Chat Logs

  1. ne555https://bpa.st/XETSU good morning, trying to debug a vaadin project. so run it with `./gradlew appRun` and then tried to «attach to process» from intellij, but there are authentication (¿?) issues
  2. cheeserne555: gemini's response: https://gist.github.com/evanchooly/fc7c0d56bad06dd0c3185a55923674df
  3. nevetgist:fc7c0d56bad06dd0c3185a55923674df
  4. dreamrealcheeser: how's gemini compared to claue and openai?
  5. cheeserlast i tried it for coding (~3 months ago) it thrashed for over half an hour and still couldn't generate code for the same prompt that took claude just a few minutes. and it was thrashing on junit code. it wasn't even that complicated.
  6. cheeseri think a newer model has come out since, though.
  7. cheeseri tend to use claude and copilot (because easy access on github)
  8. dreamrealI haven't found copilot esp useful - claude's gotten worse at containing context, codex is in the lead right now for me
  9. dreamrealthe main thing about ALL of them that I've found is that fire-and-forget instructions tend to, uh, go awry
  10. ChronosThey've all improved a bunch even in just the last 3-6 months.
  11. dreamrealYeah, Claude's been falling behind compared to codex, though, which is kinda surprising.
  12. cheeserime, they tend to leapfrog each other.
  13. cheeserbut i've wanted to get a codex sub and then dovetail the two together
  14. cheeseri just learned of an mcp that let's claude delegate to codex which would be handy
  15. dreamrealisn't that what openclaw's claim is?
  16. dreamrealI haven't used it, so I'm not clear on it
  17. cheesernot really, no. openclaw is for automating all the things.
  18. dreamrealokay
  19. Paraopenclaw is the AI powerup for curl|sh
  20. ChronosPara: haha
  21. dreamrealwell, that's justifiable, curl is madness to use! ... right? anyone?
  22. sbalmosutter insanity
  23. dreamrealnevet's THIRD UI is nearing workable state - and this one's probably the best out of them all, and probably will be the "real" front end
  24. dreamrealhopefully I'll be able to deploy it this afternoon, there's one slice of functionality I'm putting in now
  25. dreamreal(Interesting thing I did not expect: the butt-ugly reference UI has been *remarkably* useful for regression and implementation)
  26. sbalmosI demand an RSS Atom feed! :D
  27. dreamrealhas one
  28. dreamrealActually, I guess it's just RSS, not atom specifically
  29. DoofusCanadensisThe Atom format is different from RSS
  30. dreamrealyeah
  31. dreamrealI used rome to generate it, so it's RSS, could migrate to atom easily enough, I guess
  32. * dreamreal adds a line
  33. dreamrealI should have it updated in an hour, trying to round out some features
  34. dreamrealsbalmos: ++ damn it, adding atom created a whole slew of options to handle in tests
  35. nevetsbalmos now has karma of 1.
  36. sbalmosyou're welcome :D
  37. dreamrealIt's all good, I'm about to try to wire it up on the public site
  38. dreamrealit just takes a while to run the full suite before issuing the merge request
  39. dreamreal(it's about a 20 second build... but the tests take 3:17)
  40. ParaI managed to deploy my personal thing's backend today. How annoying, now I can't fiddle with the infra anymore and instead need to start building the actual thing :/
  41. dreamrealhaha
  42. dreamrealI still have a lot of infra improvements to make, but they're not actually affecting functionality: like, i want to route wordpress attack vectors into fail2ban
  43. Swayzelol i run all my things as background processes in the ai cli
  44. * Swayze remembers days when deployments were a thing
  45. Swayzevery retro
  46. Swayzekeepin it ol school
  47. * Swayze is jokin of course -- :)
  48. Swayzei bet in the 50s people watched compiler outputs the way we watch llm tv today
  49. dreamrealsbalmos: yo
  50. sbalmosdreamreal: urgh