| Time |
Nick |
Message |
| 02:30 |
|
YuGiOhJCJ joined #luanti-dev |
| 03:41 |
|
SFENCE_arch joined #luanti-dev |
| 04:00 |
|
MTDiscord joined #luanti-dev |
| 05:31 |
|
v-rob joined #luanti-dev |
| 06:07 |
|
v-rob joined #luanti-dev |
| 06:33 |
|
v-rob joined #luanti-dev |
| 06:40 |
|
SFENCE_arch joined #luanti-dev |
| 08:42 |
pgimeno |
luatic: it would be nice to add a comment explaining what test_rotation_radians actually tests, which is invariance of angles over scaling |
| 08:43 |
pgimeno |
I realized that after writing the patch |
| 10:00 |
|
rubenwardy joined #luanti-dev |
| 10:29 |
rubenwardy |
just got chatgpt to write a script for downloading and signing android artifacts, wish me luck |
| 10:30 |
rubenwardy |
it's also told me off for still using hub over gh which is fair enough |
| 11:07 |
rubenwardy |
Done |
| 11:07 |
rubenwardy |
@wsor4035 ^ |
| 11:28 |
celeron55 |
last year 50 line programs could be considered single use as you could generate one in one go using an LLM. this year you can pretty much generate a 500 line program in one go. this could mean in 2027 you could create almost a luanti sized single use project in one go. are you guys even remotely mentally prepared for this? |
| 11:29 |
sfan5 |
i look forward to a buggy luanti clone |
| 11:37 |
celeron55 |
the testing feedback loop will definitely be the bottleneck |
| 11:40 |
rubenwardy |
It generated the wrong commands but reading the gh help allowed me to fix them |
| 11:40 |
rubenwardy |
Basically was a pseducode generator |
| 11:43 |
celeron55 |
maybe it's more like 20 lines last year, 200 lines this year |
| 11:47 |
rubenwardy |
#16387 |
| 11:47 |
ShadowBot |
https://github.com/luanti-org/luanti/issues/16387 -- Add script to download and sign Android builds from GitHub Actions by rubenwardy |
| 11:51 |
rubenwardy |
gpt comfirmed hipster |
| 12:27 |
MTDiscord |
<wsor4035> https://github.com/luanti-org/luanti-org.github.io/pull/355 seeing if ci passes. anyone who wants to take a peak in the next few minutes |
| 12:29 |
MTDiscord |
<wsor4035> lets say merging in ~5 minutes |
| 12:30 |
rubenwardy |
Lgtm, assuming all links work |
| 12:30 |
rubenwardy |
Luckily ci checks that |
| 12:30 |
MTDiscord |
<wsor4035> exactly :p |
| 12:49 |
|
SFENCE joined #luanti-dev |
| 12:50 |
|
SFENCE joined #luanti-dev |
| 12:51 |
|
SFENCE_arch joined #luanti-dev |
| 13:17 |
|
SFENCE joined #luanti-dev |
| 13:17 |
|
Desour joined #luanti-dev |
| 14:16 |
|
SFENCE joined #luanti-dev |
| 14:33 |
|
BuckarooBanzai joined #luanti-dev |
| 14:34 |
|
SFENCE joined #luanti-dev |
| 15:32 |
|
nore joined #luanti-dev |
| 15:58 |
|
Desour joined #luanti-dev |
| 16:02 |
jonadab |
As any programmer can tell you, _writing_ the first draft of the code is the easy part. |
| 16:04 |
|
v-rob joined #luanti-dev |
| 16:05 |
celeron55 |
for sure. but that part is getting even easier as we speak. i guess there's hoping people will do the right thing and just delete that kind of code before it completely fills the world |
| 16:33 |
[MatrxMT] |
<birdlover32767> it would probably level off at 5000 or so, but imagine making an entire mod/game in a few moments |
| 16:54 |
|
SFENCE joined #luanti-dev |
| 17:01 |
celeron55 |
i don't think it will level off. but complexity grows exponentially and it's a very big unknown how fast they can extend the capabilities of the AImodels |
| 17:01 |
celeron55 |
+space |
| 17:02 |
celeron55 |
my point is, the observable rate has been very fast so far |
| 17:12 |
|
Desour joined #luanti-dev |
| 17:19 |
|
SFENCE joined #luanti-dev |
| 17:46 |
|
SFENCE joined #luanti-dev |
| 17:48 |
|
fluxionary joined #luanti-dev |
| 18:00 |
|
Juri joined #luanti-dev |
| 18:00 |
|
Niklp joined #luanti-dev |
| 20:43 |
|
SFENCE joined #luanti-dev |
| 20:47 |
SFENCE |
Only 500 lines in one go? Did you meet Clade Sonnet 4? |
| 21:05 |
MTDiscord |
<luatic> my little brother has been struggling with exporting b3ds, so he asked an AI (i think it was claude) to write a b3d exporter for blender. |
| 21:05 |
MTDiscord |
<luatic> i will say, i am impressed that it can generate mostly plausibly looking code, but even at first glance you can spot many issues, and unsurprisingly the resulting code did not work (and there's no way i will debug it). |
| 21:39 |
SFENCE |
Interesting. I have better experiences. I used it for OpencCL Api wrapper and it did with it in a few days, what normaly probaly takes me weeks... including test and benchmark. I guest, in 50% working code at first try. |
| 21:58 |
|
SFENCE joined #luanti-dev |
| 21:59 |
|
SFENCE joined #luanti-dev |
| 22:06 |
|
SFENCE joined #luanti-dev |
| 22:33 |
|
panwolfram joined #luanti-dev |
| 22:44 |
luatic |
pushed pgimeno's commit, also added a tiny comment as suggested by him. |
| 23:05 |
|
Eragon joined #luanti-dev |
| 23:07 |
|
alias joined #luanti-dev |
| 23:08 |
[MatrxMT] |
<Zughy> The other day I wanted to make a custom translation script for a mod of mine, but my python skills are basically zero. I "vibe coded" for not even an hour and I ended up with a quite complex script doing what I needed to. That was scary |
| 23:30 |
|
pmp-p joined #luanti-dev |
