e28eta

They delivered before you ordered? That’s amazing!

I’m guessing that’s supposed to be 8th and 17th, or 18th and 27th?

It, at the very least, adds additional complexity to exploit your service by putting another layer between attackers and secrets / capabilities held by the application.

And in a case like this, with so many exploitable PHP boxes out there for attackers to choose from instead, certainly buys you some additional protection.

I wouldn’t choose to roll my own on the day I needed it though.

The press release claims “eMudhra will complete its remaining commitments to the browser root programs, including a full root-cause report.”

Does WhatsApp have an obligation to avoid prompt injection on other chat bots?

I don’t think it’s unreasonable for them to include instructions to all chat bots, so that the WhatsApp user on the other end gets the experience they’ve chosen.

It wouldn’t surprise me if the help content including information about regulators makes any chat bots highly likely to suggest contacting the regulators to resolve an issue.

It also seems reasonable for them to want users to actually work with lawyers, instead of chat bots, for legal issues based on the content from their site.

I wonder what apple does when confronted with software that auto-approves the TCC prompt via a VNC connection to localhost.

If there’s some protections that prevent it from happening via localhost, it feels like the next step is to proxy through another host on the LAN.

Heck, I wonder if someone’s already written an app that’ll connect to VNC, look for any permission prompts, and approve. His agent software could just poke it anytime it thinks it’s blocked.

What makes you think RCEs are being found & fixed at a rate that’s faster than they’re being introduced?

I could see it going either way.

> some reason over signing metadata on-device

After my brief read, one of the main reasons they’re using PCC to produce the signed JPEGs, instead of doing everything on device, is that it maintains your privacy.

If you were signing with the iPhone, an attacker could then correlate photos taken with the same device.

Apple certainly has the data that “this sensor, in this device, took this exact photo” in PCC at the time of signing, but they discard that data.

One of the options was always on, or always on while in work mode.

I’m curious if the lack of speaker identification is a privacy thing, or a technical limitation.

It looks like it uses PCC, so it may run into usage limits

As someone on a google workspace plan, where the monthly price went up to include AI features, I’m disappointed that google won’t let me sign up for antigravity development plans with my workspace google ID. To the best of my knowledge, as of last week, the recommendation is to sign up with a different, personal google id.

It feels like such a stereotypical big corp misstep, but based on the reputation that google’s AI products have, I guess I’m not _that_ heartbroken. It’s mostly frustration that they forced the price increase on everyone, and I’m definitely not receiving the benefit.

I have a similar machine, and briefly poked at running a local LLM, but got discouraged after a couple days. The quality, responsiveness, and impact on the rest of the system didn’t seem worth it to me.

What sorts of things are you doing with the local LLM? Anything interactive? Should I take another look?

I’ve casually used Draw Things a couple of times (great tool!), and I did not find it beginner friendly. I remember feeling a lot of doubt, confusion, and frustration.

Am I getting these results because I picked the wrong model? Or I need to improve my prompt? Or the tool just can’t do what I’m trying to do? How current are these model recommendations? Have they been superseded by something newer?

As a true beginner to AI at the time, even the sizes and bits were meaningless to me. And I don’t remember having any context as to what I should be attempting to run on my mac.

So I think you need to include steps 4, 5, and 6 of swapping between different models, quants, and prompts. And step 7 is probably wading through the complicated UI, full of jargon that most people don’t know.

Don’t get me wrong, I recognize that it’s a powerful tool, and the steep learning curve exists because it exposes quite a few power-user features. But for someone graduating from commercial AI image generators that take a text prompt and maybe a choice of couple models, it’s not easy.