ShamblingMound

4 месяца назадdiscuss

Secret Agent has a slow, difficult beginning (~hour). Not much happens. And it's not clear why what's happening is happening, particularly for someone unfamiliar with Brazil's political climate in the 1970s.

As someone who's never been to Brazil, certainly not in the 1970s, watching Secret Agent still felt like being transported there. How did they make a movie that makes you feel like you're in a familiar place you've never been to?

And then after about an hour, it picks up a bit more, and by the end, it felt like they directly transmitted to the audience the horror of the Brazilian junta in all kinds of subtle and dramatic ways. We don't see the resolution of the main character's story because that moment is lost. Memories of his life are fractured (through disjointed audio recordings) or repressed (by those closest to him).

Hard to put it into words. I started out disliking it and ended up loving it.

ShamblingMound

12 месяцев назадdiscuss

Have been looking for an orchestrator for AI workflows including agentic workflows and this seemed to be the most promising (open source, free, can self-host, and supports dynamic workflows).

But have not seen anyone talk about it in that context. What do people use for AI workflow orchestration (aside from langchain)?

ShamblingMound

в прошлом годуdiscuss

Had a handful of interactions with Matt. Probably the last was sometime in the mid- to late-2010s. I essentially started my software development career using software that he was largely responsible for (unknown to me at the time):

  * Catalyst (web framework)
  * DBIx::Class (ORM)
  * Moo (object-oriented code)

These are, in my opinion, some of the best packages in Perl, and there are rarely counterparts that are as good in other languages. Have not found an ORM that is as effortless and featureful as DBIx::Class, for example.

I've read about his tough interactions with other people, and it does seem that, at times, he fell into the classic trap of loving his own ideas too much; but in our interactions (reviewing some of my code on a Perl project), he was really helpful and kind. Also amazingly quick. He read my beginner-level Perl code, instantly understood it, and instantly gave clear, concise feedback.

It's a shame he has passed.

ShamblingMound

5 лет назадdiscuss

For anyone curious about whether recycled plastics in clothing have less environmental impacts than primary raw materials, they do [0].

[0]: https://www.sciencedirect.com/science/article/abs/pii/S09213...

ShamblingMound

5 лет назадdiscuss

Thanks, that is a great feature.

ShamblingMound

5 лет назадdiscuss

If anyone knows of a flexible static site generator that can process Rmarkdown files, I'd be interested to know about it.

There is `rmarkdown::site` (see [0]), but I had to mess with the internals to get it to do what I wanted.

[0]: https://rmarkdown.rstudio.com/lesson-13.html

ShamblingMound

5 лет назадdiscuss

The author asks:

> So why don’t firms merge to get larger?

But they do. Most industries/sectors trend toward consolidation. Specific examples that come to mind are radio, agriculture, banking, and tech. I did a quick search and found [0], which describes a general tendency toward consolidation as industries mature.

One major reason for consolidation (ignored by the author) is economies of scale: as volume increases, the cost per unit decreases.

On the other side, the author ignores benefits of central control for some things. For example, private railroad companies used to have different gauges for their railroads in order to protect their routes, but that led to unnecessary inefficiencies.

Also, governments have a different mandate than the private sector. Governments are charged with pursuing citizens' values (freedom, equality, security, etc.), but for private sector companies, their goals boil down down maximizing profit.

[0]: https://hbr.org/2002/12/the-consolidation-curve

ShamblingMound

5 лет назадdiscuss

The university's IRB approved the research for the first paper as exempt. There is organization-level culpability here. It is reasonable for Linux kernel maintainers to block an organization acting in bad faith.

ShamblingMound

5 лет назадdiscuss

It certainly is considered harm. Ethical research with human subjects makes every effort to do no harm.

Taking someone's time without compensating them can be harmful. That's why researchers often give gift cards or small amounts of money for filling out surveys.

It's not always as straightforward as paying participants, but compensating participants for their time should be a consideration in ethical research with human subjects.

I don't know how much time the kernel maintainers spent on these patches or what their time is worth, but I'm certain that the time they spent on this is worth way more than the nothing they got in return.

The uncompensated time that maintainers spent on this is harm. And that anger that you imagine you would feel is harm. Ethical research with human subjects tries to avoid causing this kind of harm, and there seems to have been no such effort in this research design. This was not ethical research.

ShamblingMound

5 лет назадdiscuss

Sorry, but you're wrong. This experiment involved deceiving human research participants.

Actual humans spent their time and effort reviewing and working on the patches that were submitted.

ShamblingMound

5 лет назадdiscuss

Oh my gosh, their IRB approved this as exempt knowing all the details? Initially, my guess was that the researchers misinformed the IRB about what the experiment involved.

But the IRB approved as exempt an experiment involving (1) humans (2) who are deceived (3) and who do not provide informed consent.

Additionally, an IRB review should take into consideration all possible harms to people that could result from an experiment, and as a result of this experiment, many downstream Linux users were affected who also did not have an opportunity to provide informed consent.

I would think this would be a hard experiment to get approval for at all, let alone exempt. That's bananas. For me, this taints the reputation of the university.

Edit: The more I think about this, the worse it gets. One of the pragmatic aspects of not harming participants is compensating them for their time. Through the experiment, researchers get data, and participants get something in a trade that's not always exactly fair, but at least treats the research participants with respect. That's why it's common to receive gift cards for filling out surveys. Research participants in this experiment (skilled software developers) were presumably not compensated for their time during the experiment and their time is probably worth quite a bit. How in the world did this pass IRB as exempt.

ShamblingMound

5 лет назадdiscuss

This was (1) research with human subjects (2) where the human subjects were deceived, and (3) there was no informed consent!

If the IRB approved this as exempt and they had an accurate understanding of the experiment, it makes me question the IRB itself. Whether the researchers were dishonest with the IRB or the IRB approved this as exempt, it's outrageous.

comments