certainly you are a hazard to everyone on the road if you don't have them :)

Quebec, Canada mandates winter tires between December 1st and March 15th.

(But the winters are cold).

Didn’t know it was required, but it makes sense, as it is indeed quite cold. Good thing I always have winter tires on in the winter anyway, so I’m not breaking the law when I drive to Montreal!

I started using gemini with caution given the "much worse" benchmarking points it has gotten and still does but in practice there's very little evidence I found in comparison to claude models. It performs really well on non trivial tasks.

Worktrees are significantly faster to use and checkout than clones. They also use much less disk space, and obviously commits made in one worktree do not require git plumbing to get them visible in another etc.

They don’t use less disk space as long as you account for that correctly: clones within a file system use hardlinks for the object store, so if you du on a clone it’ll be however many megs / gigs S but if you do that encompassing n such clones they will only use a bit more than S for the object stores. And cross file system you can use “--shared” although that will break the clone if you remove the source (then again the same occurs with worktrees). It’s also not significantly faster: on a fairly hefty monorepo (a few gigs, working copy is 1.2G) a worktree takes 4.76s, a clone takes 4.94. The clone’s .git “is” 3.4G, but that falls to 5.8MB when du-ing both the source repository and the clone. Of course if you create every clone from the base repo it’s going to be slow as fuck and eat all your disk, but you can not do that.

The rest I agree with, that objects and branches are shared between worktrees, as well as the repository configuration (e.g. remotes) is why I use worktrees, having to move things around and losing the content when I’d delete the wrong clone is why I stopped using them.

Hardlinks are the wrong abstraction. You modify one copy and it changes on all linked paths. You want reflinks instead.

edit: spellchecker corrected reflinks, this is what I meant.

