Rob

You can build a better future by tearing down bad things. The old adage goes that it's easier to tear down than to build up. Right now, some pretty awful people are putting a lot of effort into building a worse future for the rest of us.

The good news is, the adage still holds. So many of the people I know are exhausted or struggling, but are able to scrape together just enough effort to keep tearing down the things being built by the morally bankrupt. It's fun, prosocial, and energizing.

So, today I learned that LetsEncrypt uses AWS for its outbound http-01 challenges. This is a problem because AWS is *also* a significant source of malicious traffic these days.

Whether it's because of any of a couple dozen different AI slop outfits aggressively crawling sites, or login abuse, or one (large) segment of the occasional DDoS, AWS has earned itself an honored place in my iptables drop rules. Except that also causes a LetsEncrypt outage.

@bennjordan.bsky.social Have you been able to play with Flock's gunshot detectors yet? That seems like a fun application of your background in acoustic engineering. Would an alert be triggered if, say, someone popped a balloon nearby? Or clapped two pieces of hardwood together?

While watching this video, I realized that OpenAI absolutely has the resources to have thoroughly done the same experiment -- but instead, Sam Altman decided to stand on a goofy stage and lie about his product. www.youtube.com/watch?v=GPWo...

Welp, I guess that's it for archive.today for me:

Pretty sure the actual cause is that I'm running adnauseam.io

@textfiles.com National Park Service is looking for anyone willing to pick up books and journals from their DC HQ tomorrow. Just in case you might know anyone in the area.

If you used 23andMe, please consider signing in and deleting your data from it ASAP: oag.ca.gov/news/press-r...

A brief DeepSeek primer, for those who have seen it mentioned recently but don't read tech news every day:

And how does all this involve China? DeepSeek is a Chinese firm. If you use the API and their commercial service, then the usual censorship is in place (i.e. it won't answer questions about Tiananmen Square). The usual geopolitics. There are some security concerns in some contexts.

A brief DeepSeek primer, for those who have seen it mentioned recently but don't read tech news every day:

What does this mean for software engineering? I've been bearish on the influence of LLMs on dev work, based on my own experiences and those of other devs. LLMs are okay at generating basic, boilerplate, junior-level code, but for complex things, they suck.

A brief DeepSeek primer, for those who have seen it mentioned recently but don't read tech news every day:

Why is it a big deal? It performs very well. It doesn't have some of the "hallucination" problems associated with other LLMs. It can handle more complex problems. Oh, and also it's open source and far more efficient than e.g. OpenAI's models. You can run it on consumer hardware.

A brief DeepSeek primer, for those who have seen it mentioned recently but don't read tech news every day:

What is a "Reasoning" model? LLMs are more-or-less text generators. They encode lots of patterns of text, and then produce as output other patterns of text. They do this well enough to seem conversational, but they don't "think". A Reasoning model appears to think.

A brief DeepSeek primer, for those who have seen it mentioned recently but don't read tech news every day:

In multi-factor authentication, a factor is a) a thing you know, or b) a thing you have, or c) a thing you are. Passwords started out as a thing you know, but now should be a thing you have (an unguessable, randomly-generated password, stored in a password manager).

It has been a long time since I had a more heartfelt "honestly, fuck you" reaction to an academic text. (From "Computing and Technology Ethics" by Burton, Goldsmith, Mattej, Siler, and Swiatek).

Of particular note, by the time the reader hits this particular excerpt, there has been *zero* discussion of differing ethical frameworks.

It has been a long time since I had a more heartfelt "honestly, fuck you" reaction to an academic text. (From "Computing and Technology Ethics" by Burton, Goldsmith, Mattej, Siler, and Swiatek).