biimugan

What's the availability story? Docker Hub has pretty severe rate-limiting even if you're not an anonymous user.

A "genuine ownership stake in the AI compute that's making things happen" sounds to me like corpo-speak for "taxpayer-funded bailout of my unprofitable company". After all, if everyone has a stake in AI, and AI crashes, then everyone (not just OpenAI) loses their money

This isn't a binary distinction -- safe vs unsafe. Things you put in your body have a risk profile. The risk profile of raw milk is much higher than pasteurized milk, regardless of how hygienic you think you're being. Cows step in mud, their own feces, the feces of other animals, dirty water, and many other things that can splash up and onto (and into) their udders, contaminating their milk with pathogens.

Letting people make their own choices always has its limits, regardless of what people say (rather casually on the Internet). When nearly all health systems in the world work through the healthy subsidizing the unhealthy, we should be attempting to limit preventable illness.

In any case, I don't think there's any country or state that bans drinking raw milk. If you're on a farm and you want to drink your own milk, go ahead. Just don't claim it's safe enough to sell, because it really isn't.

Yup. I never even use activate, even though that's what you find in docs all over the place. Something about modifying my environment rubs me the wrong way. I just call ``./venv/bin/python driver.py`` (or ``./venv/bin/driver`` if you install it as a script) which is fairly self-evident, doesn't mess with your environment, and you can call into as many virtualenvs as you need to independently from one another.

``uv`` accomplishes the same thing, but it is another dependency you need to install. In some envs it's nice that you can do everything with the built-in Python tooling.

In addition to what others have mentioned, it also just makes it easier to come back later to a code base and make changes, especially refactoring. In many cases you don't even really have to add many type hints to get benefits from it, since many popular libraries are more-or-less already well-typed. It can also substitute for many kinds of unit tests that you would end up writing even 5 years ago. If you're an infrastructure engineer or data scientist that's usually just writing a lot of glue code, then it greatly helps speed up your output (I've found)

Just run the code that provisions the infrastructure? Sandboxing is the least of your problems. You would need to fully mock out all function executions and their results to have a hope to properly execute the code let alone govern what's happening without affecting a live environment. And even still, there would be ways to fool this kind of introspection, as I mentioned. In an enterprise environment where this kind of governance is mandatory, that's not acceptable.

In any case, regardless whatever clever method you try to use, even if you're successful, it's not as straightforward and easily understood and extensible as OPA policy. Let's say you succeed in governing Rust code. OK, but now I have developers who are writing in Python and Java and TypeScript. What now? Develop a new, customized solution for each one? No thanks

I agree somewhat with the proposition that YAML is annoying for configuring something like a workflow engine (CI systems) or Kubernetes. But having it defined in YAML is actually preferable in an enterprise context. It makes it trivial to run something like OPA policy against the configuration so that enterprise standards and governance can be enforced.

When something is written in a real programming language (that doesn't just compile down to YAML or some other data format), this becomes much more challenging. What should you do in that case? Attempt to parse the configuration into an AST and operate over the AST? But in many programming languages, the AST can become arbitrarily complex. Behavior can be implemented in such a way as to make it difficult to discover or introspect.

Of course, YAML can also become difficult to parse too. If the system consuming the YAML supports in-band signalling -- i.e. proprietary non-YAML directives -- then you would need to first normalize the YAML using that system to interpret and expand those signals. But in principal, that's still at least more tractable than trying to parse an AST.

Are we already forgetting the Lebanon pager incident? It only happened last year. It seems perfectly rational to question the home country of a spyware company (of all things) when that country has in recent memory infected the supply chain of commodity electronics to booby-trap pagers and walkie talkies to explode and kill their putative enemies.

Let's also not forget that that country's paid spokesmen (both Israeli and American) were joking for months about those explosions, even lobbing it as a threat to their perceived enemies -- on live TV no less!

