sgc

They allude to this alternative tech in the article, and I think it will stay the dominant approach because the far finer dimensions of the new tech talked about in the article, even if integrated into a film using glass beads as they also did, appears to be intrinsically much more susceptible to rapid functional degradation. It's about or less than the thickness of dirt / grime / bug goo. But tests will tell.

If the application method is as rudimentary as sandblasting, it sounds rather simple to retrofit to existing aircraft. If it works as they state it does, it's a virtually free same-day fuel efficiency boost.

However, I did not see what the actual net improvement was. When they talk percentages, they are talking only about "in the transition zone". They say the coefficient improves throughout, but in theory, it could be almost irrelevant if the overall improvement throughout the profile is close to 0. It also sounds like a very difficult level of precise degradation to maintain for any period of time in real world conditions, since it would be easy to clog or abrade further.

Thanks for sharing that, I am looking to improve my agentic use, and it will be useful as I develop my own path.

OMG, not the once in a lifetime viral opportunity!

You will never win this crusade, because there are too many people here who know from experience a VPS is neither expensive, nor under-performing up to millions of users a day, nor hard.

FWIW, I once abandoned an Ubuntu server for a decade, and managed to update it painlessly in 20 minutes. That same server is still running today, now with the latest LTS. I think I might have even started with Ubuntu 4, or perhaps 6, and it has been painless all the way through. Perhaps my slow upgrades saved me from early adopter woes :).

I use Debian now much more. With all the supply chain attacks, Debian Stable feels like an absolute jewel, even if there are always a few packages I need to handle separately because I want or need a more updated version. But I love the old school no-nonsense engineering ethos.

More like very rarely (1-2 per 1000), very partially aware. I could not find anything saying that it was common, and it appears cases of actual awareness to the point of having pain / trauma are far rarer still. People who have this tend to have foggy memories or other concrete PTSD symptoms after the fact. It does not appear to be the norm.

I still think this experimentation is absolutely insane and I strongly object because there is no way to get feedback from the "patient" after the fact. Since we have no real idea of what is happening, I believe we should err on the side of caution. "But they could consent beforehand" is not morally acceptable for intrinsically inhumane actions that take away fundamental human rights and dignity. So if you think this is possibly inhumane / potentially torture, it is an irrelevant point since true consent would be impossible.

I immediately thought that was the real reason. Their models will quickly break without some sort of consensus on how to reliably exclude them.

I hope you saw others' correction to the title here that indicate Olah is just a speaker at an event on the same day, not an author (although he almost certainly was consulted for his opinions while the Pope and his assistants were working on the Encyclical). So what matters here is what the Pope is trying to do, not Olah's intentions in his minor role.

The Pope has already spoken quite a bit about ai, and exhorted priests to keep ai out of their homilies, which should be a sacred fruit of prayer and study.

Just from what I have seen he said and my Catholic Theological background, I would say he will definitely be talking about at least a couple things: 1) the relationship between ai and our intellectual labor, and how to use it fruitfully to grow without losing ourselves in it (a very similar concern to many on hn as far as I understand); and more importantly for him and again for many 2) how to use ai in society in a way that everybody can enjoy the fruits of it, instead of just the elite few (similar to the priority of Rerum novarum). This Pope chose his name because of this theme, and has consistently demonstrated that social justice is amongst his highest priority concerns - to the point that he has asked the Church to stop focusing so heavily on sexual ethics because there are such weighty injustices in the world that require our focused effort and attention.

I have the same problem on catholiclibrary.org, but insist on having something as the book date for every work. My solution is to temporarily default to the author dates until the book date can be refined. If there is no known author date I at least have a date range, hopefully to century or better.

Author dates are a much smaller data set, can be generally supplemented from public marc records (viaf, loc, etc - I don't do that, but it's an option) and at least provide basic filtering / sorting.

> good ideas ... aren’t substantially in greater supply

Which is sad because they should be. People should be freed up to think and create better things, instead these companies seem to be doing the equivalent of locking their employees in stalls like they do on some animal farms, so they can churn out 'results' ever faster.

You must be living in absolute opulence :)

Subterfuge so simple it is designed to target only the most foolish.

There is no intellectual property in ai output, at least as far as legal rulings in the US currently stand. I guess you are asking whether they will use them once you unsubscribe. My cynical take would be probably they will, but read through their fine print ;)

I think that is not an exception, but is pretty common almost anywhere in the US rural enough to have a roadside stand. Although bowls are not immune to wind, so generally the exact implementation takes that into account. Often a slotted box with no lock, etc.

We are in the context of already having to pay. You are at their mercy no matter what, so the only value of any interaction with them is based on hoping they have incentive to maintain their promises to protect their reputation etc.

It's not a good situation to be in, but still, try to make the best of it.

If you have to pay, at least try to negotiate 1) a guarantee that the hackers won't just do it again sometime later, and 2) full disclosure / assistance in repairing your vulnerabilities so you have some kind of head start for the future. Outside of politically motivated hackers, this would probably be reasonably successful.

I don't know of anything Ubuntu is doing that is significantly beyond what Debian is doing in this regard, nor that they have a distributed reproduction system set up???

To move away from organizational dependence, there should be an installable project for debian where I can dedicate some configurable small percentage of my compute when idle to reproducibly building debian components to make a robust verification system, starting with the most critical code.

Obviously, it would be a ton of work to make such a system resistant to gaming by malicious actors (see GNU Guix for useful efforts), but it would provide valuable diversity in architecture and (political or other) control.

It would be even cooler if we could have independent projects that could run on various distros and OS, and build packages for any of them. Having packages for bsd verified on linux and vice-versa with statistical logging (this code has been verified x times on y OSes) would be reassuring.

