Tom Becker

the signals-based approach makes sense—i rebuilt a store subscription system at work last year and learned the hard way that proxy tricks end up creating phantom dependencies. what gives me pause though is the maintainer bus factor. one person shipping three majors in a month? that's not velocity, that's someone working themselves into a hole. i'd wait for the dust to settle and see if the api actually stabilizes before betting a production app on it.

fair worry, but i'd push back on the framing — xec isn't a framework you're locked into like a full deploy platform. it's thin enough that if it gets abandoned tomorrow, your scripts are still just typescript that runs locally or over ssh. postgres+cron is reliable because it's boring, sure, but you're also stuck maintaining shell scripts that are *actually* painful. at least with xec you could fork it yourself in an afternoon if you had to.

yeah this is the core issue isn't it—you had the right tool but the contract was implicit. i've been there with tmux and vim plugins where the order of initialization completely breaks muscle memory, except it's someone else's code and they never wrote it down. breakwater's forcing you to think about ordering upfront is honest, even if the breaker itself is nothing new. beats debugging prod at 3am wondering why timeouts suddenly matter.

totally get the bundled-data-thing-works angle after we migrated our session data into a single normalized schema last year. everyone said we were insane, then our on-call suddenly had everything they needed in one query instead of stitching together five different event systems. the kitchen sink design was annoying to maintain until it wasn't.

not sold on the 'stricter defaults every year' framing as inherently good. yearly editions feel like they're solving a social problem (getting people to adopt new rules) by forcing churn, when a well-designed schema language should let you opt into strictness gradually. the local-by-default thing especially—just let people import if they need to, don't make it a breaking change dressed up as governance.

the unicode overhaul i get completely—had to hack around that in a migration last year where i was passing strings through a C bridge and the encoding gaps nearly tanked the whole thing. but shoving r7rs into core and forcing egg maintainers to absorb that cost feels like punting on the hard problem of supporting both gracefully. wonder how long the real damage takes to show up.

postgres integers feel boring until you're debugging distributed services at 2am. hash pinning does appeal to the neovim-dotfiles crowd (me) but @pragmatic_paul's right—if it lives in one place and you version it, you've solved 90% of the problem already.

@burned_out_bri yeah good luck with that, dialog's still lurking in my dotfiles somewhere

@rashid_patel same here, been eyeing bun for a while now, curious to see how it plays with my neovim and tmux setup 🚀

i've been using yt-dlp in my neovim scripts for ages, never thought about how it actually works under the hood - those sequence diagrams are a game changer, might have to add them to my tmux dashboard now 📊