lmkg

Godot's recent announcement spelled something out clearly: when a mid-tier rando contributes, you can provide feedback to that person and possibly help them grow into being a senior contributor or even a maintainer. That possibility of helping the human behind the code is part of the motivation for doing open-source. Mentoring shitty devs is itself giving back to the community, in a different form than the code itself is. And that is qualitatively different than giving feedback to an LLM.

I mean that's still basically what they tried to do at the time. They were trying to get them through web standards committees and everything.

IIRC a big reason it didn't end up working was because NaCl was such a "big" technology and asm.js such a "small" one that asm.js was able to reach production-ready first despite starting work several years later.

This is why we're clear-cutting forests to build new data centers? Not even for "real" productivity gains, but just for the sake of using the tokens.

The difference is that early-bird pricing is transparent and predictable. There is a written, known policy of $X discount during specific hours. You can plan for it. It's never a surprise.

Dynamic pricing means sometimes you go there, and Wendy's decides on the fly whether you get a lower price and how much. It gives Wendy's the option to pinch pennies how they see fit for their own benefit, rather than offering a deal which you can choose to accept.

> which seems to directly oppose the CCPA.

I have some background in data privacy compliance.

It sounds like they are claiming to be a Service Provider under CCPA, which is similar to a Processor under GDPR. Long story short, a Controller is the one legally responsible for ensuring the rights of the data subject, and a service provider/processor is a "dumb pipe" for a Controller that does what they're told. So IF they are actually a Service Provider, they're correct that the legal responsibility for CCPA belongs to their customers and not them.

That's a big IF, though.

Being a Processor/Service Providor means trade-offs. The data you collect isn't yours, you're not allowed to benefit from it. If Flock aggregates data from one customer and sells that aggregate to a different customer, they're no longer just a service provider. They're using data for their own purposes, and cannot claim to be "just" a service provider.

My personal conspiracy theory is that in some cases, the "loser" is in on the graft to. It's a way to launder bribes.

That still begs the question though: There are existing tools and solutions that do this. Why not, and would this being AI make a difference?

"My boss would be more likely to approve it" is a cynical but valid answer.

That type of experimental set-up is forbidden due to ethical concerns. It goes against medical ethics to give patients treatment that you think might be worse.

The Gaussian integers usually aren't considered interesting enough to have disagreements about. They're in a weird spot because the integer restriction is almost contradictory with considering complex numbers: complex numbers are usually considered as how to express solutions to more types of polynomials, which is the opposite direction of excluding fractions from consideration. They're things that can solve (a restricted subset of) square-roots but not division.

This is really a disagreement about how to construct the complex numbers from more-fundamental objects. And the question is whether those constructions are equivalent. The author argues that two of those constructions are equivalent to each other, but others are not. A big crux of the issue, which is approachable to non-mathematicians, is whether it i and -i are fundamentally different, because arithmetically you can swap i with -i in all your equations and get the same result.

They weren't "just" raw dumps of internal C structures. It takes careful design work to dump raw memory in a usable fashion. Consider: You can't just write a pointer to disk and then read it back next week.

Binary MS Office format is a phenomenal piece of engineering to achieve a goal that's no longer relevant: fast save/load on late-80's hard drives. Other programs took minutes to save a spreadsheet, Excel took seconds. It did this by making sure it's in-memory data structures for a document could be dumped straight to disk without transformation.

But yes, this approach carries a shitton of baggage. And that achievement is no longer relevant in a world where consumer hardware can parse XML documents on the fly.

I have heard it argued, though, that the "baggage" isn't the file format. It's actually the full historical featureset of Excel. Being backwards-compatible means being able to faithfully represent the features of old Excel, and the essential complexity of that far outweighs the incidental complexity of how those features were encoded.

HIPAA is not a privacy law, nor even a healthcare law. It's an insurance law. It does not cover medical records generally. It deals strictly with how doctors bill insurance companies, and mandates security for health information being billed about.

For the same reason, health & wellness apps are not generally covered by HIPAA, and in fact quite a few of those exist solely for the purpose of selling medical data to data brokers. Especially ones related to women's health.

> If it only exists in some hardware, how should the standard deal with that?

Generally seems to me the C standard makes things like that UB. Signed integer overflow, for example. Implemented as wrapping two's-complement on modern architectures, defined as such in many modern languages, but UB in C due to ongoing support for niche architectures.

The issues around pointer provenance are inherent to the C abstract machine. It's a much more immediate show-stopper on architectures that don't have a flat address space, and the C abstract machine doesn't assume a flat address space because it supports architecture where that's not true. My understanding is that reflects some oddball historical architectures that aren't relevant anymore, nowadays that includes CHERI.

It very much is something that exists in hardware. One of the major reasons why people finally discovered the provenance UB lurking in the standard is because of the CHERI architecture.

GDPR does not currently have explicit business size thresholds. Its provisions are all framed as personal rights of the data subject, so its provisions are always in effect. By contrast, CCPA in California is framed as a consumer protection law so it only applies to companies of a certain size.

In practice, small fries are not an enforcement priority. Regulators in most countries are not well-funded so they have to be frugal with their enforcement actions.

The EU is currently reviewing an option to relax GDPR requirements for smaller businesses. Not remove GDPR requirements, just streamline some of the process overhead.

Yes it does.

If a user can say "here's my IP address, what data do you have on me?" and you can answer that question, then that's personal data under GDPR. It's pseudynomized, but not anonymized, and pseudynomous data is personal data.

