Yeah but they do actually compute things the way humans did and do. Submarines dont swim the way humans do, and they arent called swimmers, and LLMs srnet intelligent the way humans are but they are marketed as artificial intelligence
When you say "structurally incapable", what do you mean? They can certainly output the words (e.g. https://github.com/Smaug123/WoofWare.Zoomies/pull/165#issuec... ); do you mean something more along the lines of "testing can prove the presence of bugs, but not their absence", like "you can't know whether you've got a false or true negative"? Because that's also a problem with human reviewers.
The solution to this is to get agents to output an importance scale (0-1) then just filter anything below whatever threshold you prefer before creating comments/change reqs.
Note that the court case that first invoked “emanations from penumbras” involved a Connecticut law banning the the use of contraceptives. Do you believe such a restriction should be constitutional?
If we’re talking about what “should be constitutional,” we’re no longer talking about “the law” but instead policy or philosophy.
Regulating the “public health, welfare, and morals” is the prerogative of state legislatures. So the question is whether there is anything in the constitution that overrides that general power. Resort to “emanations from penumbras” is a concession that there isn’t.
By the way, this isn’t even some U.S.-centric take. The constitutional law in most western democracies leaves regulation of drugs to the discretion of the legislature.
I'm asking about your personal opinion: in Griswold v. Connecticut should the Supreme Court have upheld states right to ban access to all contraceptives, including condoms?
Griswold found a person has a right to decide, together with their doctor, what course of medical treatment is best for their particular needs.
Do you support that right at all beyond contraception (and abortion)? For example, do you support my doctor's right to prescribe to me all the pain medication that he and I think is appropriate for my painful, terminal disease? Or to prescribe me LSD or other hallucinogens to treat my PTSD?
It seems to me that Griswold should be invalidating most of the role of the FDA, except in an advisory capacity. But I don't think that's what most people who were aghast at overturning Roe believe.
Would you have been willing to allow, say, Texas or Tennessee, to decide that their resident doctors could tell their patients that there was no need for taking the COVID-19 vaccine, or to wear masks, or social-distance; and there could be no repercussions against patients for exercising those rights?
Support for the arguments Roe was based on seems to be highly dependent on where you want to apply those arguments. That doesn't seem very intellectually honest.
Your question asks whether I “believe such a restriction should be constitutional.” I’m obviously not in a place to decide what “should” or “should not” be in the constitution—the document is what it is.
If your question is whether I think the Griswold was correctly decided, the answer is obviously not. Regulating access to medications obviously falls within the police power of state governments, and I don’t think the constitution has a special carve out for particular types of medications. In fact I think this was an extraordinarily easy case as a legal matter, and the fact that the Supreme Court got it wrong demonstrates how intellectually sloppy a lot of mid-20th century precedent was.
> Regulating the “public health, welfare, and morals” is the prerogative of state legislatures. So the question is whether there is anything in the constitution that overrides that general power. Resort to “emanations from penumbras” is a concession that there isn’t.
Your comments these past months seem to evince an extreme textualist view of the Constitution's limits on government power.
Would it be fair to say that, in your view, a state government can take whatever action it wants, so long as the action isn't literally prohibited by the text of the Constitution? (In other words, for state governments, anything not prohibited is allowed?)
I'm curious: What's your view of the doctrine that under the 14th Amendment, state governments must comply with the Bill of Rights to the same extent as the federal government?
You periodically mock "emanations from penumbras" — is it your view that the phrase has more significance than simply a figure of speech to communicate the underlying concepts? (I have my own issues about Justice Douglas's track record, but that's not one of them.)
> Your comments these past months seem to evince an extreme textualist view of the Constitution's limits on government power.
I have an extreme textualist view of constitutional limits on legislative power.
> Would it be fair to say that, in your view, a state government can take whatever action it wants, so long as the action isn't literally prohibited by the text of the Constitution?
I wouldn’t say “literally.” I would say it’s a fair reading, as one would apply to a commercial contract.
>,(In other words, for state governments, anything not prohibited is allowed?)
Yes, state governments are not ones of enumerated powers. They inherited the plenary legislative power of the British parliament and can do whatever they want that’s not expressly prohibited.
> I'm curious: What's your view of the doctrine that under the 14th Amendment, state governments must comply with the Bill of Rights to the same extent as the federal government?
I think it’s unsound. My secret left wing view is I think New York can regulate guns (but also Utah can make a state church).
> You periodically mock "emanations from penumbras" — is it your view that the phrase has more significance than simply a figure of speech to communicate the underlying concepts?
It’s a figure of speech that is especially revealing of the “underlying concept,” which is discovering constitutional principles in moral or political philosophy. I basically don’t think that’s legitimate. The constitution embodies a particular set of moral and political philosophies. Sometimes we add more (e.g. the concept of equal protection). But judges don’t get to be philosophers weighing in on moral issues the framers of the constitution and amendments clearly didn’t contemplate.
