Charlie Loyd 🎃

Charlie Loyd 🎃

@vruba.bsky.social

A pixel/geography/other person. Salish Sea → Oakland. Also @vruba@everything.happens.horse on mastodon. With @rahawahaile.bsky.social. He/him.

902 followers419 following764 posts

Pinned

Replying to @vruba.bsky.social

It’s got a lot of moving parts and I wanted to make it intelligible to several audiences, so the documentation gets pretty into things. I hope it makes good holiday reading.

(It is, forgive me, a bit funny that people writing programming languages competing with C have spent a decade agonizing about not being able to convincingly beat C’s speed specifically on Brotli compression, and now it turns out C’s fast Brotli was insecure.)

This kind of explainer usually contains horrible mistakes or load-bearing misleading simplifications that make it worse than nothing, Muphry’s law–style, so I am happy to repost a good one.

Natalie Gray@natgrayy.es

Here's my summary of why websites are offline right now, and why nerds are laughing at the number 4. I've attempted to explain the jargon in use, so those interested but not working in tech have a chance at understanding what's going on.

Shearlets are a more anisotropic extension of wavelets, approximately optimal at sparsely representing “cartoons” (loosely: smooth regions separated by smooth curves). You use them shrinkage-style to separate wavefront sets from noise, for example, with advantages over vanilla FFT.

Finding and predicting large fires from satellite images is an important and interesting problem with a lot of tough factors. There’s relatively little training data, you have really tricky correlates, it mixes various already-complex systems, etc. It’s just gnarly.

Anyone else noticing more Halloween stuff around these days? Seems like even a few months ago it would have been considered strange to have a skeleton or some pumpkins in your yard. Maybe just a California thing? Not complaining, only pointing it out!

The thing about measurements in recipes is I’m absolutely fine cooking by the seat of my pants. I’m not demanding precision everywhere. I just think it’s silly to say 375 g AP flour, then half a teaspoon of salt and a tablespoon of oregano. Don’t give 3 sig figs on some things and 0 on others.

I’m against this.

Daniel Swain@weatherwest.bsky.social

ECMWF is also indicating potential for *extremely* high ocean temp anomalies (>= +10F) along immediate CA coast. That would profoundly affect CA winter temps & amount of moisture available to landfalling storms. It will also have huge implications for coastal ecosystems.

Good day to be a license-respecting Gaussian splat hobbyist, I suspect.

Jason Scott@textfiles.com

One day, Jordan Koschei of HVconnect announced he was releasing 450+ 4k drone shots completely lacking watermarks and public domain, just to encourage use and beauty of what is my home county. Now, the Internet Archive has mirrored it. Lovely shots! Full metadata! archive.org/details/hvco...

My parents both had high-end educations and chose economically low-end jobs, and I think a lot of people would wisely make similar decisions if it were safe enough to do so. (I remain agnostic on UBI as a specific route; I see plusses and minuses and am not interested in debating.)

This post is unavailable.

Charlie Loyd 🎃 reposted

A neat thing with historical Landsat data - you can go back to 25th of September 1983 and see the wispy high altitude clouds over North Dakota around 17:00 UTC, roughly 3-4 hours before the false alarm. On the left, the missile fields of Malmstrom overlaid on Landsat image from 26th of September.

Stephen Schwartz@atomicanalyst.bsky.social

Today in 1983 in Moscow, just past midnight, 44 year-old Lt. Col. Stanislav Petrov’s calm decision that a satellite warning of the launch of 5 US Minuteman ICBMs was a false alarm—based on “a funny feeling in my gut”—may have averted a catastrophic nuclear war. www.washingtonpost.com/wp-srv/inatl...

Hurricane Polo yesterday. Sentinel-1 synthetic aperture radar, so we’re seeing the wind’s imprint on the water surface, not clouds. The sea-level air pressure in the eye is equivalent to ordinary pressure at about 1.1 km or 3,500 ft. Guerrero coast in the north; swath is ~410 km across.

Charlie Loyd 🎃 reposted

The Facial Expressions of Tethys by artist Julius von Bismarck. A suspended high-sea buoy that moves in real time. Sensors on a twin buoy out in the Atlantic Ocean send live satellite data to motors inside the museum. This makes the indoor buoy copy the exact, real-time movements of the ocean waves.