Well at least that doesn't require functioning DNS. This time around, it in fact could not have been an email :)

My poor fellow. You wrote about how something is a bad tool for a long list of serious reasons. Then it failed spectacularly because everybody decided to depend on it anyway - exactly what you were cautioning against. But somehow you have to respond to people who think you are the one who got it wrong! As a third party the whole affair gave me a good chuckle at least ;)

They are obvious in the links, no time for silly games.

You actually think that LLMs are not fed docs on how they work in order to help users interact with them better? Asking an LLM how to use it is based on the reasonable presumption that the company making it will prioritize making it useful for users and work on programming it with its own best practices.

Again, it makes perfect sense as well based on how they are trained in the first place. Look at how they tokenize whitespace and you will see why it's useful. Each number of repeating white spaces gets a unique token (so 2 whitespaces = token1, 3 whitespaces = token2) - so it actually does make a very clear reinforcing hierarchy readily available. And we all know if there is anything an LLM needs, it is reinforcement of important points.

Are you referring to technical implementation or the poor anti-privacy decisions they keep making when you say 'slightly better'? I have not given up, but I am profoundly disappointed and for somebody who says they have used FF for so long, it feels like I am being gaslit when you say they are peachy.

People have problems with what they choose to program, not the quality of their code. I too have used FF since the beginning, but switched to Waterfox last year (it took me about two years to make that decision - I didn't make it lightly). I chose WF in large part because its profile remains compatible with FF so I can switch back if they calm the F down and start acting normal again for long enough to rebuild some trust.

https://en.wikipedia.org/wiki/Criticism_of_Mozilla_Corporati... - start at the end for most recent.

Also go to the website of any one of the FF forks and read their reasons for existing. For example:

https://www.waterfox.com/#why-waterfox

Every LLM I have ever asked about this says they perform better when they receive pretty-printed code because it is easier to see structure and priorities. It has been an almost universal recommendation for me, and it makes sense since LLMs are just mimicking human expression.

In another field I have heard it called going for the jugular; the vivid description helps get the point across nicely.

If you want to master something, you will have to know the hardest part. So just deal with that first and then everything else is easy, because you are dealing with it as somebody who has already mastered the domain.

General tip is *iterate*. Look at what it does right and does wrong, and refine. My most complex prompt took me 2 weeks of work to get right, and I just spent a half a day improving that even more. Obviously not worth it unless you are going to be be doing something major. In my case it is for 2 years of work, so clearly worth it.

Somebody else mentioned they had great success at math heavy code. I had to develop a complex piece of software that also integrated into 4 existing systems with a lot of poorly documented constraints. I tried with the major models and Gemini provided the most structured solution that would allow me to work on it, add features etc in the future, and it created an MVP in one shot after working through the planning stage in detail. I have managed to work on that code afterwards quite successfully. It is by far the best model for language tasks like OCR and translation. In my opinion the benchmarks, which put it first for this, are far from showing how far ahead it is because it responds so well to iterating on a prompt. So I think it is good to great for a wide variety of things, but you have to iterate. If what you are doing is simple enough you don't need or want to do that, then use the best model you are already comfortable with. For me that type of work is currently done with GPT 5.5.

I think it's decidedly preliminary to compare models using the same .md file, since they respond quite differently to the same input. I try to narrow to the top 2-3 and then refine inputs for each one. For me it's unfortunately not much better than an intuitive process of trial and error.

Gemini is not at all unusable. It is quite usable for the tasks it excels at - to the point that it is the top pick for many tasks and I spend more money there than elsewhere. On the other hand it responds quite differently from the other major models - so that claude and gpt on one hand are similar and gemini requires a different approach. In my opinion people who think gemini is worthless have not learned how to prompt it correctly. Again, it's intuitive and watching concrete response difference due to small input changes, but if I had to summarize it shows its google books / google scholar roots.

I have started experimenting with qwen more than deepseek, but I have not had good results yet. Given the good press I presume I will learn how to interact with it for better results.

Curious if others have similar experiences in comparing models usefully, or if most don't bother with this, or do something else? I mainly use models for highly focused specialty tasks, so this fine tuning makes the difference between usable and unusable. I don't yet have the luxury of defining my preferred workflow and finding the tool for the task. Everything just breaks almost immediately if I try to shoehorn into my preferred flow.

Of all the standards, I find the SCAA the least useful. That was the standard 30 years ago, but nobody serious about improving their craft is brewing such a watered down espresso anymore, and a 1:2 ratio is the most common one. So 15g for 30ml. And for pressure, modern consensus is that 9 bar is the high upper limit. 10 almost always leads to a harsh cup. I find the 6 bar lower limit to be a tipping point in the concentration, such that lower pressures are more similar to drip than to espresso - the viscosity and mouthfeel changes dramatically. So I can't agree that whatever comes out of an espresso machine should be called espresso. But we are debating vernacular here, I just hope you enjoy your cup of coffee!

1-2 bars isn't really espresso, it's a moka - which can be quite good in its own right. It falls into the genus of pressurized brewing though, so same genus different species. There's a no man's land from 3-5 bars that is not really used and might be worth exploring, but most 'standards' consider 6 bar the minimum to be a true espresso.

Yes of course, but by dropping opv pressure on most home machines that also drops flow due to opv bleed off. If you have your opv at 9-10 bar you get a full flow shot on a coarser grind - which is way too fast. You need to drop pressure until you can keep your wet time the same (I use the pretty standard 17g coffee 34 ml out, 34 seconds wet, upped to 40 seconds because I use preinfusion). If you have a fancier machine you can just adjust flow directly, but then you can set pressure directly so why wouldn't you do that too.