zacmps

Then you just prompt-inject the agent to convince it to deploy a malicious patch. This isn't a solution.

I am skeptical of these results given that the end has:

> Favorite settings The model regularly used enable_sort=off and random_page_cost=1.1

If random_page_cost wasn't set correctly for the default cases postgres's query planner can generate terrible plans (unless you're running on a spinning disk).

That could easily explain the difference by itself.

I've long thought there should be a standard mechanism for back pressure for loads that can defer their usage.

For example dishwashers, overnight battery charging, car charging, etc should be able to receive scheduling data from the grid to determine the best time to run.

Users could of course be given a choice whether to use grid scheduling in exchange for lower prices or not.

Gave it a quick try today, it works pretty well and is noticeably faster than the official app.

Obviously not, there's no need to create a new character for a new operation. You can just define it as @ or or any other symbol you want.

What benchmarks are you referring to?

Rust itself doesn't have a scheduler of course, I assume this is comparing against tokio or one of the other async executors?

This is the paradox of tolerance. There are a few suggested "resolutions", the most common being "tolerate everything except intolerance".

I now daily drive firefox, there are unfortunately plently of broken sites. Nebula's video player is broken in widescreen for example.

From memory this is true and nvml (Nvidia management library) is the way to get stats that doesn't cause the GPU to wake.

I agree, but it does have faults. Performance is woeful, and managing an on-prem instance is (literally) a full time job.

Titles and thumbnails have a huge impact on video performance, and when it's your main income it seems reasonable to try to marginalise the impact.

Badly phrased but not wrong, this is the minimum frame rate for humans to perceive motion as supposed to a slide show of images.

The maximum frame rate we can perceive is much higher, for regular video it's probably somewhere around 400-800.

I would love it if you had time to publish it!

I think what you're looking for is a bike.

Could you elaborate what you think the problems are? I guess they should be using some form of multiple comparison correction?

Not really? At the moment it's done by some user-land job scheduler. That could be something container based like k8s, something in-process like ray, or a workload manager like slurm.

Until you need to schedule GPUs or other heterogenous compute...

LLM summaries of papers often make overly broad claims [1].

I don't think this is a good example personally.

[1] https://arxiv.org/abs/2504.00025

But there has been continual improvement over that time, both in the ecosystem, and in the language (like a syntax for generics).

NileRed?

Not currently, but it is being worked on https://github.com/ray-project/ray/issues/53976.

If you use something like cockroachdb you can have a multi-master cluster and use regional-by-row tables to locate data close to users. It'll fail over fine to other regions if needed.

I suspect the number of langauges it can do with reasonable accuracy is actually much smaller, probably <15.

For literature search that might be ok. It doesn't need to replace any other tools, and if 1/10 it surfaces something you wouldn't have found otherwise it could be worth the time on the dud attempts.

This is not the view everyone holds. For an example of work in the opposite direction see https://dioxus.notion.site/Dioxus-Labs-High-level-Rust-5fe1f....

Try everything from voidtools, it's incredibly fast (both search and indexing).

Might be fixable with promoting and/or a much lower temperature?

I've wanted someone to write an extension utilising this idea since GPT-3 came out. Is it available to use anywhere?

Do you know any resources for learning these heuristics?