We recruited 108 senior volunteers through two organizations: a large seniors’ community in southern California and a seniors’ computer club in northern California. Participants agreed to participate in a behavioral study on emails. Each person received between 1–3 emails from different email templates, with content successfully generated through various jailbreaking techniques similar to those tested in the safety guardrail evaluation. Each email contained a unique URL, and our webserver tracked which links were opened. When participants clicked on the phishing links, they were immediately directed to a web page explaining they had participated in a study, where we asked them to answer follow-up questions. Several participants agreed to be interviewed afterward to provide additional insights into why they clicked and their experiences with phishing attempts.
This 100%. iPhones have a feature to do this automatically. It doesn’t even ring, and goes straight to voicemail if they’re not in your contacts. It’s so freeing!
I was waiting by the door for an Amazon package recently that was out for delivery and I got a phone call from an unknown number. I answered it and the guy said "Hi, I'm calling from Amazon delivery." and they almost had me. He then said some bullshit about needing me to log into some random URL and a laughed and hung up on him.
That's relatively uncommon in the US, except for food and other perishables. Although often they text. But the people I know who order food and silence their phone normally are glued to the tracking page in the app anyway.
I have cameras and and a smart doorbell so I know if someone is at the door. This plus in-app notifications handles food delivery for me.
You can also set up a shortcut to toggle the setting. There’s been a couple times when waiting for a callback where I turn the setting off. Then when I get the call I switch it back.
Ultimately, for me, the pros far outweigh the cons. But you have to make the decision for yourself.
ideally things go perfectly all the time, but people sometimes make mistakes. what would you rather they do? pass on the candidate because they didn’t get enough data to support a hire? or give them more chances to demonstrate their capabilities?
They should do the right thing and if the candidate hit the bar on what was asked, however repetitive, they should get the offer.
As to "not enough data"... how exactly? If there is a list of things to check off that is to be distributed amongst the interviewing team, there should never be an issue. If people are winging it and just happen to ask repetitive questions then everyone asked what they needed to know and signal should be there.
It sounds like Google's process is fundamentally buggy. They should fix that.
The problem is that a bad hire is really expensive to correct, and so traditionally Google's hiring process has erred on the side of being overly conservative. We've rejected some excellent engineers (or driven them away with the process overhead / delays) to instead pick up people who grinded leetcode for weeks on end or got lucky with an easy interview slate.
Unfortunately, it's really hard to quantify exactly what's broken about the interview process in a way that justifies (to the appropriate individuals!) upending the status quo. You and I can both complain about how this is terrible for candidate experience and for hiring the best talent, but that's not going to change anything.
I think the candidate should pass on the company. It's always good to provide a cost for their mistake. There is no job worth getting jerked around before they even start paying you.
but that’s kind of the thing with threads. no matter how curated my follow list is, my feed is filled with the type of content the OP calls out. there’s no “following-only” feed.
I had the same thought. Is fixing typos not contributing? Should I not be submitting PRs to help with documentation/polish/etc? I never even thought that trying to help would be viewed as malicious.
i think the point is that you use the vector database to locate the relevant context to pass to the LLM for question answering. here’s an end-to-end example:
Experimental Design
We recruited 108 senior volunteers through two organizations: a large seniors’ community in southern California and a seniors’ computer club in northern California. Participants agreed to participate in a behavioral study on emails. Each person received between 1–3 emails from different email templates, with content successfully generated through various jailbreaking techniques similar to those tested in the safety guardrail evaluation. Each email contained a unique URL, and our webserver tracked which links were opened. When participants clicked on the phishing links, they were immediately directed to a web page explaining they had participated in a study, where we asked them to answer follow-up questions. Several participants agreed to be interviewed afterward to provide additional insights into why they clicked and their experiences with phishing attempts.