> The constitution embodies a particular set of moral and political philosophies. Sometimes we add more (e.g. the concept of equal protection). But judges don’t get to be philosophers weighing in on moral issues the framers of the constitution and amendments clearly didn’t contemplate.
I'd agree with you — if we had more than one, largely-sclerotic way of amending the Constitution to accommodate new realities and new insights. "The only constant is change" is apparently a misattribution to Heraclitus [0], but it obviously has struck a chord to be in such widespread circulation.
Judicial lawmaking is a kluge, certainly. It brings to mind a passage in Tracy Kidder's 1981 The Soul of a New Machine: A "kluge" is like a wheel made out of bricks: No engineer would be particularly proud to have designed it — but if you need a wheel, and bricks are all you have to work with, then that's what you use.
In any case: Judicial lawmaking has become more-or-less accepted. You're part of the "or less" crowd, clearly. But ISTM that judicial lawmaking is a case of the (mythical) perfect being the enemy of the good-enough.
(A possible improvement: Congress could cabin the judicial-lawmaking process by stating — under the Exceptions and Regulations Clause in Article VI — that a 3/5 majority in each chamber, and perhaps presidential acquiescence, can override any SCOTUS pronouncement about constitutionality. And maybe impose a fast track for legislation to overturn a judicial interpretation of a statute. Both would be roughly akin to the Congressional Review Act for overturning administrative-agency decisions [1].)
> state governments are not ones of enumerated powers. They inherited the plenary legislative power of the British parliament and can do whatever they want that’s not expressly prohibited. *** My secret left wing view is I think New York can regulate guns (but also Utah can make a state church).
That's certainly not an incoherent philosophical position.
EDIT: In this view, it sounds as though each U.S. state would be more like one of the member countries of the EU than, say, cantons in Switzerland or counties in England.
That's confusing the statement "I want the option to overrule it" with "I want it disabled". I don't, there's very good reason for them to always be on, but there has to be a way to tell it what I want my browser to do, rather than what a server owner, who spent all of zero minutes accepting the default helmet options, says what my browser should do. In the same way that a browser should let me say which off-site domains, specific scripts, and on-page elements should be blocked.
If it were that simple, Vanguard wouldn't have taken the additional step to prevent trading in these products on their brokerage. Telling customers "you can't buy Bitcoin because we can't sell it to you at a profit" would be ridiculous.
Singapore has some of the strictest gun control laws in the world regarding ownership and possession. If you want America to be more like Singapore, start by repealing the second amendment, not increasing policing.
I work in K-12 EdTech and this hurts to read. Every sprint spent working on a "check the box feature" that administrators want to see is time not spent improving the effectiveness and usability of the product for teachers and students. Worse, sometimes the features or changes administrators think they want actively work against the interests of the users.
Edit: And god forbid you end up working on a core curriculum product that needs to be reviewed by each state's board of education. In that case you end up building a product that appeals to state-level bureaucrats, just to get get the privilege of selling to administrators, with little thought left for the actual students.
I will never understand why more developers don't just learn effective SQL. In many cases, the database is both the most expensive and most powerful component of the system - may as well put it to use!
The N+1 query used as an example in the post could have just been a single query.
ORMs, like SQL, C++, Python, and Ruby are just a tool. You can be shit at using that tool.
ORMs also usually allow you to drop down to SQL when you have to. They are there to help you, and they expect you to know what you're doing. Like other tools that make so much easy, they do make it easy to shoot yourself in the foot. I won't argue that the average ORM-user is more likely to create shitty-performing code than the average SQL-only-user. However, this speaks to the average population of the tools, not to the tools themselves.
When building a backend, I mostly work with Python and Django, and when interviewing candidates I place heavy value on them knowing what select_related and prefetch_related are. These are ways to void N+1 (akin to includes, I gather).
You can also write shitty queries, subqueries and just terrible stuff overall in SQL. And create shitty data models. Just the other day I had to comb through a PHP codebase that used raw SQL. Do you want to know what it had? N+1 problems.
If you use the ORM with the knowledge that you, as a competent developer, should have, then you most often will get a lot of bang for your buck. You will easily avoid the N+1 problem (and others), while benefitting immensely from what the ORM gives you, such as migrations, (somewhat of a) database independence, easy index creation and manipulation, much faster development times and iteration, an (arguably) much better syntax for creating and documenting the data model, and others.
I will never understand why so many people constantly criticise ORMs while turning a blind-eye to the fact that _most_ of their criticisms are actually criticisms of junior/amateur developers doing junior/amateur developer things. (I don't think you are one of these people)
The right tool for the right job. Sticking just to the ORM and using it blindly is a terrible idea. And I'm sure that extremely large codebases can be made all out of SQL — I'm not bashing it. I'm merely actively defending ORMs (and other tools) because they have their use and they do have very clear advantages — if used correctly.
