Claude Code already does this, you can access it with /resume, /rewind and /fork. I'd imagine building a version that saves in the repo instead of in the home folder would take very minimal effort.
I'm using it on a large set of existing codebases full of extremely ugly legacy code, weird build systems, tons of business logic and shipping directly to prod at neckbreaking growth over the last two years, and it's delivering the same type of value that Karpathy writes about.
If you watch the demo video you can see how they would get this: the model is not aggressive enough. While it doesn't cut you off, which is nice, it also always waits an uncanny amount of time to chime in.
I hate to be the first one commenting to say this, but here it goes: the flashy LLM writing style, "Apple Event Dialect" in the README and in this comment is very recognizable and also quite irritating. If this is supposed to be boring then just state the facts and the benchmarks to prove them.
Nobody has to do anything, least of all massive corporations with country-sized revenues. It's /always/ a choice to comply or to put up a fight and deal with the consequences.
I can transfer money from Europe to Brazil in seconds with Wise. I press the button and the money is nearly instantly available in the Brazilian account via PIX. The same in the reverse direction is possible but only if you have a more modern bank in Europe, eg. N26 or Revolut.
This is a weakness of the Python typing system and not necessarily of individual typecheckers. Pyright has a policy of only implementing what's standardized, and the Python type system is simply inadequate to annotate most real world Python code out there. It's been years now and something as basic as properly typing kwargs is still not supported.
Ty could solve this if they rebel and decide to ignore the Python typing standards, which I honestly would appreciate, but if they take the sensible approach and follow the standards, it won't change anything.
> There is a great deal of often heated debate about these matters in the literature of the cognitive sciences, artificial intelligence, and philosophy of mind, but it is hard to see that any serious question has been posed. The question of whether a computer is playing chess, or doing long division, or translating Chinese, is like the question of whether robots can murder or airplanes can fly — or people; after all, the “flight” of the Olympic long jump champion is only an order of magnitude short of that of the chicken champion (so I’m told). These are questions of decision, not fact; decision as to whether to adopt a certain metaphoric extension of common usage.
> There is no answer to the question whether airplanes really fly (though perhaps not space shuttles). Fooling people into mistaking a submarine for a whale doesn’t show that submarines really swim; nor does it fail to establish the fact. There is no fact, no meaningful question to be answered, as all agree, in this case. The same is true of computer programs, as Turing took pains to make clear in the 1950 paper that is regularly invoked in these discussions. Here he pointed out that the question whether machines think “may be too meaningless to deserve discussion,” being a question of decision, not fact, though he speculated that in 50 years, usage may have “altered so much that one will be able to speak of machines thinking without expecting to be contradicted” — as in the case of airplanes flying (in English, at least), but not submarines swimming. Such alteration of usage amounts to the replacement of one lexical item by another one with somewhat different properties. There is no empirical question as to whether this is the right or wrong decision.
It might be instructive to consider for example the history of genocide, in particular of civilian collaboration in state lead genocide. It might be instructive to consider why the genocide convention criminalizes not only acts of genocide, but also incitement of genocide. Why it criminalizes not only the failure to prevent genocide, but also the failure to prevent incitement of genocide. The US has an extraordinarily strong position on freedom of speech, it is nowhere near a universal moral value.
People kill other people is a statement so simple as to be devoid of any positive meaning. What are you actually trying to say? Don't justice systems almost universally contain notions of incitement of crime, criminal negligence to prevent a crime, and other accessory considerations to the actual act?
Don't justice systems almost universally have several levels of responsibility in relation to intent, which at its most basic level can be established by predictable outcomes?
If, for example, you are a leader of armed forces, and also a leader of organizations capable of creating propaganda. Let's say you create and distribute some propaganda (maybe using some AI tools), and a predictable outcome of that is that soldiers will be more lenient in their consideration of the rules of engagement and international law. In that case, one could at the very least establish that you were negligent in your creation and distribution of propaganda. The actual crime would have been the people killing people, namely your soldiers, but you would certainly be given some responsibility for that.
You can similarly take a small next step after that and consider that a company producing, distributing, and profiting from a dual use technology capable of creating propaganda and disinformation that can be responsible for crimes could be held at the very least morally accountable for those crimes, if not criminally.
Responsibility, accountability, moral and criminal, are not black and white notions. They are heaviest and easiest to attribute around physical acts of damage, but they stretch far and wide. To think otherwise is to allow the people with the most power to rampage unaccounted.
Classifying image generation and manipulation as "art programs" is the most beneficial possible reading of it. When you use them to generate disinformation, incitement and propaganda, they are potentially maiming and killing humans. This failure mode is well known, the mitigations ineffective, yet here we are, about to take another leap forward after a performative period of "red teaming" where some mitigation work happens but the harsher criticism is brushed off as paranoiac.
That business model is almost certainly out the window after the Facebook acquisition. They renewed everyone's license automatically, and are probably going to remove it entirely.
Have you profiled the build? Is most of it spent in the compiler?
A more practical point is that the main competitor is C++, a language notorious for long compilation times, so not-ultra-fast compile times might not be the highest priority for people working on rustc.