BWStearns

It's mostly just annoying (for now). If you travel a lot it can amount to a couple hours extra per trip if you're unlucky.

But the bad part is that it's extrajudicial punishment/retaliation for protected speech. She hasn't been accused of anything that she can mount a defense against, so why is the government retaliating against what is presumably legal behavior? It's presumably legal because she hasn't been indicted for anything or notified of a pending indictment.

If the justification ends up being that they've marked you as a "domestic terrorist" for exercising your first amendment rights then why are they letting you travel at all? Are they going to give you random secondary every time you travel? Will they let you move your money out of the country? Why would they? After all, you're now a terrorist (but apparently not worth arresting).

when to know it’s time to replace my synchronous inter service http requests with a queue

I've found that once it's inconveniently long for a synchronous client side request, it's less about the performance or metrics and more about reasoning. Some things are queue shaped, or async job shaped. The worker -> main app communication pattern can even remain sync http calls or not (like callback based or something), but if you have something that has high variance in timing or is a background thing then just kick it off to workers.

I'd also say start simple and only go to Kafka or some other high dev-time overhead solution when you start seeing Redis/Rabbit stop being sufficient. Odds are you can make the simple solution work.

Anyone have any good literature on alert fatigue as a general concept in design?

Some 9/11 attackers were CIA assets and protected from FBI/police scrutiny as such

Citation seriously needed. This is untrue unless the words you are using are not to be understood in any sense common to English speakers. The most generous fact based interpretation I can give it is that Saudi financiers were underscrutinized for political reasons, resulting in missed opportunities to stop the attacks. The actual attackers were not CIA nor FBI assets.

Me and my wife made something very similar (https://nativi.sh/). I find it's a really good use case for LLMs to actually help you learn rather than doing everything for you.

