tbenst

17 ms·

151.
by tbenst 7y ago
Thanks for your efforts! Inspiring to see our broader community spring into action. I was part of a volunteer team that tested 3400 people on Friday/Saturday in Santa Clara country for COVID-19 antibodies [1]. It took a team of 100+ vo
152.
by tbenst 7y ago
I ran FreeBSD for years, and this title still surprised me. I thought it was gone from ports too for ____ reason. Misleading
153.
by tbenst 7y ago
PyTables includes zstd by default
154.
by tbenst 7y ago
PyTables with Zstd compression is likely the fastest performance option of them all. In my benchmarks, PyTables is up to 10x faster than H5PY
155.
by tbenst 7y ago
The best alternative right now for voice coding is https://github.com/daanzu/kaldi-active-grammar . wav2letter is also under use by some, although requires more effort to setup
156.
by tbenst 7y ago
Wish they’d review on Linux. Windows does not seem like the target audience given all the limitations.
157.
by tbenst 7y ago
I think Atom’s hydrogen and VSCode’s python are best-in-class Jupyter clients that achieve everything Jupyter Lab set out to do with more and better features. I develop scripts that function top to bottom with a notebook side-by-side that o
158.
BIOS Bug – Passwords for Asus, Acer and Sony(biosbug.com)
12 points by tbenst 7y ago|4 comments
159.
by tbenst 7y ago
Nix really shines here, has ~17000 R packages pinned and reproducibly built: https://hydra.nixos.org/eval/1560470#tabs-still-succeed
160.
by tbenst 7y ago
The part on “Make code more concise“ struck me as an anti-pattern that should instead be a simple function composition. Or the functional programmer in me notes that the author re-invented the Maybe monad. I do like the point in Learn the u
161.
by tbenst 7y ago
Fun read, I learned a lot! Thanks for writing. I’m curious if you used three or more probe servers in distinct locations if you might be able to roughly triangulate server location? To increase accuracy, seems the underlying topology should
162.
by tbenst 7y ago
Wow this article is total crap, and the brain comparison is basically absent from the paper, for good reason. Here’s what they actually did: - grow “metallic dendrites” from the bottom up, eg make a chip with random connections with memrist
163.
by tbenst 7y ago
Tried to use Wayland earlier today on Ubuntu 18.04 for fractional scaling support, but sadly there is no support for nvidia drivers. This does not seem well advertised so posting as a PSA to save others from trying to debug the login loop!
164.
by tbenst 7y ago
I wrote a lay summary of this paper that compares and contrasts with the whole brain imaging and stimulation literature for those interested in a lighter read! http://www.neuwritewest.org/blog/can-we-reverse-engineer-th
165.
by tbenst 7y ago
Curious why you use Conan vs Nix? Reproducible builds and dependency management is the Nix raison d'être.
166.
by tbenst 7y ago
A neuron uses far more than electricity to communicate. The number of neurotransmitters and receptor types alone is staggering. There’s a reason why biology evolved past using gap junctions. The computer science view of the brain compares h
167.
by tbenst 7y ago
This is interesting. Sounds like you do this on the system level? I’m interested in doing this on a shell or package level eg I’d love to have system on bleeding edge but be able to jump back to same binary blobs associated with a particula
168.
by tbenst 7y ago
Ah do you use fetchTarball for nixpkgs? Although maybe I’m not understanding how you’re using it, could you share an example? Interested in your workflow!
169.
by tbenst 7y ago
It’s largely but not fully deterministic eg backward pass is not deterministic: https://stackoverflow.com/questions/50744565/how-to-handle-n... That said, it’s hard enough to even run an old model 2 years later le
170.
by tbenst 7y ago
I had the opposite reaction--it's great for NixOS to be gaining corporate support!
171.
by tbenst 7y ago
I’d love for a way to write nix expressions that easily lead to reproducible builds. Of course this is possible right now next, but requires substantial legwork to pin Nixpkgs version or add SHA for every tarball. Especially now as the mach
172.
by tbenst 7y ago
Unfortunately, these work very poorly. It takes considerable effort to erase, even with cleaning solution, and often leaves a permanent residue. As a better solution, I would suggest the whiteboard stickers/film/vinyl that can be
173.
by tbenst 7y ago
This is trivial to do on NixOS: services.xserver.videoDrivers = [ "nvidia" ]; For a full install for eg machine learning, you'd want these lines, too: environment.systemPackages = with pkgs; [ cudatoolkit ];
174.
by tbenst 7y ago
It’s hard to justify an investment that has both significant market and technical risk as a VC. SBIR is a great mechanism to help fund technical development when companies are too early for investors and too applied for academia.
175.
by tbenst 7y ago
I’m a panel reviewer for NSF SBIR and the comment about a “preferred vendor already being chosen” is absolutely not true. In fact, firms that have been funded for years by govt grants are at a disadvantage compared to “true” startups.
176.
by tbenst 8y ago
Checkout the work on homotopy type theory and proof assistants like Coq
177.
by tbenst 8y ago
Juno is actually quite nice. I find it far more productive than Jupyter and superior to Hydrogen
178.
by tbenst 8y ago
fMRIs use a crazy strong magnetic field—typically >= 3 Tesla. The signal that is measured is not electrical but rather the blood oxygen level. Effectively, this is a correlate of the metabolic expenditure of nearby neurons. Wass it is a
179.
by tbenst 8y ago
Very different. Retinal waves are a developmental phenomena and use action potentials for propagation. The study at hand attempts to disrupt all action potential transmission.
180.
by tbenst 8y ago
Neuroscience PhD student here. The skull acts as a low-pass filter. EEG recordings are typically low pass filtered at 70Hz or lower for example. This is why you can’t decode eg speech from a EEG: the neural encoding is at a higher frequency

More ›