Beautifully old-school Web in so many ways. Besides the obvious (the layout, the "Sign the Guestbook" link) it is the whole "love" displayed by the site.
Kids, this is what the original web was like. Dedicated (maybe obsessive) site creators that (by hand) put together a site as a tribute to their passion—perhaps hoping to find other like-minded souls out there.
I wish there was a way to find these sites more easily... I know they're out there. I miss the original idea of The World Wide Web from my childhood :/
• There is overhead in learning how a specific game engine works.
• Often, due to a game engine API, it seems to herd you into writing the same game everyone else is writing with that engine.
I wanted just enough "game engine" to abstract away the pixel-buffer, windowing, user-events on the various target platforms and then do no more.
"I have been using SDL3 as it does everything I need as a cross-platform abstraction over the system - from windowing, to game controllers, to rendering."
And that is exactly where I landed as well. SDL3 [1] absolutely matched what I wanted. Then again, I enjoy writing sprite-based games. If you want to write a 1st-person shooter though I'm sure you will still want to go with one of the giant game engines.
For reasons I have been unable to articulate, I have believed that if license plate readers are a thing, someone should open-source one that the general public could share.
When a comment comes back that someone malicious will use the tech to stalk someone I really have no answer for that. That is indefensibly a bad thing.
Of course the tech exists nonetheless (not open source as far as I know) and it could be argued it is even being used maliciously—or at least without the kind of judicial oversight we assumed would be in place. But one actor behaving badly does not justify the use of the tech by everyone.
I read about the citizens of Minneapolis using decidedly low-tech means to track ICE vehicles prowling their community and it suggested a scenario where an open license-plate-tracking solution could "balance the scales" a bit (that is if you believe the scales to be imbalanced).
I imagined not a license-plate reading dragnet but rather software where you had first to enter in very specific strings of ASCII characters and the software would only announce when there was a specific string match from the camera.
To that end I vibe-coded an app for iOS in about 15 minutes using iOS's Vision framework and the built-in phone camera. Anyone could do the same.
Nonetheless I only tested it with "HELLO" and "WORLD" using scraps of paper in my kitchen and never tried it outside as the craziness in Minneapolis seemed to have quieted down.
There had been a fairly decent open APLR library called OpenALPR. It used to be updated fairly regularly, but then was "acquired" and now sits abandoned for almost a decade without any updates, while the company that acquired it (REKOR) has commercialized it and extended it as a closed source API and platform.
ALPR is fairly trivial now days, and a modern CPU can process thousands of plates per second (dozens of plates per frame, hundreds of frames per second coming multiple cameras pointing at different lanes of traffic).
The "product" around all of these is linking the plates traveling together, tracking routes through various cameras, identifying common travel patterns, and more disturbing travel patterns that are outside norm for a given plate or route (i.e. do they normally drive these 3 cameras M-F, but this Th, they went a different route and stopped somewhere for a while).
All that used to be done by the police/detectives/investigators on their own. Now the AI is automating this, and that is truly terrifying, especially for how often misreads occur.
Ive wanted to know this for a long time. Why isn't there some package I can download to a raspberry pi with a camera, and contribute to a public website that shows things like, oh I don't know real time positions of government vehicles...
I've never understood the obsession with token/s. I'm fine with asking a question and then going on to another task (which might be making coffee).
Even with a cloud-based LLM where the response is pretty snappy, I still find that I wander off and return when I am ready to digest the entire response.
Your workflow is unusual, oftentimes there is a vigorous back and forth, or a desired output like code generation, etc where a low tk/s drastically effects ux and user productivity.
But the real kicker here is the 90s ttft, that means you ask a question and don't see anything for a full minute and a half.
You are fine with it. But may be rest of the world is not. Anyway, to compare performance/benchmark, we need metrics and this is one of the basic metric to measure.
Brought a dozen or so floppies I had to a vintage computer festival last year and handed them off to someone who would archive them.
As I worked in a university computer lab briefly in the late 1980's, I had "captured" a few early Macintosh viruses on a couple of floppies. The recipient of my floppy collection seemed delighted by that, ha ha.
reply