Recent submissions

Comment · 10 months ago
https://www.youhere.org: "Want to know who showed up?" is the tag line. It's an app-based attendance service for teachers/coaches/band directors/conference organizers, etc. I'm a teacher and it scratches an itch I had (students not showing up to class). Been hitting the $500/month more and more often now (even after server fees).
Comment · 10 months ago
Hi! Thanks for trying it. The "slow sway" is part of the plan! It's so the oversized first row of a qwerty keyboard can show all possible letters on the top row. If it feels too much like a game, slow it down using the main app from your home screen.
Comment · 10 months ago
Comment · 1 year ago
Comment · 2 years ago
Uni. prof here/shameless plug: I developed a general education class for college students about the 'history of navigation' which includes the science behind navigation and why clocks and (accurate) timekeeping are needed for navigation. I wrote a book about it all I use for the class, if anyone's interested: https://www.amazon.com/Longitude-Time-Navigation-Tom-Bensky-.... I was inspired by Sobel's book, but I needed more math and science in the discussions.

Also, if you can find the book by Williams called "From Sails to Satellites: The Origin and Development of Navigational Science" you'll laugh out loud at his wit about navigation throughout the book.

Comment · 2 years ago
I work on neural nets back in the mid 1980s, so I was somewhat familiar with their structure, backpropagation, squashing functions, etc. I wanted to get back into them, so I started by rederiving all backprop formulas by hand here https://github.com/tbensky/NeuralNetworks/blob/main/Backprop....

Then I wrote up a NN from scratch in Python to train some simple vectors and even got it to train some MNIST characters https://github.com/tbensky/NeuralNetworks/blob/main/ANN/ann.... (but it was slow).

With that basic refresher, I got into PyTorch and worked on training a PiNN: https://github.com/tbensky/PiNN_Projectile (neat video of it training: https://youtu.be/0wlHa1-M7kw).

Now I'm working on understanding kernels with CNNs (this is my question, I'm making good progress on answering): https://ai.stackexchange.com/questions/46180/kernels-on-a-tr....

Having loads of fun!

Comment · 2 years ago
Comment · 2 years ago
Back during the pandemic, hardware-based contract tracers were an idea. I built one using the ESP32; see https://github.com/tbensky/npct. In a nutshell, everyone generates a (non-centralized) hash for themselves based on local entropy. This hash is set to the BLE name of the ESP32. Turn it on and throw it in your backpack as you go out. When two ESP32s pass by each other, they both log the other's BLE name (hence hash). Later on, hash logs could be inspected and uploaded to a central server so you can see who encountered who. Seems like there's still some (non-Covid) applications for this (but I can't think of any). Fun project. Learned a lot about Bluetooth.
Comment · 2 years ago
Back in the early days of Linux (~1993) or so, I was trying to convince my workgroup in grad. school that Linux was a nice alternative for running our numerical simulations (in C). But I knew I had to get the graphics working, since no one wanted to use a text-only interface. It was a Dell PC using some graphics card; the word 'tulip' is in the fog of my memory. The graphics driver was not loading, giving some error about a reference not being found in the .o file. I didn't know what to do, so I loaded up the .o file in vi, searched for the offending reference, edited it out and saved the .o file. From then on the graphics works great!
Comment · 2 years ago
XRF=x-ray fluorescence. You shoot X-rays into a metal, typically the gamma rays from an Am-241 source. This causes atoms in the metal to fluoresce their own gamma rays that are a (somewhat) unique signature of the atom itself. So in sum, XRF allow one to non-destructively determine the atomic constituents of a bulk material (but 'heavy' materials..heavier than Silicon for example).
Comment · 2 years ago
Comment · 2 years ago
Comment · 3 years ago
I remember using Turbo Lightning. It was incredible. You type a word in some DOS program, and when you hit the spacebar, a beep would sound if the word was spelled wrong. Back then Peter Norton was all the rage, so if you typed "Norton" TL would beep and suggest spellings like "notion, nothing, etc." to correct Norton. The suggestions felt so AIish and futuristic to me. Wow a "suggestion" for how to spell something? If you selected a replacement, it would stuff the keyboard buffer with backspaces, to erase the word, and retype the correction. Borland made incredible software back then. Great times.
Comment · 3 years ago

Live from the official Hacker News API

A modern reader for Hacker News · SvelteKit + Skeleton