I worked with GNU APL for a while and really liked it. It's also possible to extend it with the C foreign function interface (FFI). The best way I found to input the APL2 symbols with my normal keyboard was with a customized XCompose definition where the input chords are mnemonics of the actual symbols: https://gist.github.com/smartmic/cdb8b0b3936ab965213748813b6...
I've got some notes on setting up input on OpenBSD as well. It enables Left Ctrl and Left Alt for APL symbols, but also a Unicode escape hatch with Right Alt and Caps Lock: https://github.com/turtleyacht/ap-el-kb.github.io
That's very cool, but it's not an apples to apples comparison. The reasoning model learned how to do long multiplication. (Either from the internet, or from generated examples of long multiplication that were used to sharpen its reasoning skills. In principle, it might have invented it on its own during RL, but no, I don't think so.)
In this paper, the task is to learn how to multiply, strictly from AxB=C examples, with 4-digit numbers. Their vanilla transformer can't learn it, but the one with (their variant of) chain-of-thought can. These are transformers that have never encountered written text, and are too small to understand any of it anyway.
I assume that the porn industry will not agree to consider the training material as fair use. And this industry has quite a lot of money for good lawyers. In this respect, we will see on what basis the quality improvement in the relevant content is to be achieved.
Social media is also a wonderful tool for influencing participants and controlling them in the long term. In other words, behind the economic purposes there is a darker, more profound effect that is dangerous in the hands of a few powerful players. In the case of TikTok, that would be the Chinese state. Why shouldn't US Big Tech also be interested in this kind of power, in addition to the extra revenue?
Yup, which is also why various social media owners bent the knee to the administration, and now TikTok is about to become state-controlled too. The long term effects of subtle social media propaganda will become apparent in the years to come. Or, will be vocalized, they already are apparent - I'm convinced social media and related, 24/7 "news" media are a big factor in the right shift in politics worldwide.
Concerning Fred Brooks "No Silver Bullet", I disagree on this conclusion:
> Modern AI has thrown a wrench into Brooks’ theory, as it actually does reduce essential complexity. You can hand AI an incomplete or contradictory specification, and the AI will fill in the gaps by cribbing from similar specifications.
In your writeup, it seems like you're arguing that LLMs can't eliminate essential complexity, and I agree that they probably can't.
But I do think they can reduce essential complexity.
As a concrete example, here's me prompting Claude 4.1 Opus to define a domain-specific language for creating computer-generated paintings.[0] I just provided the requirements and left a lot of ambiguity in the specifics.
In that example, has LLM reduced essential complexity at all?
To me, the answer is clearly yes. It wrote a spec based on my requirements. I could potentially do better if I defined it from scratch, but if the LLM-generated spec is good enough, then it likely isn't worth the cost of me doing it myself for the marginal improvement in quality.
When LLMs first came out, I felt like I had no need for them because I think I can write code better than they can, and I enjoy writing code. But as I've started experimenting with them, I'm realizing that there are some problems that I can solve with software that I don't actually enjoy implementing and I don't care that much about specifying every aspect of my program's behavior, so LLMs fit well in those situations and eliminate essential complexity that would otherwise fall in my lap.
The only complexity AI reduces is the cognitive complexity of writing it. The code itself almost certainly will not be free of Brooks' nonessential complexity, and the reader is SOL.
It's like putting on an exo suit, lifting something very heavy and putting it on a shelf, then asking your teammate to go paint it.
> Is the use of a compatible licence a "re-licensing"?
> No. The original code will stay covered by the EUPL. It is the combined work only that could be, when needed, covered by the compatible licence. In this framework, a combined work results from merging functional codes covered by two (or more) different licenses. The simple action of "linking" does not merge functional codes and in such case the various linked parts will keep their primary licences. This is resulting from the European law, since Directive 2009/24/EC states that interfaces (APIs, data structures etc.) needed for making two programs interoperable can be freely reproduced/used in the various source codes, as an exception to strict copyright rules.
> To be legitimate, the use of the compatibility clause must result from necessity: using it for the sole purpose of relicensing a copy of the original work would be a copyright infringement.
Martin Tournoi, the developer of GoatCounter, wrote an article explaining why he chose the EUPL licence for GoatCounter. It provides valuable insights and a useful comparison: https://www.arp242.net/license.html
He has modified EUPL (bad idea to modify licenses) so it recreates one of the problems he has with AGPL (that a lot of companies will not use software licensed under it). If anything EUPL is less attractive to those companies if they are outside the EU as it imposes EU jurisdiction.
Other people might see the above as an advantage, as you can dual license and those who do not like AGPL can buy a commercial license.
He has misunderstood the AGPL (there is no requirement to send changes to the original developer, only make them available to users).
His modification of EUPL is to explicitly remove one of his advantages, by reducing compatibility with all but two (AGPL and OSL).
h
It looks like his EUPL is not compatible with the EU's EUPL as a result of that modification, and the only way to mix the two would be to license as OSL or AGPL.
>If anything EUPL is less attractive to those companies if they are outside the EU as it imposes EU jurisdiction.
So then for people who are living inside EU it is a good ideal to use this license instead of AGPL if they don't want their code ended up in some transnational big corp for example?
GoatCounter is an open source web analytics platform available as a free donation-supported hosted service or self-hosted app. It aims to offer easy to use and meaningful privacy-friendly web analytics as an alternative to Google Analytics or Matomo.
