Coyote Ugly

One last ai boost, sorry. 🤪 I've found that treating ai contributions as if they came from another person is really helpful- just because I vibe coded or AI assisted something doesn't mean it is exempt from my normal review process. If I give the agent's PR the same scrutiny and feedback I do one of my junior devs, the amount of slop decreases by an order of magnitude. Whether it helps my juniors write less slop, or causes my seniors to produce more slop, that says much more about the person than the AI.

One sad thing about all this politicizing of AI is that so many people put themselves in the "AI good" or "AI bad" camps, when it's always both- Good for what? Bad for what? It's not blanket one or the other. (I really do appreciate JBs nuance here- y'all are some of the few that don't just pulpit pound from one side or the other) Brandon Sanderson has a really good keynote called "We Are The Art," that hits this really well- if the intended goal is the output of a thing, AI can be really great. If the goal is what developing the thing creates in you, not so much. Sometimes we need to be the art for it to be properly appreciated and trusted. Other times we just want a cool thing. Like Chris's music- I can't get enough of that. (We seriously need an album of these- I still play "Chris is in love with another man's van" on the reg) And other music I'd pay for, I'm actually paying for their expertise and lifetime experience being a creator, not necessarily the song itself.

Mystery t9 boost 😁 All this AI blocking reminds me of the yearly hacktoberfest drama ratcheted to 11. We've relied so long on a base level of quality being baked in by virtue of the learning/effort curve being high enough that people wanting to contribute are also putting that kind of time into their code, and thus more likely to have better code. (Always a terrible assumption IMHO) Offer a bunch of people free T-shirts or the ability to outsource to AI, then the slop increases. (whether that's outsourcing to an LLM or Anonymous Indians, it kinda feels similar tbh) Blocking hacktoberfest or AI altogether is just throwing out the baby with the bathwater in my opinion. Enough people are being bad to the maintainers that we can't have nice things I guess. 🤪

Vaultwarden user over here- it's the only one I've found that has the wife approval factor while also allowing us to share creds with both each other and family/friends. (Also handy for securing and time/access limiting random stuff like when I needed to send my boss a scan of my passport) I also used it to force myself to learn and practice proper backup and HA hygiene. Been running 6 years with 0 fire drills and four 9s. If bitwarden proper goes south enough to make vaultwarden not feasible anymore I might consider a proton, but I'd probably be more likely to maintain my own vaultwarden-compatible client than store my passwords on somebody else's hardware.

Question for you guys (and anyone in the community that wants to boost in)- do you have any preferred methods of dealing with burnout? I used to get rejuvenated by doing something weird with my homelab, joining in on a JB challenge, going for a hike, or letting my adhd pick a random new hobby for me. But nowadays everything seems to just make it worse. Would love to hear strategies you guys use! (Especially Chris since he's a fellow adhd-er and it seems like we have a lot of similar stressors, at least based on what he has shared on air)

I'm in a similar boat to PJ- as a general rule, I don't fully integrate anything into my core/day to day production until I can own it and run it on my own hardware. (Or vps I control) Probably have been listening to Chris's defined networking ad reads for too long. 😆 But, that being said, one thing that is handy is that the openai api spec is quickly turning into S3 of the AI world. So in the interim, even with the cloud models, I can run them through a proxy that exposes them to my systems as a generic openai compatible model, and to my systems it doesn't matter where the model resides, and once I get a local model that performs well enough I can just swap it out and start integrating it more centrally into my production environment. Doesn't mean I won't play with it and learn it until then to be something I CAN rely on for production use. It just won't be something I WILL rely on until that point. (all caps just for emphasis since fountain doesn't support italics. Not yelling or anything)

Does anyone else have the problem of the boost audio clips slipping into your day to day vernacular? I've accidentally dropped, "they're doing a lot with mayo these days" and "that is a tasty burger" in several conversations without thinking about it and I just called my coworker a rich lobster. 😆 I guess it's just a sign of how integrated JB is in my life. 😁

One thing that gets lost in the license conversation is that as it currently stands, LLMs are just Large License Launderers. One of the great things about copyleft FOSS is that you know that all derivations will continue to benefit humanity by also being FOSS. But as it is now we lose all freedoms as soon as someone makes a proprietary derivation using an agent. "Roll over and take it because they're going to violate your code and the freedoms of your users anyway" isn't a very good stance, and neither is "put up a bunch of useless walls that also violate your users freedoms" 🤪 it just feels silly that the two leading arguments on this subject break foss ethos in different ways. I'd think that if we can get both sides to come to an agreement on an interface to let AI access the code and also respect the license, we could get the best of both worlds. Even if it's just a setting you can give your agent to include or exclude source material that is compatible with the license of your project or something. If you're also writing gpl, go ham! If you're writing proprietary, then you can consciously make the choice on whether or not you'll abide the license. Another benefit to something like that would be that folks at companies that block copyleft deps can actually enjoy all the fun AI stuff we have without worry of getting in trouble. 😁 Of course all of that is easier said than done.

(PS: One tip you may've already thought of- if you are able to plan out your route beforehand, installing/registering all the apps for chargers along your route before you leave makes things WAY easier. Especially for the plug 'n runs. Then all you've got to do is tap/enter stall id for the not as cool ones)

I'd say keep the Tesla chat coming! Always fun to see what's going on on that side of the fence. Totally hear you re chargers. I've gotten to the point where I assume the charger won't work if it's not EvGo, EA, Tesla, or has a perfect score on plugshare. EvGo is currently my favorite- they support plug and run on both Non-Tesla and Tesla. Super clean experience. And they are rolling out NACS!

Just wanted to share a little Linux and self hosted success story- thanks to what I've learned from the JB crew and community over the last 6 ish years, we were able to migrate off of GCP to a combo of our own infra and a colo, resulting in a monthly savings of the boost amount, but if it was in dollars instead of sats. And all FOSS. 😁 You guys are awesome, thank you! Also, regarding the on-site you guys did at Hadea's clinic- it was super interesting. It's always those small gotchas that you expect to be quick that end up eating up your entire time. 😬 I've experienced similar with one of our sub-companies that specializes in Linux-based home automation for short term rentals. (we call it RoboSTR) On our last install the client wanted to keep their existing chinesium cameras and APs, but plugged into their new HA and router/DVR boxes we installed- everything was going peachy until one of the 4 APs decided that it was going to max out at 4Mbps. Fast forward 6 hours and we ended up just re flashing all of them with openwrt. After that the rest of the networking and camera setup took like 15-20 minutes. Should have done that sooner. 🤪 If you guys end up wanting to do this again for the airbnbs on the way down to SCLE, we'd love to assist if you would find it useful! (At least remotely- we don't have the travel budget to come this year 🫠😞)

Everytime I hear the break song, it scratches my show tune itch... But then replaces it with another itch of needing JB: the musical. You know, like those old sitcoms that would have a random musical episode.

Guys... You just blew my mind with Nostr. I played around with it back when you first started talking about it, but I don't think I ever really understood what it actually was. I just thought it was activitypub with keys... But I was missing the point. Dunno if I'll use buzz long-term, but I'm setting it up now! And if nothing else this was my eye opening catalyst to connect all the dots. Thank you!

On the audible topic- I kept forgetting to set up my cron jobs years ago, so I was lazy and set up libation and mounted my audiobookshelf drive. It auto downloads and de-drms, follows my naming/folder conventions, and they show up in audiobookshelf within minutes of them showing up in audible. Especially handy for friends and family that want to contribute to audiobookshelf- just add their account to libation and done.