If you stop trying to find something like truly ontologically novel about it, you might be able to understand what it actually is. Okay it's not impressive. It's not incredible. It's not groundbreaking new technology. It is what it is.
It is being discussed as though it were ontologically novel, karpathy is saying it's something new and he is an authority, so to me there is something here that i am not seeing. I promise i am not trying to be a naysayer or poke holes, I am literally just trying to find out what the hell it is. I would install it but i don't want to install something without knowing what it does, and what is written in the docs is clear as mud.
It's not new in the sense that any of its components are new, and it's not new in the sense that similar things had not been done before, it's new in the sense that putting the right components together in the right way suddenly created something capable of starting a viral hype.
Essentially, as I understand it, it is a personal AI assistant running on your computer, integrated with different systems (like email, chat).
Nontechnical people I know are buying hardware to run claws on. As I understand it, the innovation here isn't the tech but in availability/ease of access.
I am not arguing, merely seeking to understand. I'm not saying it isn't novel. I'm saying that I don't understand what is novel about it and seeking clarity.
The answer seems to be simply that it is all of the extant technologies productized for normies, a la Dropbox. Satisfactory answer, got what I was looking for, thank you! As a dropbox user, I may buy a mac mini and try it :-)
That kind of blanket demand doesn't persuade anyone and doesn't solve any problem.
Even if you get people to sit and press a button every time the agent wants to do anything, you're not getting the actual alertness and rigor that would prevent disasters. You're getting a bored, inattentive person who could be doing something more valuable than micromanaging Claude.
Managing capabilities for agents is an interesting problem. Working on that seems more fun and valuable than sitting around pressing "OK" whenever the clanker wants to take actions that are harmless in a vast majority of cases.
That is a terrible assumption to make. Regular lacquer for example does poorly under temperatures commonly encountered when preparing food and it’s basically a mix of solvents.
The solvents evaporate when the lacquer cures, right? A lacquered spatula or spoon could leach some plasticizers when heated up. But who on earth would go to the trouble of spray lacquering a spatula? It doesn't seem like a real concern. Wooden spoons from IKEA aren't gonna poison you!
Flexner's "Understanding Wood Finishing" has a section about "the myth of food safety" that pretty directly states that food safety isn't a serious concern for fully cured finishes.
If you're into Nix, check out https://github.com/mbrock/filnix — not yet integrated & maintained in upstream Nixpkgs, but lets you replace Nix/NixOS packages with Fil-C versions quite easily.
I've got a pure Go journald file writer that works to some extent—it doesn't split, compress, etc, but it produces journal files that journalctl/sdjournal can read, concurrently. Only stress tested by running a bunch of parallel integration tests, will most likely not maintain it seriously, total newbie garbage, etc, but may be of interest to someone. I haven't really seen any other working journald file writers.
I see what you mean, but I think it's a lot less pernicious than astrology. There are plausible mechanisms, it's at least possible to do benchmarking, and it's all plugged into relatively short feedback cycles of people trying to do their jobs and accomplish specific tasks. Mechanical interpretability stuff might help make the magic more transparent & observable, and—surveillance concerns notwithstanding—companies like Cursor (I assume also Google and the other major labs, modulo self-imposed restrictions on using inference data for training) are building up serious data sets that can pretty directly associate prompts with results. Not only that, I think LLMs in a broader sense are actually enormously helpful specifically for understanding existing code—when you don't just order them to implement features and fix bugs, but use their tireless abilities to consume and transform a corpus in a way that helps guide you to the important modules, explains conceptual schemes, analyzes diffs, etc. There's a lot of critical points to be made but we can't ignore the upsides.
I'd been imagining taking the Zig Language Server and adding some refactorings to it—it only had a bare minimum like Rename Symbol. It seemed like a huge project with so much context to get familiar with, so I put it off indefinitely. Then on a whim I decided to just ask GPT-5 (this was before Codex, even, I think?) to give it a go. Plopped it down in the repo and said, basically, implement "Extract Function". And it just kind of... did. The code wasn't beautiful, I could barely understand it, some of which must perhaps be blamed on the existing codebase not being exactly optimized for elegance, but it actually worked. On the first try! We continued to implement a few more refactorings. Eventually I realized the code we were churning out actually needs major revision and rewriting—but it took me from less than zero to "hey, this is actually provably possible and we have a working PoC" in, like, fifteen minutes. Which is pretty insanely valuable.
reply