Fair comment, especially as you mentioned in cases where capacity is fixed as part of a reservation. With fixed reservations, we've seen examples where basic monitoring doesn't always tell the full story such as where instances appear running, but sm activity is near 0. We've also heard from teams using on-demand capacity across clouds that they haven't yet stitched together their monitoring to see exactly who is using what, and where, in a single dashboard. That's something that we help provide insights into with our monitoring dashboards.
Even before LLMs, Data Science was being replaced by more specialization, IME.
Data Engineers took over the plumbing once they moved on from Scala and Spark. ML Engineers took over the modeling (and LLMs are now killing this job too, as it’s rare to need model training outside of big labs). Data analysts have to know SQL and python these days, and most DS are now just this, but with a nicer title and higher pay.
Once upon a time I thought DS would be much more about deeper statistics and causal inference, but those have proven to be rare, niche needs outside soft science academia.
Reading a comment like this makes me realize how broad the title “Data Scientist” is, especially this tidbit:
> as it’s rare to need model training outside of big labs
Do you think there are pre-trained models for e.g. process optimization for the primary metallurgy process for steel manufacturing? Industrial engineers don’t know anything about machine learning (by trade), and there are companies that bring specialized Data Science know-how to that industry to improve processes using modern data-driven methods, especially model building.
It’s almost like 99% of comments on this topic think that DS begins at image classification and ends at LLMs, with maybe a little bit of landing page A/B testing or something. Wild.
> Once upon a time I thought DS would be much more about deeper statistics and causal inference, but those have proven to be rare, niche needs outside soft science academia.
you can’t really buy H100s except in multiples of 8. If you want fewer, you must rent. Even then, hyperscalers tend to be a bit inflexible there; GCP only recently added support for smaller shapes, and they can’t yet be reserved, only on-demand or spot iirc.
how would you explain how hard he fought to NOT buy twitter?
people seem to forget he was legally forced to buy Twitter after he tried for months to get out of his joke bid, primarily through claiming he was misled about the extent of bots on the platform
The entire idea is to buy an undervalued platform using insider information, if the stock price plunges after he committed to a price then it's no longer undervalued. This has happened between his bid and termination announcements.
I also roughly remember he had his Tesla holdings as collateral creating some liquidity crisis for him.
This elaborate explanation does not mean it isn't wrong and the original theory of idiot-with-money does not hold
I used to like allowing squashing or fast-forward merges. Most PRs would be squashed, because most developers write terrible commit messages and use merge where they could rebase. But, if you had a well-crafted set of commits, we could retain them.
I’ve recently switched to using conventional commits and release-please everywhere, but that pretty much forces us into a squash-only world, since even the devs who write nice commit messages don’t want to make each commit a conventional commit; much nicer to do it as the PR title, and more visible
It's incredible knowing this and going there today - it's a very small town with an absolutely anemic economy and extremely old homes that mostly haven't been updated. It really shows how much can change for a town in 100 years.
It is still just a little town, but the last couple years it seems to have gotten a bit more spark of money and life to it. Probably because people figured out there was cheap houses and land for sale in the area. Of course it all needs a lot of work to gut and update those old homes and other basic goods are more expensive because there ain't jack shit around except seasonal tourist shops.
for decades, big tech contributed relatively little in the way of python ecosystem tooling. There’s Facebooks Pyre, but that’s about it. Nothing for package/dependency management, linting, formatting, so folks like those at Astral have stepped up to fill the gap.
why is type checking the exception? with google and facebook and astral all writing their own mypy replacements, i’m curious why this space is suddenly so busy
Coming from a Meta background (not speaking on behalf of Meta):
"package/dependency management" - Everything is checked into a monorepo, and built with [Buck2](https://buck2.build/). There's tooling to import/update packages, but no need to reinvent pip or other package managers. Btw, Buck2 is pretty awesome and supports a ton of languages beyond python, but hasn't gotten a ton of traction outside of Meta.
"linting, formatting" - [Black](https://github.com/psf/black) and other public ecosystem tooling is great, no need to develop internally.
"why is type checking the exception" - Don't know about Astral, but for Meta / Google, most everyone else doesn't design for the scale of their monorepos. Meta moved from SVN to Git to Mercurial, then forked Mercurial into [Sapling](https://sapling-scm.com/) because simple operations were too slow for the number of files in their repo, and how frequently they receive diffs.
There are obvious safety benefits to type checking, but with how much Python code Meta has, mypy is not an option - it would take far too much time / memory to provide any value.
Probably because a large amount of AIs are churning out Python code, and they need type-checkers to sanitize/validate that output quickly. Dynamic languages are hard enough for people to make sense of half the time, and I bet AI agents are struggling even more.
how can this be? isn’t this a trivial metric to pull from any clouds monitoring service?
to get the good ones (H100+) you generally have to reserve them, a fixed cost you pay monthly and can’t pretend to not know
reply