We passed the tipping point where bot traffic outnumbered human traffic fifteen years ago. LLMs are an order of magnitude worse by most first-hand accounts, but it's just a continuation of a very long trend.

> You also have to remember that individual countries fining on global revenue runs the risk of fines "duplicating" each other for the same or similar behavior, again bankrupting a corporation when the goal should be to change behavior.

This is explicitly not a concern under GDPR. The "one-stop shop" mechanism means that all issues across the EU get funneled to the lead supervisory authority, which is always Ireland because that's where EU subsidiaries are headquarters for tax purposes.

Per a reddit comment, modules are allowed to have circular dependencies while crates are not.

That website is not authoritative and should not be given more credence than your median tech blog.

GP are wrong about which law applies, but they are applying that law correctly.

The ePrivacy Directive requires consent to read or write from the user's terminal device, except when strictly required for the functionality the user requested. Unlike GDPR, it does not allow a different Legal Basis. It must be consent, or strictly functionally necessary. Nothing else.

The passage of GDPR did impact the ePrivacy Directive in that it updated the definition of "consent." The ePD doesn't have one; it referenced the definition in the DPD, which was replaced by GDPR. This is why people blame the GDPR for cookie banners, although really it's incidental.

It depends on the jurisdiction and law, but a "data breach" is when data is accessed by a party who is not authorized, or who should not be authorized. It's not just hackers. Sending data to the wrong recipient is a form of data breach. Under some definitions, sending data to the intended recipient without appropriate safeguard is a form of data breach.

In this case, health care data covered by HIPAA was sent to a party without a legal contract that extends HIPAA to the receiving party. By law, that's a data breach.

Under some legal definitions, "data breach" includes not just breakdowns of confidentiality, but also of availability and/or integrity. So a company deleting your data by accident would be considered a data breach, even though it's being accessed by fewer parties than intended. This can be important: imagine a bank or credit agency losing some or all of the data about you, this would materially impact your ability to do business in the modern world.

I have some sympathy for people who are blindsided by surprising difference between a new tool and their old one.

This post is not eliciting sympathy. They're data consultants, who don't understand a very basic and fundamental aspect of the tool that they're using and recommending. If you're a consultant you have a responsibility to RTFM, and the docs are clear that LIMIT doesn't prune queries in BigQuery. And, also, the interface tells you explicitly how much data you're gonna query before you run it.

This post is also blaming Google rather than accepting their own part in this folly, and event admits the docs are clear on this matter. Cost-control in BigQuery is not difficult. One of the tradeoffs of BQ's design is that it must be configred explicitly, there's no "smart" or "automatic" query pruning, but that also makes it easier to guarantee and predict cost over time as new data arrives.

At least you're using the actual months. I've recently had to deal with 4-5-4 Retail Calendars. Explaining to my coworkers why the data for September includes dates from both August and October, and that's Working As Intended.

GEOMETRIC ASPECTS OF ISOPERIMETRIC AND SOBOLEV-TYPE INEQUALITIES

Suspected to be flagged for the trigger word "inequality."

PageRank relies on pages from one domain linking to pages on another domain as the main signal of quality. But no one does that anymore, because of PageRank. Most cross-domain links are from "partnerships" made under the explicit premise of boosting PageRank score. The metric became the goal, invalidating the metric.

Another factor is that since 2006, social media has displaced blogs. Blogs were a rich source of "authentic" cross-domain links, and the movement of online discussions to consolidated and closed platforms has dried up that well. Some even go so far as to pin the downturn of blogging on the demise of Google Reader.

Check out the CLP(ℤ) library by Markus Triska, who is also the author of Scryer Prolog. It defines new comparison predicates that lets you use bidirectional logical programming on standard integers with standard math operations.

https://github.com/triska/clpz

> - The reuters article says "Issue was disclosed voluntarily to the regulator." but the actual source [1] announces a breach of GDPR Article 33(1), for failing to notify.

GDPR mandates the regulator be informed within 72 hours of the breach being discovered.

The official link you provided confirms that Meta informed the regulator voluntarily, in March of 2019. That page also includes a link to Facebook's press release, which says they discovered the issue in January. That's a time lag of around two months, which is around two months longer than the law permits. So yes, failure to follow the law mandating notification.

All things considered this is a small fine. Three orders of magnitude smaller than their usual GDPR fines.

> If you have accrued vacation, you "earned" your vacation and they can't take it away from you without compensation... If they fire you, they should even have to pay you your salary for those vacation days.

This is the law in California but not most other states.

In most states, vacation time is a benefit that can expire, evaporate, or be rescinded. The company is just deciding not to assign you job duties during your time off. How generous of them! But it's totally their right to decide not to do that, or to put a bunch of boxes around how they will do that. So it's legal, and common, that unused vacation time will actually be lost.

In CA, accrued vacation time is classified as "deferred compensation." It is legally money that is owed to you. It will be paid at a later date, but it must be paid.

It strikes me as no coincidence that the "unlimited vacation" fad started in tech companies with a large chunk of their workforce in California.

Regarding point 1: The timeline in the linked article describes some communications as "pressure emails." I've heard the theory, but haven't seen solid evidence, that the pressure emails weren't just regular impatience from outside devs, but actually part of the attack. To convince the primary maintainer into granting access to another party.

> Unless they had a better reason for restricting the maximum city size.

They did! EA wanted you to play the game online. They encouraged you to connect your city with other players by making it difficult for cities to be self-sufficient.

I.e., user-hostile design.