For context, the point of the Superalignment team was to work on a problem known as scalable oversight: the problem of aligning models in a way that holds up as models become more capable [1]. The reason behind this is that current alignment techniques (like RLHF), have limitations which are expected to worsen as models are scaled up [2].
This is to say, the objective of the Superalignment team was precisely to work on techniques that would work for models which don't yet exist. They are of course aware that they don't yet have superintelligence.
Personally I think they are on the right track, which is why we're over-invested in them.
Generally in our experiments, GPT-4/Claude Opus is able to generate Next-flavored React Server Components (RSC) a lot easier than anything else, because their APIs are simple.
The needs of internal apps & dashboards are so spiritually different than the needs of consumer-facing software, and Next has been all about the consumer-facing side for the last few years. I think betting on Next is correctly simply because they're massive, but in some ways it's a bet despite everything that Next has been about for the last few years.
these days i'd choose remix instead. next 14 is fine, but i feel more productive with remix and when i don't know how to do something, remix usually has a way better "principle of least surprise"
Sometimes they get sidelined by vcs, but you can't see it from lower levels. Other times they started something else, and have insider information that the company will not work
You should realize that faang swe income, on a risk adjusted basis, is the best you will get. Hold on as long as possible and invest your earnings into the stock market. Ignore the noise. You quit when you have enough to retire
I think there is something to be said for starting a company when you're younger and more risk-tolerant though. If you wait until you're middle aged and set for life you might never do it.
conversely, money early on in life is worth more than money later on in life. This can be especially true once you get into a house you're happy with.
Your housing expenses are likely to be your single largest (fixed) expense, so if someone can work hard for a few years and buy a place, it's a lot easier to coast on lower paying jobs.
Agreed from personal experience. And a huge component of housing is where you live.
I'd argue that the single most financially-beneficial choice you can make as a SWE is finding a remote or hybrid job, then leveraging that to live somewhere you enjoy that allows you to save much more money.
I'm a stickler for diversified asset class allocation
Edit:
Normally I am not a fan of explaining myself, but the most fundamental principle of wealth management, diversification, calibrating your risk exposure to various markets, is being downvoted. Thank you for the reminder people
A company that lets you "invest" in fine art. They sell shares in famous paintings & such that they buy, hold, and sell (and charge big fees to do so, naturally). I think they predate the NFT craze, but when they saw it happening they started running extremely aggressive astroturf / content marketing campaigns trying to convince people that fine art was an "alternative asset class" worthy of investment. These articles/posts had a bait and switch quality to them: you would start reading something that sounded conservative and reasonable about the importance of diversification and then BAM, they hit you with "alternative asset classes like fine art" and then you would realize you had been reading an ad. That's what I'm parodying here.
I always found it strange that VCs were so focused on GPT for X companies. In my view, there is no reason why GPT for X isn't just GPT. Once the model is fine tuned for this task, there isn't anything terribly special about a given X.
That being said, enhancing existing products to be chat friendly will definitely be an improvement.
reply