00:16 <+bridge_> <efetee.> it was only for ppl used discord support i guess ?
00:20 <+bridge_> <teero777> https://cdn.discordapp.com/attachments/293493549758939136/1470560107380080793/image.png?ex=698bbd53&is=698a6bd3&hm=50be4283c6adffd0ef1d12258750c95109e2dff8797b6be20011ed8773c90432&
00:22 <+bridge_> <aegisub> fwiw that was an external vendor that they only used for under-13 ban appeals and that they have since stopped using; they'll use better methods for the new age verification
00:22 <+bridge_> <aegisub> not that I like either discord or their new age verification policies but this is still important context
00:23 <+bridge_> <12944qwerty> They may be trying, but even with the discord as support leak they were supposed to delete all info when it was closed, but never did so
00:23 <+bridge_> <12944qwerty> They may be trying, but even with the discord support leak they were supposed to delete all info when it was closed, but never did so
02:44 <+bridge_> <pilonpl> At least it is a ban for adult content only
02:47 <+bridge_> <pilonpl> In the UK they started banning random stuff like Spotify
02:47 <+bridge_> <pilonpl> Or Wikipedia
02:55 <+bridge_> <12944qwerty> That's crazy
02:55 <+bridge_> <12944qwerty> Only to think they were all for freedom of content not long ago
02:56 <+bridge_> <12944qwerty> crazy to think that they were all for freedom of content not long ago
02:56 <+bridge_> <12944qwerty> crazy that they were all for freedom of content not long ago
03:13 <+bridge_> <12944qwerty> I'm more concerned with the fact that they're gonna use an ai model to "guess" your age by your activity...
03:13 <+bridge_> <12944qwerty> Which has already backfired for YouTube and Roblox and possibly others. And it's a terrible invasion of privacy
08:57 <+bridge_> <Assa> I am lost, where do I find the map submission bot code again? I even wrote a PR a year ago and can't find the repo
09:42 <+bridge_> <ryozuki> https://www.theverge.com/tech/875309/discord-age-verification-global-roll-out
10:19 <+bridge_> <Assa> I wonder if they take fucking account age into consideration
10:22 <+bridge_> <Assa> my dc account is only 10 years old, so I must at least be 11 right?
11:47 <+bridge_> <Assa> I now tested https://opencode.ai/ and we are cooked. I asked it to fix my webdav mounts and it just worked.
11:48 <+bridge_> <Assa> AI in terminal :justatest:
12:32 <+bridge_> <remakepower> I basicly finished the work of server-side custom resource. Would it be accepted by DDNet :3? If so, I could port it from our teeworlds fork to DDNet. Anyway, there's a lot of thing to discuss. :3
12:33 <+bridge_> <remakepower> Also #7296
12:33 <+bridge_> <DDNet> https://github.com/ddnet/ddnet/issues/7296
12:37 <+bridge_> <heinrich5991> have you read the thread? it's not that long
12:37 <+bridge_> <heinrich5991> are the server side custom resources embedded in the map?
12:37 <+bridge_> <heinrich5991> you could give a technical overview about your implementation in that thread
12:37 <+bridge_> <remakepower> I have read a long time ago
12:38 <+bridge_> <remakepower> I still disagree with every custom resource must be embedded in the map though.
12:39 <+bridge_> <heinrich5991> would be good to give your thoughts in the thread
12:39 <+bridge_> <heinrich5991> I don't see you making that point in the thread yet
12:42 <+bridge_> <learath2> This is how you get those “Post-Mortem: Claude deleted my entire production database” posts
12:44 <+bridge_> <aegisub> You're absolutely right! I
12:44 <+bridge_> <aegisub> - Deleted your entire database.
12:44 <+bridge_> <aegisub> - Did not ask you for confirmation.
12:44 <+bridge_> <aegisub> - Did not make any backups.
12:44 <+bridge_> <aegisub> 
12:44 <+bridge_> <aegisub> I should have
12:44 <+bridge_> <aegisub> - Checked my command before running it
12:44 <+bridge_> <aegisub> - Made backup copies before running dangerous commands
12:44 <+bridge_> <aegisub> 
12:44 <+bridge_> <aegisub> Would you like me to make a detailed damage report?