The difference here is that git objects don't change. When you do a `git gc`, you do get new objects at new inodes, but they don't get written to the existing ones. Try this out and you'll see local git clones do indeed create hardlinks:

``` cd $(mktemp -d) git init test ( cd test; touch README.md; git add README.md; git commit -m "Test"; ) git clone test/.git test_clone find test_clone/.git/objects test/.git/objects -type f -printf "%i\t%p\n" | sort ```

You'll see the same inodes used in both clones.

Yeah, but this only works for the immutable object store. The rest of the tree is still copied.

CoW and reflinks let you share everything, including the non-immutable paths, and even stuff like node_modules.

Not relevant to comparing clones and worktrees, both create a fresh working copy from the repository data, and object store files are not modified in place (or at all really).

You're about to be very surprised!

  echo old > a
  ln a b
  echo new > a  
  cat a b                   # prints "new" and "new"

Some programs like GNU sed [1] do create new inodes, but that is a property of the application, not the filesystem:

  echo old > a
  ln a b
  sed -i 'c new' a
  cat a b                   # prints "new" and "old"

With reflinks in a CoW filesystem the inodes are different from the start, but share the same underlying blocks:

  echo old > a
  cp --reflink=always a b
  ls -i a b                 # prints different inodes
  echo new > a
  cat a b                   # prints "new" and "old"

[1] https://www.gnu.org/software/sed/manual/html_node/Command_00...

Cool, I love surprises :D

Thanks for examples and especially the doc link. I don't normally use hard links, and for some reason was sure that's how they behave.

Maybe my belief indeed stemmed from observing software which overwrites opened file with a new copy.

No this is completely incorrect. The filesystem doesn’t break a link just if you open it. Your editor or higher level utility is free to do so (by unlinking and copying the data into a new file/inode), but this is not how the filesystem works.

Changing a hardlinked file changes it everywhere, unless your specific editor happens to create a new inode. Thus it depends on your editor.

CoW is similar except the new inode is created by the file system itself and is independent of the editor.

For anyone unfamiliar with this strategy: it's literally just "clone from a /path".

I wonder what the target audience for this price range is? At that price point I would rather go for a single H100.

I would rather take this option than a single H100:

> 96 GB Dual NVIDIA RTX PRO 5000 w/ 1000W PSU (non-refundable) +$19,169

Though I don't think there's anything stopping you from trying to stuff an H100 into this machine if you want to BYO.

For that price you get the H100. The machine is overpriced, and I'd personally rather go for a second-hand workstation with server-grade components.

Hopper is ~4 years old at this point though, compared to Blackwell which is ~2 years old, the difference isn't nil.

Depending on your use case, you might prefer native FP4 and FP6 low-precision support and 5th-generation Tensor Cores rather than what Hopper offers.

I think for training Hopper makes sense as it's generally a bit cheaper and the difference isn't that big, but for inference the difference widens a bunch makes a lot more sense to go with Blackwell.

You're correct, the difference isn't nil - H100 is data-center grade GPU with loads of bandwidth and compute while RTX PRO is a consumer grade GPU. The difference between the microarchitectures would make sense to point out if we had been comparing apples to apples and not apples to pits.

I would assume an H100 selling for less than $30k is a scam. I see one for 11,329€, and one for $25k but most are over $30k and I don't know how to judge these random storefronts. You've bought an H100 for $20k that was in working order recently?

What kind of comment is that? No, I have not bought the h100 obviously, and, yes, you can buy the new one for 20k. Used one for even less.

In certain, more complex, software engineering domains this almost became true as of today.

If I strip the dependencies out, I'm using it on a 5MLoC C++ codebase, and I found it performs really really well. I am using the Opus 5/Fable in parallel and I couldn't tell the difference. Both models make mistakes here and there.

I don't see how useful this benchmark at all is for tracking the progression of models. I am not intending to bash on you personally but this is useless. People who are using AI models everyday are for sure not interested how close the AI model can visualize the pelican but they are interested in how they will perform on their daily tasks at work or private use. Correlation between doing good on pelican task and doing good on actual work you need to do is close to zero.

Look at the difference between the Muse 1.2 and Muse 1.3 results.

I did. And?

1.3 is better.

The benchmark is irrelevant.

You do know it's mainly a joke, right? https://simonwillison.net/2026/Jul/16/kimi-k3/#what-can-we-l...

It doesn't appear to me so

Which part of "let's compare advanced AI systems being then sketch out a child-like drawing of a pelican riding a bicycle and then argue over whether or not they should be wearing a hat" isn't funny to you?

When the combustion engine was invented, and we got the means by which we could accelerate our trip by 10x, and at the same time scale it to multiple people, we didn't tell the humans to keep pushing the vehicles by their hands, didn't we?

As a developer you are more akin to a car mechanic, who’s still expected to know how the engine works even a century after it was invented, rather than a driver, who is just the user in this parallel.

Mechanics I've been to generally have no idea about how things actually in the car work - from what I could understand, area is full of recurring problems, and this exactly benefits mechanics to solve them through trial and error approach rather than understanding much how underlying things work. Pretty much close to what I would say SWE will turn into, there's no other choice IMO in foreseeable future.

I think you underestimate the skill of more experienced mechanics. Tractor (as in 18 wheeler) and master mechanics tend to know what’s going on under the hood in detail.

Like what exactly?

I’m sure there are many of those, but like everything, there are tiers. I’m sure we’ll get something similar in SWE, but my hope is that you’ll still be able to choose to be a high tier engineer if you wanted to

Speed doesn't mean anything if you end up driving off a cliff.

Evidence so far does not suggest people driving off a cliff so while a valid concern it's not a likely one. We also acknowledged that the benefit of using a vehicle is such that it outnumbers the risks it may introduce. We never said there will be no risks attached.

We’ve only been on this train for 3-4 years, and in that period, we’ve already invented terms for this kind of vibe-coded, highly-breakable crud. We’ve also had a few, relatively high-profile in their spheres bugs come out and hurt people’s real, lived experiences.

And that’s just the stuff where the house of cards failed quickly.

Software by definition has always been and will remain broken in one or another way. AI makes no difference or whatsoever. Arguably, it will help raising the quality of software.

No, but the US and UK did enact Red Flag laws.