BWStearns

I've found my GA flying to really be about trips that you _wouldn't have made otherwise_, but the biggest barrier is cost. I've done little adventure trips with my wife that were super fun and without a PPL we wouldn't have done the equivalent. But when it comes to boring stuff it's just too expensive to justify flying so the trips don't happen (since I have to rent).

My suspicion is that the middle of the Venn diagram between "can afford a $500K aircraft" and "doesn't fly GA because they're not going to stay proficient enough to be safe" is vanishingly small. The doctors who fly their Cirrus once a year think they're Maverick so they don't care about the safety features. The guys who fly every chance they get aren't the ones stall spinning on base to final.

Always happy to see new GA aircraft, and the tech here looks very cool. The emergency autoland and chute will definitely be lifesavers. If your system can reliably prevent base-final stall spins then that would clearly be awesome.

One thing that jumped out at me is what are you trading to get that protection? While I have no issues with fly by wire, I'd be nervous about a GA aircraft that's overly opinionated about what it should be doing vs how to do it.

* Can I put it into a forward slip, or will it refuse or return to coordinated flight? This could be critical in an engine related emergency where you're too high and can't or don't want to risk getting farther from the airport.

* If I need to make a steep turn to avoid someone heading the wrong way on base is it going to let me?

I see the value in stall prevention but, outside of stall spin on final, that's largely been solved in GA via passive aero design characteristics making it tough to stall and harder to stay stalled.

Writing assistant for your second language(s). Right now it’s a web/mobile app but we’re getting ready to launch our browser extension so hopefully that will be more convenient for our users.

https://nativi.sh

Also since the backend is in rust I’ve been contributing to a bunch of the libraries I’m using. There’s definitely been a cost to using rust since a lot of the web dev packages are on the younger side, but the reliability has been huge for a side project. Other than one time I set the wrong send grid API key I don’t think we’ve had a single 500.

I am using Timescale at work and I like them a lot. If your data is nicely structured it's a breeze. But my data is kind of pathological (source can just change the structure and I gotta put up with it), so I'd honestly use Influx in a heartbeat if their pricing wasn't totally insane.

Worked on a billionaires yacht when I was a kid. It always had to be stocked in case they showed up but they hardly ever did. So we had to eat all the lobster and steak before it went bad, and the owner’s chef made the meals for practice and experimenting. I’ve never worked at FAANG but I’d be shocked if anyone south of the C suite had as good office lunch on a regular basis.

They also tried to kill the CEO of Rheinmetall. The GRU in particular (but also the FSB) occasionally just goes and does some crazy shit.

There was a drone that had a new return home failsafe mechanism (for when there's low battery or some fault etc). Unfortunately it was possible to misconfigure it so that home was not set and the failsafe was still armed. This led to several drones perishing in the Atlantic as they bravely tried to fly home to Null Island.

Making sane DB indices and constraints. It's amazing how often people just don't add indices even when the access pattern is clear from the outset. "Premature optimization is the root of all evil!" ok, so when are we actually going to add that index? (Answer: never)

I wonder if the harsh post impact environment selected for small, short lived creatures due to the lower caloric energy requirements, and faster genetic iteration speed respectively. If your environment is particularly hostile with low resources then taking many small bets quickly might be the favored play.

I like the 401->autokick back to homepage pattern.

I think the type signatures are really helpful in this regard. It makes it really easy to tell what’s happening and if I have to worry about mutations.

Writing the signatures can become a pain because you have to think everything through up front, but we spend more time reading code than writing it.

No idea how it is for new grads. I hear it’s bad but also you only need one job, so I wouldn’t stress too much. Developing some domain expertise in the business that you want to work in can be a pretty important differentiator.

What worked for me was counterintuitively not applying to a ton of places, but focusing on places I'd really want to work and then reaching out directly to relevant team leads. It might feel like you're wasting effort since it's more investment and people might still just not reply, but I found it to be more effective. And it led to a really interesting job that I like a lot. Good luck!

I got laid off last year, it was pretty rough. Took a couple months to work on some projects and my wedding which was nice. But then it took about 3-4 months of intensive job hunting.

Tons of ghosting companies, lowball offers, etc. Maybe it was just vibes from it being the mass layoff era, but it also felt like the interview processes at a lot of places became a lot more unpleasant than usual.

It's not a fun time to be looking. To anyone who's still on the pointy end of a layoff, good luck and don't get too down on yourself.

Linear is pretty great. They have a lot of more advanced features but the core interface is nice and clean. I like their project view with a nice graph of completed/remaining that gives a nice sense of whether the scope is still expanding or if you're on the other side of the hill and cruising towards complete.

Maybe? I don't use Wechat. But I think it's a good model if it is, though it would be nice if it were an open standard so it wasn't stuck in one app's ecosystem.

This might be a weird one but if you already know rust loco-rs is pretty great.

I've been using it on a nights/weekends project. It's axum + some rails-esque QOL features.

Because it's rust the time to make a given feature might be slightly slower (10-20%?) than in python/js, but I've found you gain that time back almost immediately from things just working nicely the first time. Also refactors/feature rework are super easy.