12:47 <+bridge_> <heinrich5991> I'm deeply sorry, that shouldn't have happened.
12:53 <+bridge_> <Assa> the first step is to run the AI as a user, that doesn't have that many permissions
12:53 <+bridge_> <Assa> also you can run it in "suggestion mode" so you need to enter the commands yourself
12:54 <+bridge_> <learath2> Then you get a little tired, it seems to not make mistakes anyway, the command looks reasonable
12:54 <+bridge_> <learath2> Boom
12:55 <+bridge_> <Assa> But then it's *YOUR* fault and not the AIs :pepeRage: 
12:55 <+bridge_> <Assa> 
12:55 <+bridge_> <Assa> Also people seem to delete entire wikis without AI ...
12:57 <+bridge_> <heinrich5991> I tried that. it asked me whether I'd allow sqlite3 commands in the current directory. how would I evaluate that?
12:57 <+bridge_> <heinrich5991> sqlite3 surely contains a privilege escalation (in that scenario), like arbitrary file write in the whole filesystem
12:58 <+bridge_> <Assa> you could ask for the exact command it wants to do and evaluate it - I know this is a lot of work if you do this properly
12:58 <+bridge_> <heinrich5991> yes, that's what I did
12:58 <+bridge_> <heinrich5991> but I don't think that's what people do
12:59 <+bridge_> <aegisub> manual command approval ruins the vibes
12:59 <+bridge_> <Assa> I agree with all of you tbf - this is totally a double edged sword and you can damage yourself with it
12:59 <+bridge_> <aegisub> you're supposed to write an apple script that hits enter every two seconds and leave claude running
12:59 <+bridge_> <aegisub> ||this is not a joke; there was a blog post where someone actually did this||
13:00 <+bridge_> <heinrich5991> > Also people seem to delete entire wikis without AI ...
13:00 <+bridge_> <heinrich5991> we probably fixed that now. https://archive.org/search?query=subject%3A"Miraheze"+subject%3A"DDNet"&sort=-addeddate
13:00 <+bridge_> <heinrich5991> although I haven't tested that backup yet…
13:00 <+bridge_> <learath2> Overtime people lose the skill to evaluate it because they offload the task to the ai gods
13:01 <+bridge_> <Assa> it has 165 MB, looks _reasonable_ without downloading it
13:01 <+bridge_> <heinrich5991> the old wiki backups also had 165 MB
13:01 <+bridge_> <heinrich5991> of the deleted wiki
13:01 <+bridge_> <heinrich5991> they were missing the page texts though
13:01 <+bridge_> <heinrich5991> only had the images
13:02 <+bridge_> <Assa> I'll ask the AI questions until I fully understand what it's doing - but I am also not vibe coding
13:03 <+bridge_> <learath2> I ask it to search for tools for me
13:03 <+bridge_> <learath2> It found `witr` for me recently, quite useful actually
13:03 <+bridge_> <aegisub> this is the way
13:04 <+bridge_> <aegisub> when you have a question that needs enough context to be tricky to google, describe your use case and ask it for tools/conventions/idioms that apply here, and then look them up yourself
13:05 <+bridge_> <aegisub> oh that looks neat
13:05 <+bridge_> <learath2> Another one I do is asking it to find me documents. It does dig up some very interesting literature, just don’t ask it to read the pdfs it finds, it will misquote them
13:06 <+bridge_> <learath2> Like I needed to find a bitsliced implementation of the PRESENT algorithm
13:06 <+bridge_> <aegisub> a while ago an LLM gave me ddrescue which was great
13:06 <+bridge_> <aegisub> but in the same breath it told me that ddrescue and dd_rescue are the same thing which is bullshit
13:08 <+bridge_> <learath2> Oh yeah, it loves doing that. I was looking for an alternative to GNU Parallel and it kept telling me xargs can do it, by just passing flags that belong to parallel to xargs
