Just in general these questions are probes on curiosity and ability to show depth too. I’m astounded by suggestions of stating flat out refusal to even try out LLMs or suggestions to over praise the merits as if the interviewers want to hear binary answers. A well thought out pros and cons story wins over binary yes/no answers at pro and anti ai companies alike.

> A well thought out pros and cons story wins over binary yes/no answers at pro and anti ai companies alike.

The issue with this is, you need to know how to really program to be able to articulate the pros and cons, which a new grad would mostly likely not have.

For example, if you want to include how AI can onboard quickly, you really need to understand the pain points like, I tried asking people but really, everybody is busy. Or I've found coding agents help me speed up making code changes, but it some situations, they can help accelerate making mistakes.

I think the issue that a lot new grad are faced with is, you don't know, what you don't know.

This is fascinating that it worked though. Can we just merge all the open weight models and get something better?

I imagine it'd work the same as merging all the good-tasting foods to get an even tastier one

If you go to Civitai this is pretty how it works in that corner of the image generation world

Everything is using Stable Diffusion as underlying model, then most of the usage is merged of checkpoints

most merge improve a small subset of "feeling" benchmark (too small, too specific, or out of distribution) and tend to show degradation on actual benchmark, with especially punishing result on long chain benchmarks.

also only work on matching architectures (i.e. finetunes/loras of the same model)

that kinda worked in llama 1/2 era, not between different models but between finetunes of the same model. the briefly legendary Mythomax was IIRC a merge of 5+ tunes, some of which were merges themselves.

Merging related models has been a very common practice for years. See the Stable Diffusion community.

No, they need the same arch, but you can distill them into a single model. And yes, if you use the API directly Claude will often say it’s an open weight model (likely the ones it was distilled from)

I feel like a very minor tweak to comply specifically with whatever the issue the directive stated and release it under a new name (since the directive specifically names Fable and Mythos, not Opus or Sonnet) while the courts sort it out is reasonable.

Yep took a while but it's down. It's still in the model picker but it's broken

Restart Claude Code and pick up the update to see the acknowledgement that Fable is gone.

This is exactly like a much older Apple 2 game called Old Ironsides fwiw.

On the Apple 2 it was 2 player only but a lot of fun back in the day. With an added ability to ram ships (pointy front into flat side one if you landed it before they sunk you).

Alto hug the iphone sigoff is hilaripus sonce fhe meyboard is so bad it always comes across asa an ask doe forgivebeds

— Sent from my iPhone

The mathematical term for this is the probability of a number being b-smooth. Here ‘b’ is 2^32

Related but not strong enough. 17 x 17 x 17 = 4,913 is 2^8-smooth - no prime factors larger than 2^8 - and it is less than 2^16, but 17 x 17 = 289 does not fit into a byte. Smoothness is required but not sufficient for a product representation to exist.

It's related, but not the same thing. For example, for b=10, the number 70=2x5x7 is b-smooth, but it cannot be written as the product of two numbers less than b. Here are the other b-smooth (counter)examples for b=10:

    | n   | factorization  | products of two numbers
    |-----|----------------|------------------------------------
    | 50  | 2 * 5^2        | 1x50, 2x25, 5x10
    | 60  | 2^2 * 3 * 5    | 1x60, 2x30, 3x20, 4x15, 5x12, 6x10
    | 70  | 2 * 5 * 7      | 1x70, 2x35, 5x14, 7x10
    | 75  | 3 * 5^2        | 1x75, 3x25, 5x15
    | 80  | 2^4 * 5        | 1x80, 2x40, 4x20, 5x16, 8x10
    | 84  | 2^2 * 3 * 7    | 1x84, 2x42, 3x28, 4x21, 6x14, 7x12
    | 90  | 2 * 3^2 * 5    | 1x90, 2x45, 3x30, 5x18, 6x15, 9x10
    | 96  | 2^5 * 3        | 1x96, 2x48, 3x32, 4x24, 6x16, 8x12
    | 98  | 2 * 7^2        | 1x98, 2x49, 7x14

This very thread was an example where it unintentionally got root access though.

Because of how Docker works, not because of how Unix permissions work.

Unix has always had incredibly weak protections between users. You shouldn't rely on it as a security boundary. Think of it as a "keep honest users honest" protection. And llms are not honest.

The protections between users are reasonably strong. Android uses them with great success, by isolating every vendor within their own user. Things start going to hell when everything runs under root for "practicality reasons", like the default, not-rootless Docker setup.

I've seen this sentiment a few times on HN recently I wonder where it comes from?

The only thing I can think of is that if the protected files are on a unencrypted drive, then you could boot from a live-usb(or similar) where you have root and read anything. But that's completely irrelevant as we're talking about a piece of software running on a system without root. In this scenario Unix user permissions are safe, barring user error (such as accidentally granting root, like in this instance)

Of course security holes happens, such as copy-fail, but it's pretty rare in the grand scheme of things, and tend to get patched quickly(like copy-fail was)

That's a terrible distinction to make on a topic about how the coding agent gained root inadvertently.

Fwiw separate machines for the agents is awesome in general anyway.

I have agent frontends running on a low power server where every session is in tmux. So i can just resume from my home pc and pickup where i left off without reestablishing context. I do have to manually feed it data it can access bit that’s also a feature. Also let’s me shutdown the home pc if it’s some long running task since the server is much more power efficient.

I don't think that's the only reason but you're spot on about OpenAI marketing being absolutely terrible. The primary product names of "Claude" vs "ChatGPT" highlights this remarkable difference. To the point where I'm seeing Claude completely take over the generic term for agent.

I do think OpenAI is doomed due to bad leadership. What you said (that the marketing is relatively terrible) and what others are saying here (that the product is worse) is damning isn't it? Are they really failing on all fronts?

The marketing of Claude relies primarily on fear, and I don't think that will have lasting success. Using fear like that tends to backfire once people see past false taking points.