tpoacher

> Now there is a third place: code generation time. The model can often avoid the mistake before the compiler ever sees the code.

This is doing a lot of heavy lifting here. How do you even know a mistake has been "avoided", and safely, for that matter? The kinds of mistakes the AI will miss are the worst kind: subtle logical mistakes, buried inconspicuously within boilerplate code, hard to detect or reason about, until one day you wake up to find that your database has been replaced by a rickroll video.

Also, it's odd that they moved from Haskell to completely untyped python ... python may be no Haskell when it comes to type safety, but typing+mypy/mypyc goes a very long way.

I know, but there's no harm in confirming that is indeed what they had in mind :)

Most people commenting here seem to be focusing on memorization, which seems to miss the main point of the article. The big insight here is the chunking part, not the memorization bit.

I use anki in a similar way to the author and it has been truly transformative to my learning and daily habits.

And, like the author, anki has become part of my daily workflow even for things that it's not really specifically designed for; e.g. easy frictionless LaTeX notetaking, cross-referenceable database (via note ids), incremental learning repo, long-term reminders (gtd-style) etc.

If you're relying on government regulation to solve this, I guess the obvious bigger question would be "Quis custodiet ipsos custodes?

That's just it though isn't it. Good tools that are invisible to you won't easily come to mind because they tend to be, well, invisible.

It's not until you randomly end up on a system which doesn't have that tool that its usefulness becomes visible; and I mean really visible.

Reminds me of this quote:

"We notice the person who is for ever bowing and fussily servile, and perhaps say, How humble he is! But the truly humble person escapes notice: the world does not know him."

~ Tito Colliander

would you mind linking to the channel? not sure which one you mean

Do I need to be "specific"?

You made the general claim. I'm pointing out that the general claim when applied to a different minority population suddenly does not sound as reasonable, and could easily cause distress and discussions of persecution and discrimination.

Pivoting to wanting to exceptionally examine an exact opinion for this particular minority is simply Special Pleading. The point stands that, no, people should not be expected to shut up and "defer to reasonable people" just because their opinions and interests are in the minority.

I clearly didn't "post generated text" masquerading as human insight in the manner cautioned in that context.

I, the human, demonstrated that even an LLM, which in this context can be viewed as an impartial source, could easily find legit examples.

I could of course have simply stated that this is the case without pasting the LLM text, but in that case I'm fairly sure OP could have simply replied "Nah-uh".

Rear entrance then

Black people constitute 13% of the US. Should opinions exclusive to black communities be ignored and dismissed as unreasonable?

Is it bait? I'm pretty sure it's a reasonably factual, albeit general claim. Asking chatGPT for country-specific examples for instance gives this:

> Yes—some countries have (at various times, and in some cases still today) adopted policies aimed at making the population more “homogeneous,” through segregation, assimilation pressure, or exclusion/deportation. Concrete examples:

- South Africa (apartheid era, 1948–1990s): An official system of racial classification and enforced separation (“separate development”).

- Germany (Nazi period, 1933–1945): State ideology enforced a racial hierarchy and pursued forced removal and mass murder of those deemed “undesirable.”

- Israel (state policies affecting Palestinian citizens and occupied territory, especially since 1967): Includes laws and administrative practices that many observers describe as producing or enforcing unequal status by group; key issues include citizenship status differences and restrictions tied to national/ethnic identity.

- Myanmar (Rohingya): Policies and law enforcement that stripped/blocked citizenship for Rohingya and enabled persecution, culminating in mass violence and displacement.

- Canada (Indigenous assimilation policy, especially 19th–20th century into 1996): Forced assimilation via residential schools and bans on language/cultural practices; many have characterized this as cultural genocide.

- United States (Jim Crow + earlier immigration/citizenship rules; and internment): Historical legal regimes created segregation and restricted citizenship/naturalization based on race/national origin (e.g., earlier Asian-exclusion immigration restrictions).

You may disagree with some examples on this list, but I'm sure even you would consider that the first two are clear examples of diversity-fearing 'cultures' rather than 'bait'. And this is even before considering the wider definition of the word 'culture', which can be even more exclusionary.

> Google Management Has Lost Its Moral Compass

This needs a "2004" date.

My favourite entry was the "leaky redaction" one, which created a "redacted" version of an image in the text-based ppm image format, but in such a manner that most of the redacted information could be restored directly from the file.

The trick was to redact by converting characters to zeroes, as opposed to words. So "8" would become "0", but "256" would become "000" rather than "0". In a black and white text document, distinguishing between "0" and "000" effectively faithfully recreates the original document! But even in color pictures, it was very interesting to see how much information is still retained going from 256bits per channel down to just 3.

The method was underhanded in that the code at that section was written in a seemingly very sensible manner, enough to convince an astute auditor that this was an honest and defensible bug.

The main thing I feel this article gets wrong (or at least does not properly address) is that all the reasons provided seem subjective. Is there no objective (or at least semi-objective) process to do a cost-benefit analysis of a rewrite?