I think people should be highly skeptical of articles like this, even without knowing anything about the subject in question. No byline/author. No citations/links to the studies in question. Confirmation of preconceived notions that people would like to be true (e.g. the sun as a wellness remedy instead of damaging to skin), including unfounded "just so" stories and claims about evolution, diabetes, and other unrelated topics. Named individuals seem to "specialize" in sunlight as a wellness remedy (seems like a big red flag to me). No actual physical theory as to how it could be true (more vitamin D reduces death by up to 50%? how? your body only needs so much vitamin D and it's not actually all that much).

And sure enough, if you look up any details on the studies in question, they are highly questionable. Vastly different populations studied with very weak controls. For example, sunscreen use -- both chemical and physical, i.e. hats -- was not controlled for. Seems like a big problem since that's the primary claim being made! And it seems like such an obvious thing. It makes one wonder why it was omitted.

The facts of the "status quo" of sun exposure dangers, on the other hand, have quite a lot more going for them, both in terms of study quality and in terms of physical explanation/interpretation. UV radiation physically damages DNA, even when you don't burn. Tanning is a response to skin cell damage, so any additional melanin production in your skin is indication that your DNA is being damaged. Damaged DNA means when your cells reproduce, they reproduce the damage and/or otherwise mutate. If that damage or mutation happens to be cancerous, then you have a big problem. Tanning, contrary to what people seem to think, doesn't inoculate you against skin cancer or damage. It merely helps absorb a higher percentage of UV radiation -- meaning your skin is still getting damaged, just at a slightly lower rate (a helpful, though marginal, evolutionary advantage).

I'm not saying any schmuck could have done that. I'm saying that the engineers and managers at Apple (to use your example) are just as (if not more) responsible for the success than Jobs. Those lower-level engineers and managers also explain the repeated successes. And that, I would say, is the case in most market successes. The CEO is not remotely deserving of all the credit, or even most of the credit, in most situations. They don't really deserve to be paid what they're paid.

There's this certain anti-historical proclivity to create heroes for worship. Because it's a simple story to tell and it gives you the opportunity to put yourself in the hero's shoes. But the simple story is almost always wrong.

The only difference was leadership.

That's quite the leap though, and is just confusing correlation and causation. Maybe the previous leadership was simply getting in the way of the engineers and managers that had the good ideas. And the new leadership was more hands-off, or focused in other areas like marketing. Or those cases are just flukes. For every case like the ones you cite, I could find two where the exact opposite happened.

If you're downing a shot of vodka every morning, and suddenly stop, then yeah, your health is going to improve.

In my opinion, many (if not most) of these CEOs are business-focused people with no technical (or even non-technical) knowledge of anything they purport to manage. And on the whole, they really don't affect the value of the company one way or the other.

You don’t have to randomize the first part of your object keys to ensure they get spread around and avoid hotspots.

From my understanding, I don't think this is completely accurate. But, to be fair, AWS doesn't really document this very well.

From my (informal) conversations with AWS engineers a few months ago, it works approximately like this (modulo some details I'm sure the engineers didn't really want to share):

S3 requests scale based on something called a 'partition'. Partitions form automatically based on the smallest common prefixes among objects in your bucket, and how many requests objects with that prefix receive. And the bucket starts out with a single partition.

So as an example, if you have a bucket with objects "2025-08-20/foo.txt" and "2025-08-19/foo.txt", the smallest common prefix is "2" (or maybe it considers the root as the generator partition, I don't actually know). (As a reminder, a / in an object key has no special significance in S3 -- it's just another character. There are no "sub-directories"). Therefore a partition forms based on that prefix. You start with a single partition.

Now if the object "2025-08-20/foo.txt" suddenly receives a ton of requests, what you'll see happen is S3 throttle those requests for approximately 30-60 minutes. That's the amount of time it takes for a new partition to form. In this case, the smallest common prefix for "2025-08-20/foo.txt" is "2025-08-2". So a 2nd partition forms for that prefix. (Again, the details here may not be fully accurate, but this is the example conveyed to me). Once the partition forms, you're good to go.

