gcr

I think the latter half of your post is missing

What do you mean? Data science is advancing at an incredible pace IMO! DuckDB is the coolest thing to happen to data science in my neck of the woods

'btrfs send' is more like diff/patch than it is like rsync. It requires both source and destination to agree about the source snapshot.

Is this the “Improve model for everyone” setting under “Data Controls” or is that a different checkbox?

The report pdf confirms Sebastian was the Bubeck in question

it's possible to do a RLHF or RLVR pass pretty quickly. I'm almost certain a full pretraining run isn't possible within that time frame.

300e9 output tokens at the current Astra per-token API pricing ($50 per 1e6 output tokens) would be roughly $15,000,000 ignoring input tokens.

this is a highly marketable problem and specific teams at OpenAI were aware of specific competitor efforts. I would be surprised if this were happening on a large scale, but

1. Less than a hundred people in the world are working at this problem, 2. A significant fraction of those happen to work at competing hyperscalers, 3. Those hyperscalers repeatedly show themselves not to take user privacy seriously

The employee was named as “Sebastien Bubeck.” It helps to read the article.

I was wondering if this was going to be some Rust fork of Emacs, à la Java Minecraft vs Bedrock Minecraft

The URL should probably be its landing page, which does a good job with the overview: https://www.audacityteam.org/audacity-4/

Could you explain where this comes from? I dislike DHH as much as the rest, but I thought Hitler was just referred to as Hitler in danish?

That would be a vulnerability in the IDNA filter that they’re responsible for fixing.

This is great!

A couple years ago I adapted a quite similar concept for the PUCO 8: https://www.lexaloffle.com/bbs/?pid=106664 My version allowed players to re-use letters though. Yours is much more interesting and challenging.

I remember being pretty proud of my compression system that let me fit several thousands of words on the cartridge.

In NYC, most independent shops give a debit card discount / credit card surcharge.

People don’t rely on AI because they’re worried about political correctness. That’s a pretty weak assumption.

That ignores the men with guns, who would probably say that their alert is very important.

Take a look at the article. The issue is more complex.

You already can’t download and run an unnotarized binary on M-series Macs without stripping the gatekeeper quarantine flag, so…

Amp will probably claim the agent wrote the code and the human approved it

TIL! When was this functionality added? It wasn’t in llamacpp when I looked in June

Seconding this. I’ve found that llama-cpp doesn’t completely evict all of the GPU memory with its idle sleep parameter configured.

There is no curvature in the screenshot, only a semitransparent background

Sounds like a great way to get your account revoked.

I think this is an evolution of Zed’s “multiplayer” CRDT features from a few years back, but with some new agent affordances.

For others considering this, note that Claude removes session history files from ~/.claude after 30 days

The problem with combining mailing lists, patch exchanges, and Git is that you have to combine mailing lists, patch exchanges, and Git.

Most projects are not created under those conditions.

It's not just a step, it's a leap towards real productivity that your agents will actually use!

The reason why the attack works without “biometrics or device unlock” is because it only works on passkeys that were issued without requiring user verification:

> The Pass-ta-key attack is effective when the relying party does not strictly require user verification. Many relying parties configure WebAuthn’s userVerification parameter as preferred rather than required to support diverse devices and user experiences, making them susceptible to this attack.

So it’s possible for a relying party to mitigate this attack by requiring user verification and checking that the proper bit was set.

Later on, the article outlines an issue with the way that Chrome interacts with Windows Hello during passkey registration and another issue where Chrome dumps the TPM’s master key in process memory, but these are endpoint concerns and have since been patched

Mac users are used to checking ~/Library/Application Support/App, ~/.app, and ~/.local/config/app

why can’t it be undone?

for new installations going forward, why not put the folder in XDG_CONFIG_DIR and an optional symlink from ~/.cargo that can be opted in