In “Jurassic Park” (1993), Dr Sattler is realizing that they’re growing extinct plants just as we first see the dinosaurs. Yet the dinosaurs are eating eucalypts (E. camaldulensis?). But (1) eucs are post-K/Pg and (2) hyper-oily leaves were very likely basal. So what the hell?

A post saying “China using bamboo to replace plastic” might mean it’s a high-priority government policy that’s measurably changing the manufacturing sector. Or that one hobbyist who happens to be Chinese was in a tourist’s widely reposted photo. Or anything in between. This post is not about China.

My plan to move public opinion on matters important to me is to spend a portion of my day reposting every article and zinger I see go by that agrees with me to an audience of people who follow me because they find this behavior tolerable. I’m pretty sure this will work eventually.

Sad to see about Annie Dillard, a good writer. She used to live part-time on the island where I grew up. Oldtimers said she would often chop her firewood nude. They considered this unwise: there are splinters, and what if the axe hits a nail?

I guess I don’t know what it is to make things in public right now. For all the obvious reasons. I know I like when others do, but I have very little to offer them. (This is me trying to name something that’s been bothering me, not a request for validation or suggestions.)

When I think about posting more stuff I’m thinking about, I find myself imagining trying to use video or something. I think this would be like trying to pet a cat backwards. I don’t know. I see the slippery slope to clickbait and the kind of #scicomm BS that makes me want to throw away my monitor.

Acquaintances are like “Ooh, you should do the kind of work you want to get hired for” or “Ooh, you should get back to writing online” and I want to say: I did. No one really liked it. I got like five extremely kind social posts about it and then crickets. Nothing about it stuck.

I feel mild disappointment about this outcome but no serious regrets about the effort. I’m glad I did it, glad it seemed to interest some friends at least, and glad that it’s there to be found by people looking in certain directions in future years or decades.

But it has given me a gnawing sense that “Here’s a thing I did in more detail than you’ll think you want at first” is no longer a useful strategy on the public internet. It’s cooled my desire to work up several similar projects to a presentable state.

And so little has come of it. Industry did not hire me off it; there were some conversations but they led nowhere. Academics (some of whom I respectfully e-mailed) have not cited it so far, nor silently adopted its ideas in the way that, cough, sometimes happens.

There’s an Easter egg tucked away that I was using to see if anyone got really interested in the details. Not yet. I think Potato hit people uninterested in pansharpening as being about a weird niche, and people interested in pansharpening as not being in their sub-niche.

I do wonder if maybe it should have been a kind of pseudo-preprint instead of a repo. Or an interactive version that you could watch work (and adjust parameters of) in the browser. Either would have been a big time investment, but so was what I did.

There’s a narrow range of paths that could possibly bring an adult of sound mind into caring about a kind of image processing they’d never heard of before, and I was building trail somewhere else. I was worried about the wrong things.

And I think I packed a lot in there. There’s some in-the-stacks stuff about pansharpening narrowly, but also materials on how we use multispectral bands, how we scale satellite images, and how we think about them that are, I think, nontrivial contributions to bigger conversations.

Where I think I failed was in crafting a graspable surface. There’s a reason journal articles require an abstract and a stately procession of topics. I made something I would find fun to explore, not something of legible appeal. That’s not only word choices but actual structural weakness.

I learned whole new skills from the project and I think it did a good job of its technical goals. As far as I know, it’s on the Pareto front of quality and speed. So it advanced the state of the art in its little niche.

I’m about halfway convinced that the reason we haven’t had a proper Nasa mission to Uranus – one of the most interesting places in the solar system – is that Congress doesn’t want the hyuk-hyuk headlines.

I love re-upping this: for decades the main library at UCLA had the generic name University Research Library because the person most responsible for getting it built was an English professor named Hugh G. Dick and the university refused to name it after him.

There’s a reason you don’t see me walking back old opinions and predictions on “AI” as the situation changes (with public opinion, market outlook, technical progress, etc.). It’s that I’ve been careful to only say true things and not talk out of my ass nonstop. Hope others find this inspiring!