But the key issue here with the above situation is you have to wait for that warm up time. So if you have some workload generating or reading a ton of small objects, that workload may get throttled for a non-trivial amount of time until partitions can form. If the workload is sensitive to multi-minute latency, then that's basically an outage condition.

The way around this is that you can submit an AWS support ticket and have them pre-generate partitions for you before your workload actually goes live. Or you could simulate load to generate the partitions. But obviously, neither of these is ideal. Ideally, you should just really not try and store billions of tiny objects and expect unlimited scalability and no latency. For example, you could use some kind of caching layer in front of S3.

Presumably the superior solution is the product that bears the same name as this blog post. Which I take it is in the process of being released since I can't find many technical details about it.

That builders can't construct housing in whatever form-factor they want and whatever part of a municipality they want might be a problem. But the analysis of why these kinds of restrictions exist, to my mind, is not correct.

Do niche interest groups have influence on housing policy? Sure. But these niche interest groups don't usually have a monetary interest in the outcomes they're promoting. But individual homeowners (who often band into interest groups of their own) and large real estate conglomerates do have such a monetary interest. They consider certain types of housing built in a certain way to have *more value* and be worth more. So they promote politicians who introduce zoning and other rules to protect that value.

For someone looking to make a profit off of housing (or even to invest in housing), what is more appealing? A traditional U.S. suburb? Or a Kowloon walled city? One is denser, cheaper per capita, and (if you're not careful) unappealing to look at. In other words, it's worth less. So there is a great monetary pressure from people who already own homes to prevent "mixing" this type of housing into existing planned communities. People who need homes, on the other hand, are a little bit less discerning (to say the least). They don't have a monetary interest necessarily. They're primarily looking for a permanent residence.

So I just don't buy this "abundance" stuff in general. If you remove all of these restrictions, will some companies start building housing? Some will. But my guess is most will say the juice is not worth the squeeze -- the profit margins and the long-term values of these properties will make it unappealing. Just like it's unappealing for grocery stores to set up in big urban areas. Or for hospital providers to set up in rural areas. Food deserts don't exist because of too much government intervention. A lack of rural hospitals is not a problem because of too much government intervention. It's because those things are not profitable.

So in my opinion, if you want to reform zoning rules or things of that nature, it's only really going to be effective if you *force* (or if you want to be politically correct, "incentivize") housing companies to build in these areas too.

I don't really consider this an "anti-trust" argument. It can be equally true if there's a lot of competition in the housing market and if there's next to no competition. It's more of an incentives argument. This is an argument that, like with medical care, we're treating something that is a fundamental need of every living human to have a stable and peaceful and fruitful life as if it were a standard market commodity. And when you do that, you get poor outcomes. We need to support the building of housing *even if* it's unaffordable or has low (or even non-existent) profit margins

I'll just note that this seems entirely predictable. So much so that I can't help but see it as purposeful. The federal court system itself only has about 25,000 employees. SCOTUS has 9 judges plus a couple dozen clerks and other assistants. Lower courts already do not have enough employees to contend with an executive branch made up of millions of individuals, especially when that executive is ordering its employees (seemingly) to just ignore or purposefully misinterpret laws, leading to an ever-increasing number of lawsuits. To further reduce the power of lower courts at this time (which this SCOTUS seems to do in almost every decision involving the executive) means even more cases for SCOTUS and even less time for arguments.

Conveniently, we have the shadow docket. A way to issue diktats without any arguments before the court and in many cases without any reasoning whatsoever.

And conveniently, lower courts can then interpret a lack of details from a shadow docket decision however they want. So that the executive can appeal yet again, to get another thumbs down from SCOTUS (without any explanation), and round and round we go. The executive gets to keep the plates spinning while it essentially does whatever it wants.

With this situation, why shouldn't we simply pack the courts? If SCOTUS is going to take more cases than it can handle and not provide any real guidance to lower courts, then clearly they need more employees.

In the sense that LIDAR can detect some things that visible light sensors cannot, it can make up (in some respects) for the lack of those other capabilities. It's not meant to replace them per se, but to provide another dimension of detection.

