| Time |
Nick |
Message |
| 04:00 |
|
MTDiscord joined #luanti-dev |
| 07:18 |
|
pmp-p joined #luanti-dev |
| 09:43 |
[MatrxMT]_ |
<Zughy> #17347 this is the second vibecoded PR OP has opened in 24h. Can we just ban them? |
| 09:43 |
ShadowBot |
https://github.com/luanti-org/luanti/issues/17347 -- Clarify palette pixel repetition by coyaSONG |
| 09:52 |
sfan5 |
well we don't have a rule against that yet |
| 09:56 |
MTDiscord |
<.zenonseth> Counterpoint: that PR you linked Zughy is +5 -3 and having now read the clarifiation it's useful, imho |
| 10:00 |
MTDiscord |
<.zenonseth> my question would be: what do you want to guard against? bad code + frequent huge 'slop' code? or just anything LLM coded in general? The former can be automated to some degree, re-linking this I posted in the engine channel: https://github.com/peakoss/anti-slop The latter may result in some people just outright lying about llm usage anyway. |
| 10:24 |
[MatrxMT]_ |
<Zughy> sfan5: last time I checked, you had banned an account for the same reason |
| 10:26 |
[MatrxMT]_ |
<Zughy> Counterpoint: where is the line of "this is small and innocent enough that an LLM can handle it"? |
| 10:40 |
sfan5 |
I banned an account presumably used by an "agent" to make undisclosed AI contributions to collect bounties |
| 10:58 |
MTDiscord |
<.zenonseth> > Counterpoint: where is the line I don't believe there is an objective and clear line. But that doesn't mean that the extreme cases aren't obvious, nor does it mean that it's not a useful distinction to make on the more obvious examples. Similar to the "how many grains of sand make a heap?" paradox |
| 11:01 |
[MatrxMT]_ |
<Zughy> fair enough. Change of perspective: should an user should be credited as a contributor when what they did was writing a prompt so that a machine could make the work? Isn't this the user agent case, just without automation? |
| 11:01 |
[MatrxMT]_ |
<Zughy> too many shoulds, whatever |
| 16:45 |
|
Eragon joined #luanti-dev |
| 18:51 |
|
ZenonSeth joined #luanti-dev |
| 19:39 |
Krock |
Will pushy push #17325 in 15 minutes |
| 19:39 |
ShadowBot |
https://github.com/luanti-org/luanti/issues/17325 -- [no squash] Fixes related to node waving by SmallJoker |
| 19:40 |
Krock |
#17340 too |
| 19:40 |
ShadowBot |
https://github.com/luanti-org/luanti/issues/17340 -- Main menu: Make render info text copyable (closes #16953) by siliconsniffer |
| 19:45 |
|
calculon joined #luanti-dev |
| 19:56 |
Krock |
done |
| 20:13 |
ZenonSeth |
sfan5: I think you were talking about supporting lines of various thickness for any new line api before, I just pushed changes to support a hex tube, on top of the pixel-line primitive, for the lines api PR #17335 |
| 20:13 |
ShadowBot |
https://github.com/luanti-org/luanti/issues/17335 -- Add new Line SAO/CAO and lua API + parabola helper by ZenonSeth |
| 20:25 |
|
siliconsniffer joined #luanti-dev |
| 20:26 |
siliconsniffer |
Krock: thanks for the merge! |
| 20:27 |
siliconsniffer |
I assume the overlapping element is fine then? :-) |
| 22:24 |
MTDiscord |
<luatic> it is occurring to me that maybe my suggestion of vendoring a fallback libzip similar to how we do it for jsoncpp might have been stupid (or "ill advised", as they say in britain) |
| 22:24 |
MTDiscord |
<luatic> we can probably get away with just demanding it be taken care of by the system, more similar to libpng |
| 22:24 |
MTDiscord |
<luatic> what do others think? |
| 22:32 |
|
panwolfram joined #luanti-dev |
| 22:36 |
rubenwardy |
I was thinking this - we should use the system one |
