I use the Apple Vision Pro pretty much every workday for at least one to three hours. Sometimes, when I’m deep in coding or other major projects, I’ve been in it for over five or more hours at a stretch. I use it exclusively with my MacBook Pro or iMac as an immersive display, with the Magic Keyboard and Trackpad paired to the computer. This setup is basically perfect. I frequently have some mellow ambient music in the background, which, in combination with the immersive work environment, helps me concentrate like never before for long stretches without getting distracted.
A major improvement happened with the release of beta version 2 of their operating system, which made the keyboard visible even when the immersive environment is fully turned up. This change allowed me to have the best of both worlds. Obviously, I have it plugged into an outlet through the battery or it would never last for such a long time. I’ve tried going back to the Oculus, but there’s no comparison. I love it!
I saw that but here’s an alternative take on what happened:
While the session file definitely shows the AI agent using sudo, these commands were executed with the presumption that the user session already had sudo privileges. There is no indication that the agent escalated its privileges on its own; rather, it used existing permissions that the user (buck) already had access to.
The sudo usage here is consistent with executing commands that require elevated privileges, but it doesn’t demonstrate any unauthorized or unexpected privilege escalation or a self-promotion to sysadmin. It relied on the user’s permissions and would have required the user’s password if prompted.
So he sudo commands executed successfully without any visible prompt for a password, which suggests one of the following scenarios:
1. The session was started by a user with sudo privileges (buck), allowing the agent to run sudo commands without requiring additional authentication.
2. The password may have been provided earlier in the session (before the captured commands), and the session is still within the sudo timeout window, meaning no re-authentication was needed.
3. Or maybe the sudoers file on this system was configured to allow passwordless sudo for the user buck, making it unnecessary to re-enter the password (I just discovered this one, actually!).
In any case, the key point is that the session already had the required privileges to run these commands, and no evidence suggests that the AI agent autonomously escalated its privileges.
Is this take reasonable or am I really missing something big?
Thanks for clarifying, and I’m sorry to hear this is happening. LLM agents have a lot of promise, and it’s frustrating to see baseless fear being stirred up. There’s already enough uncertainty around what’s legitimately needed to get the tech and usage right.
Among the good reasons for SB-1047 to have been vetoed are that it would have regulated the wrong thing. Here’s a great statement of this basic flaw: https://law.mit.edu/pub/regulatesystemsnotmodels
Not speaking for MIT here, but that bill needs a veto and a deep redraft.
I used Hackmd in the past to share the draft of my book (1) and liked that people don't need to have an account to comment. Google Docs was no option as no markdown support and account required. The process worked well but I found Hackmd too expensive for just getting feedback. Stash looks promising for this use case.
(1) Written fully in Markdown in Obsidian at this point. I moved to Asciidoc since because of formatting. The early draft is still available on Hackmd though. Details in my bio.
We've used HackMD in my OSS project since Google Docs was failing us as far as version control goes. What's great is that you can sync a HackMD document to GitHub or similar, so you can collaborate on a document and then push it to a repo.
It’s about agents. This is the first clear public signal from OpenAI that they are in fact going to release agentic AI capabilities (enabling the model to take actions on behalf of the user, and this will inevitably eventually include consumer and other commercial transactions). Altman says the model will “…. the ability to take actions on your behalf”. The mention is rather buried near the bottom but it hit me as the most important and profoundly transformative (mostly in good ways but also potentially very risky ways) roadmap item yet announced.
This looks really promising for complex legal reasoning tasks and other challenges. How can I track progress? Is there an email list or something? Thanks!
I’m way out of my depth here but if I’m reading this right a flow model would not base outputs on a prompt but instead, after it’s sampling/inverse flow pipeline (if I got that right?) is activated it would somehow just output whatever image it’s statistically sampling at the moment. Is that even close?
A major improvement happened with the release of beta version 2 of their operating system, which made the keyboard visible even when the immersive environment is fully turned up. This change allowed me to have the best of both worlds. Obviously, I have it plugged into an outlet through the battery or it would never last for such a long time. I’ve tried going back to the Oculus, but there’s no comparison. I love it!