I've never understood this argument. When I drive, I'm relying on more than just vision. I'm relying on sound, tactile feedback, an understanding of traffic law, an understanding of human behavior on the road (some of which is specific to my locality, the time of day, etc.), an understanding of weather conditions and geography, past problems I've run into on the road and heuristics I've learned over time to improve safety, weighing the risks vs benefits of certain maneuvers (especially considering how specific cars around me are behaving), and so forth. Some of this a purely vision focused FSD could simulate. But much of it it cannot. That's why alternate sensing systems really are needed, whether it's LIDAR or something else.

This is decidedly not equivalent to those things, even roughly. Interpreting a statute (which by the way this Supreme Court has expressly limited, which is ironic given your comment) is quite different from interpreting an amendment to the Constitution. What you're describing is like saying a president can re-interpret the meaning of 'militia' in the 2nd amendment and start confiscating everyone's firearms because they're not members of a militia. Even despite rulings like Heller. This is exactly the argument that the dissent in this case makes.

What you say may be true with respect to breaking laws. But illegal immigration is one of those relatively small infractions, and only now is there some sudden insistence to prosecute all of them and deport them. So this is a self-made problem.

All of the evidence available to us shows us that migrants, on average, commit less crime than U.S. citizens. The evidence shows us that they pay into social programs without reaping almost any benefit. The evidence shows us that they take jobs that the average American isn't interested in. An evidence-based political program would not target migrants as a first priority, except to provide some more straightforward way to become documented and legal.

The other issue is -- the U.S. has 300 million+ citizens. This argument that migrants will "clog up" the courts seems ridiculous if you also believe U.S. citizens deserve due process. If your court system can't handle a relatively small percentage of your residents committing the crimes you have on the books, then maybe those crimes aren't really serious crimes are they? Or else not funding the courts appropriately to satisfy the political program is purposeful. The goal is to avoid due process and accountability, for citizens and non-citizens alike.

They are bothered by it. But the anti-liberal, extra-judicial, law-ignoring method this administration is levying against Harvard is also being levied against many, many other progressive priorities and interests worth even more than taxes on a $52B trust fund.

That's fine, but allegations in a law suit aren't prima facie evidence of anything. Especially when the text of that law suit is filled with political invective (calling protesting students and faculty "uncontrolled antisemitic mobs" and so forth).

There's a very easy determination to be made here about which students are or are not being victimized. If my knowledge of current events is still accurate, not a single pro-Israel student has been extra-judicially kidnapped and imprisoned. Pro-Israel Jewish students very well may feel victimized or scared. But put into perspective, I can imagine that pro-Palestinian students feel it much more so.

In an American context, this part also struck me:

> We did that. But only the Prime Minister could actually cut through all the bureaucracy and say, Ignore these EU rules on Blah. Ignore treasury guidance on Blah. Ignore this. Ignore that. “I am personally saying do this and I will accept full legal responsibility for everything.”

By taking over responsibility, Johnson loosened the accountability of the civil servants and allowed them to actually solve the problem instead of being stuck following the rigid formal process.

Of course this also can have pretty severe negative consequences. In the U.S., thanks to a recent Supreme Court ruling, the president has immunity from criminal prosecution under certain (yet to be fully determined) circumstances. If the president then "takes over the responsibility" for obviously illegal actions, and is immune from prosecution for those actions, you now have a civil service unburdened by any responsibility to follow the law. And there are some 3 million odd workers in the U.S. federal government.

That the conservatives on the Supreme Court did not consider this danger, especially in light of who occupies the office, is still astounding to me.

You linked an anonymous Reddit post and some European blog post with dubious credibility. Just because someone states as a matter of fact that these people are "pro-Hamas" doesn't make it a matter of fact.

The same individual making these legal threats has appeared on Russian state media over 100 times.

He allegedly supports Hamas? Is that in the article somewhere? It seems to me people are simply making things up to justify what seems pretty unjustifiable.

