Hacker Newsnew | past | comments | ask | show | jobs | submit | e12e's commentslogin

In my mind, "yolo ai" application (throwaway code on one hand, unrestrained assistants on the other) - is a little like better spreadsheets and smart documents were in the 90s; just run macros! Everywhere! No need for developers - just Word an macros!

Then came macro viri - and practically - everyone cut back hard on distributing code via Word and Excel (in favour of web apps and we got the dot.com bubble).


The main thing that held me back from using Asahi on my M2 MacBook air - was missing external display. If I read TFA correctly - that should now work with a custom kernel.

If that's true - I'd say MacBook air M2 is probably the new sweetspot - depending on how cheap you could get an M1.

My impression is that until now, MacBook air M1 was the sweetspot.


yep it should work on M2 Macbook Air.

> (...) and the only communication channel would be towards me (enforced with things like API key permissions).

> This should prevent any kind of leaks due to prompt injection, right ?

It might be harder than you think. Any conditional fetch of an URL or DNS query could reveal some information.


DNS Queries are fine, and also conditional URL fetches, as long as they are not arbitrary, should be okay too.

I don't mind the agent searching my GMail using keywords from some discord private messages for example, but I would mind if it did a web search because it could give anything to the search result URLs.


> did not have egress to the Internet. It did have curl and Python, but not much else.

So trade exfiltration via curl with exfiltration via DNS lookup?


Exfiltrate what? It's an empty container.

There do exist container escaping exploits.

Is this referring to some section of the announcement?

This doesn't seem to align with the parent comment?

> As with every new Claude model, we’ve run extensive safety evaluations of Sonnet 4.6, which overall showed it to be as safe as, or safer than, our other recent Claude models. Our safety researchers concluded that Sonnet 4.6 has “a broadly warm, honest, prosocial, and at times funny character, very strong safety behaviors, and no signs of major concerns around high-stakes forms of misalignment.”


Does snikket recommend/facilitate federation with other servers?

Yes, definitely. To me, the idea of a chat server that doesn't federate is as absurd as setting up an email server that doesn't federate. I understand that today people know more contacts with email addresses than XMPP addresses, but if we ever want to free ourselves from the current walled gardens, we need to stop treating chat as something that only happens in walled gardens.

Some people get worried about the idea of "federation", thinking that it somehow means their server is less private, and their data is being spread across a mesh of servers, and stuff like that. That's true in some decentralized/distributed chat protocols, but not in XMPP. Connections between servers only happen on-demand, similar how when you send email between different email providers, they will connect to each other to deliver the messages.

However we do have a feature which allows disabling federation access for specific accounts, for example to prevent kids from communicating with anyone outside their own Snikket server. This is a feature I want to expand on, so that you can permit communication with a limited number of approved contacts on other servers.


> (...) he should be fired for it.

I don't know about that - I'd say it's the managers responsibility to make sure employees don't feel pressured to work when they're to ill to function.

And also brings to mind the IBM one million dollars story:

(...)

A very large government bid, approaching a million dollars, was on the table. The IBM Corporation—no, Thomas J. Watson Sr.—needed every deal. Unfortunately, the salesman failed. IBM lost the bid. That day, the sales rep showed up at Mr. Watson’s office. He sat down and rested an envelope with his resignation on the CEO’s desk. Without looking, Mr. Watson knew what it was. He was expecting it.

He asked, “What happened?”

The sales rep outlined every step of the deal. He highlighted where mistakes had been made and what he could have done differently. Finally he said, “Thank you, Mr. Watson, for giving me a chance to explain. I know we needed this deal. I know what it meant to us.” He rose to leave.

Tom Watson met him at the door, looked him in the eye and handed the envelope back to him saying, “Why would I accept this when I have just invested one million dollars in your education?”


Did it happen? I'd like to believe it but it's a lot of money even now and in Thomas Watson's time it was worth a great deal more.

> An expert panel is set to discuss whether Japan's health ministry should approve two products using iPS cells. The treatments are for heart disease and Parkinson's disease.

Nice link - still relevant IMNHO - even though it's from 2021.

Discussed at the time:

https://news.ycombinator.com/item?id=28892933


It has updates at the bottom. Most recently 14 Sep 2025.

I think this continued anthropomorphism "Have you tried asking about..." is a real problem.

I get it. It quacks like a duck, so seems like if you feed it peas it should get bigger ". But it's not a duck.

There's a distinction between "I need to tell my LLM friend what I want" and "I need to adjust the context for my statistical LLM tool and provide guardrails in the form of linting etc".

It's not that adding prose description doesn't shift the context - but it assume a wrong model about what is going on, that I think is ultimately limiting.

The LLM doesn't really have that kind of agency.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: