Back to the channels

#java

11:32 dreamreal https://bytecode.news/posts/2026/04/how-would-you-avoid-ai
14:07 sbalmos dreamreal: by disappearing into a cave, rolling a stone in front, and returning in 3 years
14:23 dreamreal sbalmos: makes sense
15:27 johnjay you'd miss out on the 2x boost in salary from being fired and then begging for you to come back though
15:27 dreamreal We're still waiting for the influx of "please come back" unfortunately
15:28 cheeser the ai models haven't been trained on such interactions and so it's taking a bit of time to figure out how that works.
15:35 kcomhnall cheeser: are you talking about the LiteLLM chain attack or something else?
15:36 dreamreal ~dynamo
15:36 javabot DynamoDB is a NoSQL database engine created by Amazon. It supports key/value and document models. It's very (VERY) scalable, but it comes at a cost: there is no guarantee WHEN things will happen, but THAT they will happen. See https://aws.amazon.com/dynamodb/
16:08 sbalmos hence why I'll wait about 3 years. there won't be a "please come back". it'll be more like "please, we need someone to fix all this and no one actually knows what it's doing when it was AI-written"
16:23 dreamreal sbalmos: ++
16:23 nevet sbalmos now has karma of 2.
16:24 Para It's a bit of a high stakes play, but indeed that's a very likely end result. Big question is if there's going to be any money left to fix the situation.
16:25 dreamreal We'll just ask the AIs to do it </blinder>
16:25 sbalmos that's a problem for the company
16:25 Para Well, we could of course ask AI to create succesful cryptocoin ICO startup thing.
16:25 Para (remember to add "succesful" to the prompt!)
16:26 sbalmos I am free to charge whatever I want to feel is fair to come in and attempt to fix what is broken
16:26 sbalmos the company is free to deem the fixes critical enough to accept my offer
16:28 johnjay i'm looking for some small java projects to study that aren't toy projects. any ideas? debian tells me i can use runescape or treeview or svngit which are written in java.
16:29 Para I guess for a kernel dev anything below millions lines is small :D
16:30 Para I don't know what to point to, though. Library landscape probably has something interesting and not too complex, Google Guava, Apache Commons things, that sort of things.
16:37 johnjay that's a good point. and i dont' think github lets me sort by LoC does it?
16:37 johnjay i'm not logged in on this machine so I don't recall.
16:38 Para Probably not. Although Java isn't very LOC oriented, or rather that the number would be that meaningful.
16:39 Para Would be kinda interesting side project to create quality metrics for every single JAR in the Maven index and produce public stats from those for ranking through everything.
16:40 sbalmos define "quality"
16:41 Para Cyclomatic complexity, coverage, coupling, cohesion...those kinds of things.
16:42 Para Hm. Lots of c:s in quality metric names.
16:47 johnjay i was thinking about coupling i guess the other day. do you want lots of modules tightly copuled as the ideal system? or is the tradeoff more modules means less coupling?
16:48 Para I've forgotten the name of metric which counts call depth.
16:52 johnjay i meant more like how should you design one layer to interface with another layer
16:52 johnjay i think coupling is part of that
17:10 dreamreal this is not a new question
17:10 dreamreal see bob martin, hexagonal architecture, etc
17:20 Tenchi Para: the name of the metric which counts call depth is "call depth" :P
19:07 NeXeN hexagonal architecture is a good name for an EDM group
19:17 * NeXeN ponders if he should stick with lit or switch to hilla for a vaadin project