The issue is that this could prevent even opening up the discussion of whether a transaction was a bribe if the former president can claim it was an official act.

Nixon was never prosecuted because he was pardoned so the official acts stuff is moot. Crimes committed in war/putting down rebellion/etc are bad but more clearly within the scope of official acts than anything Trump is accused of.

Reagan, to the extent that he did/didn't coordinate with Iran to delay the hostage release, would be the closest parallel since it would have been a crime (Hatch act violation) in the direct pursuit of winning an election. But that was before he was president so this also wouldn't be relevant for today's ruling.

We don't prosecute ex presidents often because most of them either don't commit crimes or the ones they do are arguably part of the job. Trying to prevent the legitimate transfer of power is not close to a part of the job. The opposite in fact. Nixon only gets pardoned because he agrees to fuck off and spare the country the shitshow we currently have.

Basalt on the ocean floor records the magnetic field of the earth from when it cooled. Since it radiates out from midoceanic rifts and the rate of its creation is known, we can use it as a diary of the magnetic field. Iirc we first figured it out during WW2 doing magnetic bathyometry to help ships navigate.

I did a remote calc class and nerd sniped myself into learning latex via using overleaf to do homework instead of taking photos of handwritten stuff. 10/10, would self nerdsnipe again.

Not quite web dev but I've always thought there might be room for a browseresque class of software for mobile apart from the literal browser. Instead of every mom and pop shop having basically similar apps that everyone feels resentful of installing, just have a bunch of the shared functionalities bundled into a meta app. The app browser could handle things like identity/billing/notifications etc.

The mini-apps could be mostly a JSON object defining behaviors and then maybe some styling/assets. It could lower the cost of getting a proper mobile experience out for businesses as well as lower the friction for a user debating installing an app.

It seems a shame that the choice is between a responsive webapp and a really expensive (dev time wise) native app.

Do not taunt happy fun ball.

> 10 years ago, code was on average relatively clean and easy to understand

We have very different memories of 10 years ago. I just remember peak OOP cargo culting with 10 layers of inheritance and searching `Factory` would OOM your IDE.

So for context our app (https://nativi.sh) is a language correction app. It takes in text and cleans it up to make it sound more fluent/correct, it's basically geared towards being grammarly for your second language.

For some of our deterministic LLM tests, we have inputs that have known spelling errors but no wrong word errors, or some other combination of errors. If the config under test doesn't identify the issue, or identifies issues that we know aren't there then it's marked as being wrong for that test case. Then we can test across config x language x kind_of_error.

For the LLM vibe driven scoring we have it set up to just do a head to head between the current leading config (usually what's in prod) and the new candidate config rather than generating an abstract score. It will flag "x config straight up failed question N based on some_reason(s)" so that we can manually check it.

My partner wrote the testing framework. She's been thinking about cleaning it up and open sourcing it.

That's more or less the definition of AI app for the moment, no? If it's not using an LLM[0] for whatever magic trick it's offering then it's not really an AI app. And if it mediates input/output between the user and the LLM then it's an LLM wrapper.

Every software product is a wrapper of sorts and necessarily masks the some functionality of the ingredients it's made from in order to provide a productive interface.

[0] I know there are other kinds of AI but that's not what people mean when they say AI app now, for better or worse.

We have a script and a input library that has a bunch of scoring dimensions and allows a head to head comparison of a new candidate prompt vs what's in prod. It takes a configuration (prompt, which LLM to use, temperature etc.) and then gets run with all the various inputs and makes a json blob of the outputs for scoring.

Most of the score dimensions are deterministic but we've added some where we integrated an LLM to do the scoring (... which brings the new problem of scoring the scoring prompt!). We also do a manual scan of the outputs to sanity check. Not doing any fine tuning yet as we're getting pretty good results with just prompting.

Not super successful yet but I've been working on https://nativi.sh on nights/weekends. It's a tool to help make your writing more fluent when you're working in a non-native language.

We have a fair amount of users who've worked it into their daily professional and personal writing, which I'm pretty stoked about. I've pretty much only worked on B2B stuff professionally so it's fun making something that an individual gets utility/happiness out of.

Middle square is actually pretty good. Solid odds of the opponent giving you a layup with a bad break.

Not even sure these guys fit the standard mantle of language hyping charlatans. Most of the catastrophic choices were things you could do in any language. With the bonus of no meaningful comments and just committing to master so there's no PR comment history to help the future team's archeology project.

I work at a place that has ancestral Gooby code. It is truly horrific and an affront to whatever machine god (GPT-X?) may judge us in the future. The badness has literally nothing to do with the language. All the bad parts are due to the original team having reached for the most esoteric thing available at the time. I like playing with weird tech shit but there is a time and a place, and that place is not in prod for a startup.

They way overspent their strangeness budget[0] at every step of the stack. Even though I love (this) Gooby in general, in this case it has caused extreme damage to the organization.

I am still angry at the original devs for their choices because they basically poisoned the whole org against the language even though it would be useful for the org as a whole to adopt it in a non psychotic manner. It basically resulted in a reflexive ban for all Gooby in the future even when it might make sense.

[0] https://steveklabnik.com/writing/the-language-strangeness-bu...