Also, not sure if you're getting hugged to death but I'm getting this in the interface but not seeing any network failures.

``` Unable to process correction

Error processing your text. Please try again. ```

https://www.reddit.com/r/drones/comments/1fvqiqz/angry_flori...

At least one Florida man is out there plinking Walmart drones at 400 feet with a 9mm. Saw another who took one down from a boat also with a pistol (probably 9mm also), but can't find the video now.

The Venn diagram of "shoots at drones" and "concerned with other people's safety" is two separate circles.

I consider [having a big benefit at 100% vs an 80/20 rule] a characteristic of type systems in general; a type system that you can rely on is vastly more useful than a type system you can almost rely on, and it doesn’t take much “almost” to greatly diminish the utility of a given type system.

This! This is why I don't particularly care for gradual typing in languages like Python. It's a lot of extra overhead but you still can't really rely on it for much. Typescript types are just barely over the hump in terms of being "always" enough reliable to really lean on it.

It was a big miss calling them "prompt engineers" and not robopsychologists.

I wonder if lunar space elevators might be the fix here. If I understand correctly, such an elevator would not be as subject to the perturbations since the tension would keep it's orbit stable (is it still an orbit if it's tethered?).

Another option might be a LORAN style system put up on towers. With lower gravity and no atmosphere I imagine we could stick transmitters up very high without super complex construction, maybe even just a giant carbon fiber tube with a transmitter at the top.

What if we make it accomplish 20% less but spend 10% more money?

Thanks for the clarification. I expected it worked like that but couldn't find it spelled out after a brief perusal of the docs.

I have file A that's in two places and I run this.

I modify A_0. Does this modify A_1 as well or just kind of reify the new state of A_0 while leaving A_1 untouched?

Built a writing assistant for second languages (https://nativi.sh). Still working on it a bit but mostly trying to figure out how marketing works.

Currently working on a few rust tools for drone stuff.

https://github.com/BWStearns/ulog-rs/ https://github.com/BWStearns/flight-engineer

I was mostly just joking about him going off and running for Ohio governor. It does give "send the defeated noble to the frontier" vibes.

https://github.com/lox-space/lox/blob/main/crates/lox-space/...

The example code is helpful for seeing how it'd be used (might be cool to link to it from the README while the docs are still todo)

That's what happened to Vivek.

As far as I know there's not really LLMs good enough that can run locally. Maybe with the R1 improvements and future derivative work that'll change.

We use about 4-6 calls per improvement and use a mix of Anthropic and OpenAI. Interestingly we really couldn't get sufficiently good performance from just one model. It's interesting how they can be good or bad at different tasks where one task doesn't seem materially harder than the other.

What did they do?

Check the plug

I just spent a whole day trying to figure out what was going on with a radio. Turns out I had tx/rx swapped. When I went to check tx/rx alignment I misread the documentation in the same way as the first. So, I would even add "try switching things anyways" to the list. If you have solid (but wrong) reasoning for why you did something then you won't see the error later even if it's right in front of you.

Treat your LLM input like it's user input and don't execute SQL directly from a LLM output.

Anything that speaks mavlink is this. It's also much harder to get from 0 to 1 with this than it should be.

So all the batteries are on the belly and a door opens up in flight when they get heavy and they get yeeted? Now you need to have sensors and actuators and inspections on all these doors. The cost of the plane just meaningfully increased.

Mechanical systems experience failure, so now your pilots need to train on the procedure for a stuck drop pod door. It'll probably change the flight characteristics and fuel efficiency while stuck open, so any time that happens it is now an emergency. That just added some operational costs for stranded passengers. You also would probably need a whole new cert for the airframe for any changes to your drop pods.

Throw them in little nacelles under the wings? Now slipping it onto the runway gets harder, more pilot training. What's the procedure for a battery pod strike? More training and more procedure and more certification. More redundant safety systems.

Throw them in the wings themselves? Now if the drop doors get stuck you really have a suboptimal flight condition.

I don't doubt that the technical challenges are surmountable, but all these considerations could literally double the cost of operating a plane compared to just accepting a lower MTOW. The military does all this crazy stuff because it's a hard mission requirement and they don't care how much it costs.

Awesome writeup! As someone who is roughly in this same stage of hardware learning, this had me laughing.

I went into "screw with it until it works" mode. I took it apart, re-applied the magnet, swapped wires around, did all kinds of stuff. After a while of this, the error finally went away! No idea why though.

I imagine the extra complexity, parachute and release system weight, and risk would nix that.

Interesting. So if it was used for aviation then your takeoff weight would be lower than your landing weight. That'll be counterintuitive.

A 30mm round plowing through someone’s living room because some bath water IQ nitwit thought they saw an alien is not a valid use of government resources. Even if they’re from Jersey.

I've used it for French and German (and some maintenance practice in Russian). For German it was really the only thing I used and it got me to high A1/low A2. I was able to have decently complex conversations while traveling where German was our only shared language, so it can work to a point.

I think you realistically need to supplement it with other input/practice but I've found it really good for two things:

1) Just doing _something_ every day. Unless you go to DLI or Middlebury, you're just not going to learn a language in a month even if you do nothing but study it. If you're not living in your target language then it's easy to just not use it at all. Duo at least makes you light up the neurons for your target language every day.

2) For French I've found it's good at covering random blank spots that I didn't realize I had. My wife is french so I mostly learn from conversing with her and I'm at like B2/C1(on a good day). But she's not going to correct me every time I misuse a tense or preposition. So _most_ days on Duo I mostly get everything correct and learn like 1 or 2 new words, but sometimes it covers some grammatical nuance that I wouldn't have thought to go improve.

Thanks for posting. I was confused as hell why I couldn't install Axios and I was getting a 404 earlier.

Having lived in most of those cities and in Miami I can say Miami is definitely the most dangerous driving I've seen. And my car insurance company certainly seems to agree.

I am morbidly curious to see how Waymos interact with the Good Vibes Only crazies. Miami is definitely going to be hard mode for self driving.