I've been using sqlc, sqlx, pgtyped, & kysely and I found they have ORM-like productivity with full type-safety but they don't bring the baggage of leaky abstractions.
Unfortunately I don't know if there are Ruby or Python equivalents yet.
There is good value in having ORM for all those boring 90% data fetch tasks. It removes a whole lot of boilerplate and you can guarantee that your code representation equals database representation.
The actual problem here lies in the Active Record (anti) Pattern. It becomes impossible to distinguish in-memory access from db access. I’m not surprised to find this being particularly prevalent in dynamically typed scripting languages…
> The actual problem here lies in the Active Record (anti) Pattern. It becomes impossible to distinguish in-memory access from db access
I'm not a big fan of Active Record either, but conflating the two is pretty much the point. I think the biggest antipatterns around AR are in trying to treat it as a transparent abstraction rather than always being aware that it's more or less an "immediate mode" DB interface. Yes, you can treat your model objects as plain old domain objects here and there, but they don't stop being intimately tied to the database at all times, and you need proper DTOs and/or extra interfaces if such separation ever becomes important. But many apps don't have to care, and those are the ones where AR isn't such an albatross.
My main problem with AR is crappy implementations like Eloquent. No identity cache, so many methods and props on the Model base God Class that prevent using a column with that name, magic methods everywhere... Still, it serves all right as long as you treat it as a DB interface and not your canonical business object model.
> But many apps don't have to care, and those are the ones where AR isn't such an albatross.
I am super defensive on this one. Apps can keep not caring about it... until they have to.
Is it that much more effort to have a clear boundary between domain and db-access? I think it is only effort if you don't know what you are doing... which is _precisely_ when you need it the most.
> Still, it serves all right as long as you treat it as a DB interface and not your canonical business object model.
I still don't get the benefits. If you decouple db-model from domain model you lose all "advantages" of AR. It's like having plain DTOs but with its sharp spikes sticking out if you ever do something wrong.
You're preaching to the choir here: I much prefer datamapper ORMs when I even want an ORM at all (I'm not quite the OOP astronaut I once was). I'm saying AR's poor design isn't a noticeable drag in small scopes, especially for short-lived models whose only purpose is to populate a view. It's like having really fancy DTOs, and keeping them in the DTO bin is what keeps you from getting stuck with all the sharp spikes. Life can be comfy in a minefield when you have a map and don't wander too much :)
Rails can optionally use SQL migrations, UUIDs for pks, enforce referential integrity with additional options, and create additional indexes easily. The value of ORMs comes in separating domain data modeling from the peculiarities of a DBMS'es SQL flavor.
An ORM should be used to automate the commonplace rather than as a substitute for understanding how to manage and operate a DBMS.
SQL knowledge is often needed for cleanup and ETL work outside of the monolith. There is no substitute for understanding how things deeper in the stack operate.
I’m increasingly cold on ORMs existing to abstract away the specifics of your DBMS. In my entire career I think I’ve seen one migration between different databases (MySQL to Postgres), and it was an undertaking despite using an ORM.
Where I do appreciate ORMs is in being able to bridge between the relational world of SQL and the object oriented or functional world of the application. Occasionally I’ll decide an application is stupidly simple and it’s not worth the overhead of an ORM, I’ll just write SQL directly. I regret it every single time, because inevitably I end up implementing a half-baked ORM to make the results usable in the application layer.
And yes, as you say, regardless of an ORM you need to understand the database underneath and the SQL being generated. I’ve seen so many developers who are completely baffled by why their code is running slowly when a quick EXPLAIN will surface the fact they’re doing a sequential scan over several million rows of data.
> I’m increasingly cold on ORMs existing to abstract away the specifics of your DBMS. In my entire career I think I’ve seen one migration between different databases (MySQL to Postgres), and it was an undertaking despite using an ORM.
I've always viewed it more as "we can start fresh with another database with ease" and not "we can migrate from a database to another quickly".
And when I say "start fresh", I don't literally mean starting fresh. It can just mean I write my packages (e.g. Django's "Applications") in a database-independent way and whoever ends up using it can use it with whatever DB they have.
Perhaps your vision is indeed what is often sold, though I never saw it that way.
While this feels like common knowledge, I’m willing to bet you’ve learned extensively why you should profile everything when it comes to performance.
I remember my engineering manager was freaking out over me using strings instead of symbols because they’re less performant, meanwhile he didn’t know to add indexes to fks lmao.
He clearly cares about performance! Just never came to understand relational databases, which was my day 1 obsession. And doesn’t know how to profile the entire thing to make sure he’s stressing over the things with the largest effect size.
Anyway he got fired and works as a systems guy now. Happy for him lol.
