singhrac

21 ms·

241.
by singhrac 3y ago
I encountered this a lot in my previous work and found workarounds that I write about here: https://rachitsingh.com/collaborating-jupyter/ . At a high level I think being scared of rerunning your kernel is indicative of
242.
by singhrac 3y ago
Can you explain a little more about how it works? Does it handle cases like loops correctly (or self-referencing cells)? Is this running a CPython fork, or how does the lineage tracking work? Are the values “x” and “y” in the quickstart exa
243.
by singhrac 3y ago
It was very disappointing but somewhat heartwarming to see him attempt to give honest feedback until he realized it was just all nonsense. Seems like a genuinely nice guy (love his blog).
244.
by singhrac 3y ago
I don't really understand (I mean, I do understand on some level) why we don't have a decent US government agency that specializes in consulting work like this. Why didn't the BLM get the USDS to work on this website and then
245.
by singhrac 3y ago
I've never once regretted my decision to write GPU code in CUDA... I mean, I wish there were alternatives because being locked into Nvidia isn't fun, but CUDA is a great developer experience.
246.
by singhrac 3y ago
Strongly disagree with this statement (and agree with the parent). In many hard engineering disciplines, appropriate solvers (e.g. SuiteSparse) are barely supported, if at all (e.g. scikit-sparse only supports CHOLMOD, and UMFPACK is suppor
247.
by singhrac 3y ago
No, not really. It was never really as large, and most of their output was in probabilistic programming (e.g. Pyro) and work that was relevant to self-driving cars (point cloud compression, etc.). But they shut down Uber AI in the layoffs l
248.
by singhrac 4y ago
I cannot comment on the feasibility of building a profitable open source business around Python tooling, but I can say that I'm very very glad someone is taking that risk: both Charlie in career/business risks, and the VCs in taki
249.
by singhrac 4y ago
I’m sorry, but this is just not true. Active managers are measured by their deviation from the S&P (i.e. a beta 1 manager is extremely passive). This index has also always been concentrated to some extent, so not sure it’s any kind of r
250.
by singhrac 4y ago
Really? That's essentially just the market cap weights, which is the most default thing you could do. There's a slight deviation because it's float-adjusted and I'm assuming their marks are at some frequency, but those n
251.
by singhrac 4y ago
I'm fairly certain he already guaranteed that with his $150mm previous donation in 2014, which at the time was the largest gift in Harvard College's history. Note that in that case, $125mm was earmarked specifically for financial
252.
Former NYC Transmit President Andy Byford Joins Amtrak(bloomberg.com)
1 points by singhrac 4y ago|0 comments
253.
Interview with NYC Landlord on Policy Changes [audio](bloomberg.com)
3 points by singhrac 4y ago|0 comments
254.
by singhrac 4y ago
I am a Scratch-taught programmer in that I played with Scratch in 5th grade and ended up programming professionally. It was great and I'm sure they've done a good job with it since. I'm very, very grateful.
255.
by singhrac 4y ago
It isn't fast (really as far as you can get from "zero cost abstractions") but you can have Rusty enums via sum types in Python, and a poor man's match expression. It's not even that bad, really. With Pylance (or po
256.
by singhrac 4y ago
People continue to say that Intel has lost its way, and maybe it has in hardware, but for a long time they have had the best numerical computing software teams outside of NVIDIA. You can look at mkl, their recent updates to numpy, etc. This
257.
by singhrac 4y ago
That’s the Mark Cuban Cost Plus Drugs company. This is a competitor. Here’s a good overview of their business model: https://d3.harvard.edu/platform-digit/submission/goodrx-or-g...
258.
by singhrac 4y ago
The spread between wholesale energy prices between high and low renewable times (with low-compared-to-future renewable penetration) is around $20/MW (it’s not $2 and it’s not usually $200), or about $0.02/kwh, so this isn’t a vi
259.
by singhrac 4y ago
For what it's worth, it ran fine on my 2070 (8GB of VRAM), even with the GPU being used to render my desktop (Windows), which used another ~800MB of VRAM. I was running it under WSL, which also worked fine. Note the level of investment
260.
by singhrac 4y ago
I have no idea how far along they are, but I remembered seeing this exchange earlier: https://twitter.com/getoutline/status/1589765154882797568 . If you’re seriously considering switching off Confluence, maybe give
261.
by singhrac 4y ago
Super curious about this - would you be willing to share the order of magnitude of money this brings in? I like your website a lot, though the use of serif text is a bit jarring on the web in this context.
262.
by singhrac 4y ago
What about Outline [0]? The self install process isn’t very hard and you lose a decade of enterprise cruft. [0]: https://www.getoutline.com/
263.
by singhrac 4y ago
I’m not really that qualified either but fwiw the quotes in the article from the NREL paper on grid stability in low inertia settings ignore that the NREL also mentions that ERCOT has managed to deal with higher renewables penetration succe
264.
by singhrac 4y ago
This is kind of a crazy story to me, and likely pointing to other issues. The "4GB just to run the OS" is probably just macOS preemptively caching... and all Macs use swap extensively so using e.g. 10GB of memory actively is total
265.
by singhrac 4y ago
Honestly, I think this is also partly them patting themselves (or their silicon team) on the back for a job well done. And as everyone else has said, their competitor is the fastest Intel/AMD chips if they're trying to win market
266.
by singhrac 4y ago
In my opinion the only thing that would break the NVIDIA monopoly is if some other vendor (maybe Intel) sells a graphics card that breaks NVIDIA's consumer/datacenter divide (e.g. offers A100-size memory for consumers to purchase,
267.
Collaborating with Jupyter Notebooks(rachitsingh.com)
2 points by singhrac 4y ago|0 comments
268.
by singhrac 4y ago
That's not an entire summary, I think. The lack of local (or nodal, as it's often called in the US) pricing means that there's comparatively little incentive to build new transmission. The UK is not a large country, and we ro
269.
by singhrac 4y ago
I lived above L'Industrie for a year in 2018-2019, and I would say that it's very, very good. However, it's possible that Emmy Squared is just as good (can highly recommend Mama's Too as GP mentioned). I'm almost re
270.
Intel Sapphire Rapids finally released(phoronix.com)
2 points by singhrac 4y ago|0 comments

More ›