I am guessing the author is either criticizing people who are anti-social (in the pop culture definition) or believes he was before and after some thinking arrived at the conclusion that antisociety was not the way. But I don't feel it describes my internal motivations, so I've translated them to my behaviors:
- if someone is confusing or upsetting you, assume it is your fault
- interpret others' actions in the context of your fears (this one is spot on)
- assume your assumptions are wrong and that you shouldn't even bother
- pivot conversations when someone asks you about something you actually know or are good at, it might be a trick, tell them you're dumb instead
- if you must ask questions, convince yourself you must not, just figure it out instead
- dig in your heels at no point in time whatsoever and just tell people the minimum they need to hear so they leave
- do not develop narratives or it means you will have an immediate network
- do not research the acumen or credentials of anyone
- do not grant grace to those who make mistakes, they might actually be wrong and you're not a judge
- when all hope is lost in conversation, pretend to take their side to end the conversation
It's as if you know me, I can feel myself panic whenever anyone asks me anything, and I end up saying completely the wrong thing as if I wasn't fully in control of my thoughts.
gpt-image-1.5 works decently for generating images compared to old Sora, but you pay per generation. It's possible that monthly flat rates were too much of a loss leader for OpenAI. I imagine the server side cost for generating video for Sora 2 is much higher as well.
"Anymore" is right though. This should be a call to change the global mindset regarding dependencies. We have to realize that the "good ol days" are behind us in order to take action.
Otherwise people will naysay and detract from the cause. "It worked before" they will say. "Why don't we do it like before?"
DISA STIG already forbids use of the EPEL for Red Hat Enterprise Linux. Enterprise software install instructions are littered with commands to turn off gpgcheck and install rpm's from sourceforge. The times are changing and we need cryptographically verifiable guarantees of safety!
This might be well-intended to restrict bot posting, but it also silences dissent. HN is one of the few places left on the internet where dissenting voices can post. A dissenting voice already has to work against the hivemind, adding more restrictions will increase the echo chamber effect.
To add on to this, in some organizations it's easier to assess risk according to RMF and similar frameworks if the application ships with stunnel and is configured from within than it is for the application to require a system-level VPN like Wireguard.
That said, I think Wireguard is easier to analyze on the wire since it has a known binary signature from the first 4 bytes, while stunnel tunnel is indifferentiable from web browsing traffic. For a bad actor looking into exfil or C2, this means an stunnel is probably the sneakier and thus more reliable method of encryption on the wire compared to wireguard.
Counterpoint: if they had reached for sharding early, they would have avoided the technical debt of having to refactor their existing database. I don't think sharding is necessarily that complex either, especially for a SaaS style app like ChatGPT where users are mostly siloed in.
I think economies of scale, while only mentioned in the penultimate paragraph in TFA, is an underrated factor. Whenever something looks like alien technology but is available for $200-300, I assume an economy of scale helped.
TFA goes into the industrial engineering efforts associated with LCD manufacturing, but I don't think those wins would have shown up without a huge market for TVs.