In ML e.g., there is the concept of a "ceiling analysis". I don't see why a variant of this wouldn't work here too.

I agree for the most part, though it's worth pointing out that HN specifically has a mitigating characteristic in this case, which is that repeat posts are not moderated away, and are in fact encouraged.

Case in point, one if today's top posts is on knoppix. Definitely not early adopter material! :)

I agree more generally though. While I understand the benefits of a 14day response window, it really does destroy the ability to find a thread that is useful in a more anachronistic manner.

> If anything, that they thought this was acceptable makes me wonder what else they're harvesting from my machine? PII?

Hah, you just reminded me of this meme I spotted the other day: https://img-9gag-fun.9cache.com/photo/an76Wnz_460swp.webp

Possibly, but that still doesn't mean there isn't a real benefit to developers developing for ReactOS as a proxy for developing for Windows in a manner that makes their products instantly compatible with Mac and Linux.

The "real work" part is mostly about the user. And that depends on available programs on the platform, not on the user's boundless love for telemetry.

OP seems to be mostly addressing developers here, so the point stands.

Then again, this goes completely against the "O" in the SOLID principles (and possibly also the "D" too).

Personally I prefer SOLID to YAGNI.

How about this analogy then:

"OP says Windows needs webdevs to make webpages, but then says Windows Lite internals shouldn't be tightly knit with Internet Explorer, one of the primary means of viewing webpages on windows."

There is already a "Windows Lite", and it's called ReactOS [0].

Except, the last thing we need is for Microsoft to own / embrace-extend-exterminate this. It's best that it's kept as an independent project, enabling developers to target the Windows API if they feel like it, while ensuring that they only target the Wine-compatible parts of it, which then enables their software to also run seamlessly on other platforms at no extra effort, and without requiring either the users or the devs to have to lock themselves into this ungodly telemetry-spying-ads-AI-.NET mess that the article talks about.

Microsoft, on the other hand, has zero to gain from making a Windows Lite variant themselves given their lifeline is exactly this telemetry-spying-ads-AI-.NET ecosystem; and they have zero to gain from absorbing ReactOS or similar projects, other than in order to eliminate them.

[0] https://reactos.org/

Most people here seem to thank that the only realistic alternative for submitting code / written reports in the age of AI is handwritten invigilated exams.

So just mentioning my own anecdata here; I've had reasonable success converting a "code + written report" assignment into a "code + structured video report", where the code is assessed "exclusively" through the video report, and where the student needs to demonstrate ownership and understanding of the code, explain and defend design decisions, etc.

Yes, in principle it is possible for the student to generate a "script" and narrate it during the video presentation, but this is fairly easy to detect, and in general, given enough flexibility / subjectivity in the marking scheme, you can always downmark such students on the basis that their presentation demonstrated lack of understanding and ownership of the codebase, without having to go into too much specifics.

Obviously one implication of the above is that you no longer care if their code was "manually" written vs AI "assisted"; only if they have a good grasp / ownership of it, and can defend the broader architecture and design decisions. But I feel that this is realistic in the age of AI, since this is largely what will be expected from their future employers too.

> Writing in block letters was seen as a sign of illiteracy

How interesting. I've always thought the cyrillic block letters look beautiful, whereas the cursive variant not so much. I learnt some russian and I've never bothered writing cursive because of this perception. Odd to realise now that the average russian seeing my writing would be dismissive instead of impressed!

Ah, really? I had no idea! Very surprising too, since they've made their reputation as an indie+abandonware promoter.

Though, to be fair, I haven't seen your game in the Dreamlist [0] or Community Wishlist [1] pages either. You could at least add it and try promote it a bit there to help GoG gauge interest?

[0] https://www.gog.com/dreamlist [1] https://www.gog.com/wishlist

GoG, pls.

Nice name, "Akrites".

Probably not as impressive to a non-Greek, but to a Greek person it creates very strong imagery.

While the "thankfully not all books are worth reading" adage is presumably very true of blogging as well, I do still feel some admiration for bloggers, even when their blogs could be perceived as "useless". Even to put something "useless" out there can be an act of bravery itself sometimes (in a "In a time of universal deceit, telling the truth is a revolutionary act" George Orwell sense).

I've felt the "chilling effect" way too often when considering if I really want to put an idea out there that might later turn back to haunt me (even for things I would not consider controversial). I am very aware of the irony that oddly enough, psychologically I don't perceive the same level of danger from random HN comment threads though.

I reluctantly confess that I have indeed on occasion had to write in a way that makes the reader have to do a couple of extra mental steps to follow the logic, to avoid reviewers rejecting the manuscript on the grounds of the theoretical contribution being "trivial".

Combine this with added fees for longer papers and you have your answer.

A plug for my favourite book here. Richard Webb's 80/20 Japanese covers these topics extremely well, and in a very logical / productive order for a foreign speaker.

(and it comes with very enjoyable audio tapes and anki cards to boot)

Highly recommended.

In some sense, you should still act on this, since if an external auditor relies on the same stack, it'll still cause you headaches.