Well I don't see how it even could be open source, since the source is distilling closed weights models, lol.
Also, who cares about having the source? Are you going to tweak it + spend a billion dollars training to get slightly different weights? Who would do that?
This isn't how it works. Hundreds of billions are spend on existing NVIDIA AI hardware that must first be recuperated. This kind of speculation doesn't make sense to me.
Totally agreed. I find it distracting working through a book that's been marked up by someone else. In fact, I get the same feeling even if I marked it up myself, which is why I stopped writing in books. I would prefer to keep my thoughts and the author's thoughts separate. When I want to know what the author thinks, I'll go to the book; when I want to remember what I thought, I'll go to my notes.
You could... but I view JSON as a simple format that you could write a recursive parser for in a few hundred lines yourself.
Bringing in compile-time code generation just for defining static JSONs (which is not that common outside of tests, as most JSONs are serialized and deserialized at runtime) sounds like a hard sell.
If AI were to become a super weapon why should I trust a private company to own it?
If AI becomes a super weapon, it's possessor will no longer care whether you trust it. It would be nice if whether "we" trusted OpenAI et al mattered even now but I see little evidence of that.
But if you need to split the commit in hunks, and add a commentary per hunk, isn't that just a sign that you really should just split your commits?
That's the most killer feature of git, that it's so easy to slice your commits any way you desire, and then redo again. The use case of taking a chunk and commit separately is so common it even got a special mode in the add command. That, and the super fast jumping between branches is what set it apart from contemporary version control systems.
The extra context provided by the review tool is gone when the review is done anyway. Review systems come and go, but the commit log is for eternity.
A stated goal of the API is to have "low ceremony" when reading arbitrary JSON files. I agree the JSON creation API looks pretty clunky, but the parsing API looks fairly simple.
Claude is a far superior _model_ to whatever Siri uses, but in terms of the Claude app correctly transcribing speech, in my experience it's as bad or worse than Siri (which is already bad). Doesn't matter how good the model is if it misinterprets what you say.
Hello Skynet, we have 1.6T debt to hide as there are no chances to recoup that money and investors are getting really angry at us.
Hello, say no more I have a perfect solution for this problem…
The 2025 paper showed how people could fool themselves. How easy it seems that you fool yourself.
The 2026 paper admitted it's different and has issues.
> A lot of software projects don't make money directly. Our company makes money.
That's an evasive answer and you know it. It's about AI making an impact on the bottom line, justifying it's cost.
> A lot of software projects don't make money directly. Our company makes money. This would have cost thousands of dollars paid to a contractor previously with unclear results so yes it's 100% justified.
Thousands of dollars is pocket change even for a small company. That's not the kind of impact AI is supposed to have. You see why I'm not convinced?
You also don't talk about the unsubsidised cost of AI nor the negative externalities.
> though I don't know if it's possible to have that same social experience on the modern web.
The great thing I've found about running an oldschool web forum is this - if you don't run any advertising, Google doesn't care about you and you won't show up in search results.
So only people who are told about it actually come to the site. It's really quite peaceful.
It's interesting to hear the logic he lays out, which seems parsimonious and highly probably at unfolding. I have not been able to think of high probably counter-logic.
It amounts to: robots (physically manifested AI) + digital AI ==> any job can be done vastly better than any human.
Consequently, goods and services will far outstrip what any human can consume. We'll end up with deflation.
So we won't have to work and won't need money. Very exhilarating and scary road ahead, but there's no way to stop it.
If I have two editor tabs open as two browser tabs they should already be able to talk if they're from the same origin, which different editor tabs would be.
The JS code won't be able to modify the browser itself, but neither can ELisp modify the actual Emacs binary code. The customizability in both cases is really coming from an "OS" being implemented in user-controlled scriptspace. You'd have just as much ability to customize that HTML-andJS OS if it was running in the browser as you have to customize Emacs I think...
100% - his articles were a major inspiration, and first line in the credits:
- Thanks to Bartosz Ciechanowski, whose exceptionally well written and designed articles inspired this one.
I'd love nothing more than measure 3), but quality is really hard to assess IMO as even with the frontier models I find myself disagreeing on what is good and maintainable code; so that disqualifies LLM-as-judge, which even at it's best is somewhat RNG and costly.
+ I don't see this replacing anyones workflow to make it clear. This is mostly for fully autonomous agents that are given a task & are asked to execute. For instance, subagents are a perfect example where you can get the aforementioned savings for free, even if you don't run a "software factory" per-se, as you generally do not interact with them.
If memory serves, it has been since version 4 or 5 that they started including real workloads. I agree that it shouldn’t be considered synthetic any more.
Also, who cares about having the source? Are you going to tweak it + spend a billion dollars training to get slightly different weights? Who would do that?