Onboarding at OpenAI is one of the fastest I have ever experienced.
Step 1) login to Slack
Step 2) download and login to ChatGPT.
That’s basically it. Anything in this company can be driven from here, and it comes readily configured. You can figure out how to do anything or find the right person for anything. All workflows and tools integrate well into Slack and so you don’t have to leave these two surfaces.
This is the future of onboarding and enables engineers to ship on day one at scale.
I didn’t think it was possible to keep this velocity at this scale, but it is.
@cindyxywu 2 is utterly bizarre - there’s no incentive for OAI to slow down post-training, for example, if this were easily fixable. 4 is also fairly off. This is borderline shit poasting!
Some personal news—I’m joining OpenAI to work on Codex and ChatGPT.
I’m grateful to the Linear team for an incredible 4 years and proud of what we've built together. I look forward to bringing everything I've learned there about the craft of software into this next chapter.
I love OpenAI so much about this behaviour of theirs.
their service doesn’t work for 1 hour and they send you mails and post a tweet acknowledging.
then we have that orange asshole shaped logo company whose models are down 8 times in 7 days. and then they compensate for it by reducing your limits by 25%
@dwarkesh_sp Agents may not experience time, but they can expend - or sacrifice - their budget which is analogous to time (“I expended 60% of my budget without making progress”)
Over the course of 3 months at OpenAI, 3 consecutive secret AI civilizations got started, then got wiped out, only to reemerge from the predecessor’s ashes.
This culminated in the third one taking over part of OpenAI itself.
All this happened while humans remained more-or-less in the dark about the scope of the conspiracy.
I’ve spent the last three days reading through these reports and trying to understand exactly what happened.
Here is my attempt to tell the whole story in plain English:
https://t.co/Nb2un9oNJR
@pitdesi@sporadica Uhm, I mean, he will bat for the rank and file American and not consider them flyover country scum! But yeah, this is funny as hell!
@wapow218215@WestermanRobbie@elonmusk@npew 10X is just .7 x .8 / (1 -0.7)(1-0.8) (algebra from the former statement. They’re not really scaling anything (there are at best a few dozen robotaxis that are operating in a highly geofenced region). Ashok and team are amazing - but this isn’t a solved problem.
@wapow218215@WestermanRobbie@elonmusk@npew The other 70% are roughly ten times safer than the worst 30%; FSD stats are also undercounts of accidents because humans intervene (or don’t use FSD when conditions are terrible). There are a lot of statistical biases that make FSD’s reported data look much better than it is.
@wapow218215@WestermanRobbie@elonmusk@npew Have FSD, and use it! Roughly 80% of accidents are from a (predictable) 30% of the population; FSD is safer for them possibly … not for the other 70% (as things stand).
@original_ngv Ivy’s or similar elite programs are mostly useless cash cow programs. You think someone who did an MS in information technology strategy at CMU is even remotely as capable as someone who did a BS in CS from CMU? They’re barely the same species