TOMDM

With all the "Claude rewrote X in rust!" posts lately, I'm waiting on the inevitable Linux kernel rewrite post.

Vibe code aside I feel targeted rewrites are becoming more and more justified.

Have you worked with agents on tasks with high capital and long latencies?

Having worked with Fable 5, the feeling I get is that it's fairly capable of accounting for these tradeoffs and will depend fast more time on planning and testing.

At the end of the day though, with horizons like that the best use of an AI is to get it to help you with those things, not so much delegate fully.

Well, you can say the same thing about HSMs, or the CPU in your device.

If there's no trust afforded to the device holder, or it's manufacturer, there's no trust in anything.

Always to degrees, and never fully, but trust can still be had.

I'd always thought the usefulness of C2PA was limited to verified devices in custody by trusted actors. Like a security camera with a tamper evident enclosure, or an organisation being able to attest that they recorded the imagery.

The idea that it could be used to attest the authenticity of any random person or device surely wasn't a thing serious people expected was it?

Yeah I'm getting this feeling too, that Opus 5 collaborates better with other Claudes, but that some of the older Opus models collaborated with people better.

In a situation where you don't have enough people, keeping them on this task takes up someone who could be processing people, people who may be panicking too much to quickly give a yes/no.

> You’re unable to view this Post because this account owner limits who can view their Posts.

12gpm

I've seen a lot of people claiming that these recent examples are just trivial combinations of existing prior work.

I'm yet to see anyone cite the prior work. If it's true I think some credit is due for the authors Fable/Sol are branching from.

You're not running an Opus 4.8/GPT 5.6 tier model on $5k of hardware at a useful Tok/s

I can delete cookies to obscure usage history

No, the people growing their food built modern civilization.

(Or millions of disconnected stakeholders with different incentives collectively built modern civilization, but who wants to put that on a bumper sticker)

My intuition is that there would be a fairly stable base load, but doing something like switching on a new training run of a frontier model would be incredibly spiky, thousands of GPUs going from somewhat idle to 100% in seconds.

Would demanding that large spikey users of energy like data centers implement some sort of demand ramping/isolation from the grid in the form of a massive capacitor bank or flywheel generator between them and the grid help reduce the risk here?