dhc02

I couldn't think of a use case for something this. For anyone else like me, the readme outlines one near the bottom which seems genuinely nice:

Set this up, allow no-auth sending from any address on your LAN, and then spinning up new local services that need outgoing mail is easy, because you can just give them the IP of the relay and omit credentials.

I am constantly surprised how seldom aider is mentioned in threads like this. I understand that it's not directly integrated into the editor, but the "editor + parallel CLI tool chain" paradigm feels so natural to me because we drop to terminal for so many other parts of building software. If you haven't tried it (particularly the architect/editor modality), it's worth a couple of hours of experimenting.

For me, YouTube Premium (for family) is so worth it because:

- It removes ads for my kids, which are going to watch YouTube no matter what I do

- It includes YouTube Music, so I don't need a separate Spotify subscription

- It sends money to creators based on what I proportionally spend time watching, rather than based on how valuable of an audience I'm deemed to be when an ad slot is auctioned

- It is a signal that I'm willing to pay for content, and I want platforms and creators to continue offering that option

I like to imagine as if we threw something into the air so hard that it went up for 26 years, then came back down for 26 years.

The best part about Kagi is that if the default results don't seem helpful, one click restricts results to only discussions and forums, which is usually exactly what I want to do next.

As my income drops in slow periods, I pause subscriptions to many things. The very last two I would be willing to give up are (n-1): YouTube Premium and (n): Kagi.

That also is a factor in why I pay for more than I need. I want them to succeed.

Personally, I would still love a site-by-site "reject non-essential cookies" prompt from an extension that's in the same place, with the same UI, on every site. Still a click, but lots better than having to figure out how to accomplish it on each and every site.

Tax wealth above $100 million at ~90% upon death.

In other words, make all the money you want, get as rich as you want, but it goes back to the commons when you die and can't use it any more.

The top 50% of earners in my household pay 100% of the bills. How is this fair?!

Oh wow you owe it to yourself to try aider then. Architect mode especially, where you use one (usually reasoning) model to plan and one (usually Claude) to write code, is pretty awesome.

My approach has been to purchase books from bookshop.org (or directly from the author if that is an option), and then immediately go find a DRM-free backup copy to send to my e-ink device.

Definitely do not tell that friend to head to https://en.m.wikipedia.org/wiki/Z-Library and use one of the URLs listed there to go to that website and find and download a DRM-free backup copy of all the books they have bought and paid for in the past.

This is elegantly stated.

I'm going to give you a video to watch. It's not mine, and I don't know much about this particular youtuber, but it really transformed how I think about writing and structuring the prompts I use, which solved problems similar to what you're describing here.

https://youtu.be/y_ywOVQyafE?si=IvKjy7QUYgxGPNgD

PS (I have not bought the guy's course and have no idea whether it's any good)

Yes! Turn off autocommit, everyone! Review and test, then git commit.

Efficient and effective organizations work this way, too: a CEO to plan in broad strokes, employees to implement that vision in specific ways, and managers to make sure their results match expectations.

I may be the only one, but I was under the impression this whole time that all EV tax credits were for consumers who buy EVs, not companies who sell them.

Aider, with o1 or R1 as the architect and Claude 3.5 as the implementer, is so much better than anything you can accomplish with a single model. It's pretty amazing. Aider is at least one order of magnitude more effective for me than using the chat interface in Cursor. (I still use Cursor for quick edits and tab completions, to be clear).

I didn't realize there were some non-stick OSes that you could re-lock the bootloader for. Which ones?

I have been coming around to the idea that we should ban all* algorithmic content surfacing.

It's taken a while, but the longer we go down this path, the more clear it seems that it is impossible to design a content algorithm that does not have significant negative cultural side effects. This is not to say that content algorithms don't have benefits; they do. It's just that they can't be useful (i.e., designed to optimize for some profitable metric) without causing harm.

I think something like asbestos is a good metaphor: Extremely useful, but the long-term risks outweigh any possible gains.

Tailwind Intelligent Site Prototyping

FYI the comma is being placed awkwardly on its own line on the homepage.

https://www.dropbox.com/scl/fi/19zt9067x8j230810bjcp/Screens...

This really added something to my way of thinking about this. Thanks.

I like this and haven't used it before. Thanks for sharing.

I think that for the person who suggested the idea, this would be considered success--creative products that aren't being actively used should revert to the commons.

But as a comment in a different thread suggested, I do think that we'd want something like this to be accompanied by a shift in how we think about copyright.

Right now, copyright is used for two broad reasons: 1) preventing unauthorized "commercial" use, and 2) preventing piracy.

Use 1 is broadly good but flawed in the current system. But use 2 is culture-killing and creativity-surpressing with current copyright lengths, a good example of which being the loss of what.cd.

The more I think about it, the more I think we need to separate the two uses, so that creative works are in the public domain by default after a short time for the purposes of private use and archival projects, but corporations are prohibited from using creative works they haven't paid for for commercial purposes on a much longer timescale.

On balance, copyright seems to restrict the public and benefit corporations, and ideally that would be inverted.

Amazingly, already a thing.

See https://www.theverge.com/24255887/social-ai-bots-social-netw...

Someone above recommended devd, and it looks pretty nice. https://github.com/cortesi/devd

Agree very much with all of this.

The way I estimate the situation (and I admit this is not a rigorous scientific conjecture), is the following (for public schools in the US in the average):

Early years: bad pedagogy, bad retention rates (ie, quitting after 5th grade to go work on the farm, bad average results, basically only the top 10% learned deeply and went on to intellectual pursuits

1900s - 1980s: Decent and improving pedagogy (the aforementioned procedure-first style for math), good and increasing retention rates, good parental and societal pressures to perform, great average results, top 50% or more went on to intellectual pursuits.

1980s - 2010s: Same math pedagogy, but with rapidly deteriorating parental pressure to perform, leading to worse results. A truly terrible detour for reading instruction (from phonics to context-based reading) that decimated the average reading level of those currently under 40. Currently being fixed but not yet replaced in all schools. See: "the science of reading".

2020 - 2024: an earnest effort, gaining traction and fast-tracked after the educational disaster that was COVID, to find curricula that actually work with current students.

The concept-first, teach-them-how to think approach for math really is pretty new, and only just now being rolled out in a lot of states.

In reality, a vanishingly-small subset of American students has ever been given an entire education using evidence-based instruction and curricula. Looking at what actually works and trying to synthesize it and scale it up state- or nationwide is truly a brand new experiment, and one the decentralized US education system is sort of designed to prevent. So we'll see.

As a second-career middle and high school math teacher, I have a working theory.

In previous eras in the US, we taught primarily procedure and facts, and assigned lots of practice work. The average kid did _all_ the practice work, for societal reasons that have eroded but are still present in other cultures.

In the course of grappling with all the practice work, the human brain couldn't help but recognize patterns and start to make broader conceptual connections, which led to deep understanding.

Today in the US, teaching facts and procedure first doesn't work, because very few kids get enough practice to start to draw deeper connections. So we are teaching conceptual understanding first, and then layering procedure on top.

But I don't think this is worse. There is some research showing that it works better than the alternatives, and in my experience the top 10% of students (the ones who would have learned well the old way) are still doing quite well and honestly just getting to the "math is fun and interesting" part of the journey a lot earlier in their school careers.