Chris Schneider

@lorigrego @SouthwestAir Thunderstorms in Denver at the airport apparently. Stuck in Vegas for the same reason.

Hey @SouthwestAir - what’s going on? Sitting at BWI airport- weather is fine - and my flight to Denver was cancelled. Did you rebook me to fill later flights? Not OK. Your BWI staff was lovely, but it’s not their fault. Tell me why I should stay loyal?

@CasJam I do, but mostly because I still like cursor's autocomplete. cmd-k is nice still too for quick edits. But claude code seems better at doing back & forth with bigger agent requests. For an extra $20, cursor isn't that big a deal.

Does anyone use Claude Code inside of Cursor? (if so, why?) Or is it more useful if you're not a Cursor user?

@nateberkopec I have an Unraid server, in a Fractal Define case (excellent). Don’t worry about SATA port count, PCI expansion cards for SAS can easily add 8-16 extra ‘sata’ ports with little effort Airflow over spinning disks is certainly important, but decent case fans are pretty quiet

Anyone ever built their own NAS? Seems kinda fun. My main concern is noise, but it feels like if you BYO then you can bling out the fans and not have to worry too much.

@nateberkopec are there any hosting companies even close in the US? And what downsides for a US company using an EU hosted server, that’s + some milliseconds on response time due to distance I guess?

If you're starting a new Rails app and live in the EU, the optimal strategy is pretty clear: Get a 64-thread box on Hetzner for 250 euro a month, allocate 20% of it to the DB, 40% to web, 40% to bg, then just forget about scaling until you're making money hand over fist.

@allanbranch if you need to put her down, have somebody come to the house. It is so much better than a strange place. Other piece of advice: don't wait until it's an emergency. going a day early is much kinder than a day late.

@xgigglypuff Senior dev here. I don’t find them all that useful, I have copilot via work and it’s just a slightly better autocomplete. It’s not ‘ai’ in any meaningful way. At least not the way interacting w/ ChatGPT or whatever is.

I still haven’t used chatGPT or GitHub Copilot or any form of AI assistance for programming. Am I just too much of an old school personality? Anyone else out there? 😂 Should I experiment with it on stream?

@ass_beaters ah shit, what was that video game called. Genesis era? Punks roller blading and grabbing on the back of cars.

@patio11 Just played some with the free version of aidungeon, quite impressive. There must be a ton of prompting work behind the scenes to make this work smoothly. I’ve used SillyTavern, and never got this quality of responses. But perhaps it’s just a better setting card.

Annals of “nobody expected that” AI capabilities: I have a brother who publishes RPG modules as a semi-commercial hobby. (He did well on a consulting business; long story.) Apparently if you flip the script on AI Dungeon it will happily let you be the DM and it will be party.

@_swanson Specifically about counts like your example

@_swanson An excellent essay on just how hard it is to i18n correctly across languages. I refer back and share this every year or two. perldoc.perl.org/Locale::Makete…

Pluralization is a basic concept in i18n that can be handled without much extra work In the Rails i18n gem: ``` en: ownership_message: one: "%{count} day as a Proud Owner" other: "%{count} days as a Proud Owner" ``` and you call it using `I18n.t` (or just `t` alias): ``` t('ownership_message', count: 2) > 2 days as a Proud Owner ``` In fact, using localization tools can sometimes even save you from an extra if/else in your code. This approach is better than simple `count > 1 ? "days" : "day"` because it lets you move the count to the correct "spot" in the sentence when localizing.

@ioquatix Immersed is the software or something else?

@ioquatix Which headset? I tried quest 2 and it just didn't have resolution to be usable

Okay, so I set up Immersed, anyone want to come and do some hacking in VR with me?

@nonzenz @patio11 My guess would be a western union or equivalent at a regular retail store

@patio11 I asked there but I’ll ask again: what’s the actual cash out method for scammers? He says they directed him to a “Federal Payment Processing center” - but surely no such place exists. I just don’t understand the last mile.

Interesting narrative of a scam. Anyhow, inform vulnerable people in your life, and yourself, that the government will almost never do anything consequential by telephone.

@mateosfo Note that i agree it's bullshit that it happens. But wondered what the situation that causes this in particular.

@mateosfo Is this the result of curb cuts being part of a road project budget and sidewalk from some other spot. So the do the sidewalk stub to match some rule during road construction and just never get to building an actual sidewalk?

The thing you have to remember about traffic engineering in the United States is, someone with an actual civil engineering license stamped this design and spent taxpayer dollars to build it. These dead-end, useless sidewalks and are not accidents. They are malicious.

@nothings @eggkopf strong agree. This is the right way to do things. Business programmer here, but any heuristic of 'too large' is a poor proxy for what's actually a problem: 'too complex'. "How does xyz joker work" is one quick in-file find away. That's great!

@Rockettheband1 Enjoyed the show last night!

@tuple @BetterStackHQ Oh, we're just moving over to Betterstack - do you have any tips / advice?

We've been working hard to level up our observability, and the logs-to-metrics functionality from @BetterStackHQ has been making that a lot easier 🙌

@_swanson @nateberkopec Yeah, I was at ScoutAPM for a long time, and this was a very common request. Thinking about it now, I’d build it with Flipper, to check if a given client should be traced. By default whatever % chance, then high value, or complaining customers get put into the always-trace bucket

@watchchrislearn @nateberkopec yeah this is good -- we are spamming our APM with a bunch of background jobs that we don't really care about performance on

@nateberkopec any good rules of thumb for performance sampling percentage in APMs? 10%? 1%? 0.1%?