I don't think you have the facts of the matter or the context right here. The government absolutely does not have the right to expel legal permanent residents without due process. And 'keeping their mouths shut' and pretending otherwise is not even remotely, feasibly "reasonable". The Supreme Court literally just ruled that residents with much less legal status than this person must be given due process -- an entirely obvious statement. But one which apparently needs to be reaffirmed by the highest court in the land because of how casually this administration seems to discard basic tenets of U.S. law. So you're just outright wrong.

The government absolutely did prevent this person from 'saying whatever he wants' -- he was essentially abducted from New Jersey and sent to detention in Louisiana without the ability to contact his family or his lawyer, or get any sort of hearing. It's only after the fact, due to the work of his lawyers and the public outcry, that he was able to submit a habeas petition. You seem to be downplaying how violent and seemingly corrupt the government's actions were in this case (I guess because you ultimately disagree with whatever you imagine this person did). Just because one Louisiana DoJ immigration judge sided with their own branch of government doesn't really lend plausibility to the government's case.

"Blue states" are not airgapped enclaves, completely distinct and separate from other states. And there's a lot of diversity just within those states. The approach you're suggesting is not remotely viable. People who vaccinate can still contract illnesses and die from those illnesses. Many people can't vaccinate at all. Herd immunity is an absolutely essential part of the public health benefit of vaccination.

Your first link goes to a 2023 arXiv pre-print that never landed in any journals as far as I can tell (could be wrong though). And there seems to be some controversy about whether Kerr's math shows what he says it shows.

This is the danger of trying to sensationalize science and putting any special weight on science influencers, especially ones who very often seem gung-ho about any story that challenges the status quo despite the evidence.

This whole situation really goes to show that both the judiciary and the legislature need to be greatly expanded -- probably by 10 fold or more. Even if you greatly reduce the size of the federal government.

The executive employs approximately 3,000,000 employees. The federal judiciary only employs about 30,000 total, and the legislature about 20,000 total. The sheer velocity with which the executive can ram through questionable directives *and have them executed upon* (despite the law) means the other "co-equal" branches of government are always potentially on the back foot. It's just a personnel game. Trump has only highlighted how absurdly easy it is to abuse this imbalance.

And after the Chevron and Trump decisions, it's only going to get worse and worse. I do think these Federalist Society types who pushed these unitary executive theory ideas have now created a monster. They've created a situation where the executive has immunity to simply apply the law however it wants, clogging up the judiciary with civil and federal suits, and where the Congress cannot pass laws fast enough or with great enough specificity to avoid defying Chevron or avoid executive misapplication. Meanwhile the executive has long since moved on from the original issue to the next 10 issues, and the next 10, and the next 10, while the courts and the Congress are still only getting started on the first few problems. And the executive will never really get "punished" for these actions because of its supposed immunity.

And presidential elections don't really help with this problem. Because one president has 4 years to drastically reshape everything, and the next president will spend all of their 4 years reverting it, dropping all the previous suits, creating its own litany of new suits, and rinse and repeat. The hysteresis of this process is too long and leads to instability and chaos.

Only tangential to this post, but if you need a way to share secrets with your teams (or whoever), Hashicorp Vault is pretty decent. They don't even need login access. Instead of sharing secret values directly, you wrap the secret which generates a one-time unwrapping token with a TTL. You share this unwrapping token over a traditional communication method like Slack or e-mail or whatever, and the recipient uses Vault's unwrap tool to retrieve the secret. Since the unwrapping token is one time use, you can easily detect if someone intercepted the token and unwrapped the secret (by virtue of the unwrapping token no longer working). This hint tells you the secret was compromised and needs to be rotated (you just need to follow-up with the person to confirm they were able to retrieve the secret). And since you can set a TTL, you can place an expectation on the recipient too -- for example, that you expect them to retrieve the secret within 10 minutes or else the unwrapping token expires.

All of this has the added benefit that you're not sharing ciphertext over open channels (which could be intercepted and stored for future decryption by adversaries).