hamandcheese

If you decode it and open the link, you are potentially leaking data to whoever controls the domain of the URL.

This somewhat validates a fear I've had (but hadn't tested) of codemode with smaller models. It works fantastically with Opus and Sol but I've always wondered how well it scales down.

In its current form, Meta's muse is actually the most pro-user agent among Muse, Instinct, and Grok bot. Instinct and Grok bot are both prompted to hide details from you, not reveal how they work, etc. which is user-hostile.

Muse on the other hand is very much Your Agent and does not (yet) have silly limitations that work against your interests.

I hate Meta as much as the next guy, but Muse is well designed.

I wonder if increased reliance on VPPs increases the difficulty of a black start. Especially since these things are internet controlled, I can't help but feel like there's a bit of a circular dependency. Can't coordinate the VPP without internet, can't have internet without power, can we black start without coordinating the VPP?

Are there peaker plants that come online for seasonal variance?

...is it? I'm looking and it seems like it doesn't have any data. It might be useful if they keep it up.

> The same can be said of Linux, it can be a real PITA finding and configuring drivers for something that should not take more than a few seconds.

If you value your own dignity you should see that these scenarios are worlds apart. In one case you have a trillion-dollar company actively working against your own interests. On the other hand, you had some trouble finding drivers for a volunteer-driven project, which is a problem that you are actually empowered to fix.

Presumably, if you are positioned to plug in Jev (or an LLM classifier), then you are also positioned to collect training data.

If we were talking about speed instead of size, i think it would be perfectly reasonable to say 9x faster. I'm not sure I agree that 9x smaller is unintuitive. It makes sense to me.

Why not? If its the difference between the information being available at all, then I choose login any day of the week.

And what a lot of people seem to miss is that with AI, there is going to be (already is?) orders of magnitude more software in this world. I'm sure a lot of jobs will be eliminated, but new jobs will be created as well. Hopefully enough to balance things out, but we'll see.

> if you can actually afford to use the cloud.

If you are an individual and you want to store a media library, you'll go broke instantly.

I have roughly 200TB of storage capacity, 100TB used, in my NAS. That would be $2300/mo on S3 before even getting to access fees and network egress.

> Creating an issue first is pointless spamming akin to the classic "asking to ask"

Many projects disagree with you on this one. It is quite common to have a policy that forbids PRs without an issue or discussion first.

I wonder how much is for training vs for LLMs doing research. On several occasions Claude has gone digging through kernel archives on my behalf (sometimes at my direction, other times all on its own). Usually to determine the current status of some kernel bug I'm experiencing. Apologies for the load, but I'm sure it was much less than an actual crawler trying to slurp up everything.

Sometimes Earth is the one that is in between.

This seams neat, but I'm not sure I understand the purpose/difference between this and using the iOS simulator.

I believe you were referring to one kind of eclipsed and with my subtle edit I was referring to a different kind of eclipse :)

And when it’s in between we cannot even see it (when it’s exactly in line).

The story of how the code got to a certain state belongs primarily in git commit messages, not in comments. Comments should serve to aid in understanding of the code, highlight footguns or hidden dependencies, etc.

Claude very often litters code with comments about decisions that were made within a single session/pull request, its just noise.

If I'm paying you per-second of screen on time: yes.

> No, you paid for the end result.

No, I literally am paying for the thought process, per token. "Pay only for the result" is not how these things are billed.

I'm curious why people care about such tracking? I hardly use an editor anymore, and I'm not alone. All that matters is who is signing off on the commit.

> Text a human wrote or edited should be considered close to sacred: an agent should be hesitant and have a very good reason to touch it.

Not sure I agree (particularly with code, not prose). If code is risky to change for reasons that aren't obvious, it should be commented as such. It doesn't matter how the bytes were generated.

> Another use case: the README.md, originally generated, where you rewrite the opening paragraphs. The agent should feel free to redo or append parts further downwards but should really think twice changing anything in the opener.

This, I understand.

Doesn't it being a phone chip make the efficiency claim more impressive, not less?

Moreover, from tfa:

> This Dell also beats out both M4 and M3 Mac Studios, only failing to the M4 Mac Mini at 7.57 Gflops/W.

Slightly off topic, but Claude Code (and likely other models/harnesses) are incredibly effective at Nix. It can trivially self-verify, without side effects, which is a perfect match for an LLM.

If you've ever been put off by the difficulty of the language, it's worth checking it out again with AI assistance.

That doesn't sound expensive to me. Not compared to, say, a $1.7M public toilet[0].

To be clear, I do not support Flock cameras.

[0]: https://www.nbcbayarea.com/news/local/san-francisco/san-fran...

> > We’ve reached a point where an update to macOS — or to any app I rely on, really — is a source of dread rather than excitement. I now expect the new version to be worse.

You know what software doesn't fill me with dread to update? FOSS.

Proprietary software isn't built for users anymore. But (some) FOSS still is.

> and not purchase any more.

why would anyone sign such a contract?

Is Claude's paraphrasing of Lord of the Rings equivalent to the original?

Cheap tokens only benefits SaaS that depends on AI. Otherwise, cheap tokens means it is only more cost effective than it already is to cut out the SaaS and build instead of buy.