13:51 <+bridge_> <remakepower> Also if anyone is interested in this, you can view https://github.com/TeeworldsArchive/teeworlds/pull/46 😄
13:58 <+bridge_> <avolicious> This could happen to a human as well
13:58 <+bridge_> <learath2> At least then you are a regular moron
14:00 <+bridge_> <avolicious> The same if you just allow any command to be executed from any LLM agentic driven tooling
14:00 <+bridge_> <heinrich5991> this seems to be the modus operandi of a lot of vocal people on the internet
14:01 <+bridge_> <avolicious> But this was the same before LLM where here for the mass
14:01 <+bridge_> <avolicious> But nobody wanted to get blamed on X (formerly X) for deleting prod after executing a stackoverflow command
14:02 <+bridge_> <avolicious> Now you just blame LLMs that its their fault
14:03 <+bridge_> <heinrich5991> I don't think you can take that from the first post
14:03 <+bridge_> <heinrich5991> > This is how you get those “Post-Mortem: Claude deleted my entire production database” posts
14:03 <+bridge_> <avolicious> https://axcient.com/blog/how-google-cloud-deleted-a-125-billion-account/
14:04 <+bridge_> <heinrich5991> it seems we're in agreement? I fail to find any difference in opinion
14:05 <+bridge_> <avolicious> I would rename it to: "Post-Mortem: How I deleted my entire production database"
14:06 <+bridge_> <heinrich5991> but these blog posts exists
14:07 <+bridge_> <heinrich5991> you can't rename them for the people who write them
14:08 <+bridge_> <heinrich5991> see e.g. https://old.reddit.com/r/ClaudeAI/comments/1pgxckk/claude_cli_deleted_my_entire_home_directory_wiped/
14:09 <+bridge_> <avolicious> But its the same here, human error
14:09 <+bridge_> <avolicious> He just allowed the rm -rf command to be executed
14:09 <+bridge_> <heinrich5991> yes, the blog post is called "claude cli deleted my entire home directory"
14:09 <+bridge_> <heinrich5991> so I think a "post-mortem: claude deleted my entire production database" is also likely
14:10 <+bridge_> <avolicious> Its just the tool
14:10 <+bridge_> <avolicious> You could also say: "rm -rf deleted my entire home directory"
14:10 <+bridge_> <avolicious> Which is intended
14:10 <+bridge_> <heinrich5991> @avolicious would like to blame someone else, but agrees that the blog post is likely to be called 'This is how you get those “Post-Mortem: Claude deleted my entire production database” posts'?
14:10 <+bridge_> <heinrich5991> @avolicious would like to blame someone else, but agrees that the blog post is likely to be called "Post-Mortem: Claude deleted my entire production database"?
14:11 <+bridge_> <avolicious> "I deleted"*
14:11 <+bridge_> <avolicious> Not claude
14:11 <+bridge_> <heinrich5991> you don't believe someone would write a blog post with that title?
14:11 <+bridge_> <heinrich5991> I guess we're really disagreeing then
14:11 <+bridge_> <avolicious> I do, because obv. they are using "claude" already
14:11 <+bridge_> <avolicious> But blaming an LLM for their fault is not correct
14:12 <+bridge_> <heinrich5991> that seems to be a discussion you've opened
14:12 <+bridge_> <heinrich5991> I'm not really interested in discussing who's to blame, I just wanted to understand where your difference comes from
14:13 <+bridge_> <heinrich5991> you apparently don't want to cmoment on whether you think such a blog post title is likely
14:13 <+bridge_> <heinrich5991> so I guess the discussion is over, with us discussing different things
14:13 <+bridge_> <Assa> time to check ddnet for vulnerabilities for AI
14:13 <+bridge_> <Assa> time to check ddnet for vulnerabilities with AI
14:13 <+bridge_> <heinrich5991> time to fix existing known vulnerabilities 😉
14:14 <+bridge_> <Assa> are there many? Except in the map io
14:14 <+bridge_> <avolicious> UDP
14:14 <+bridge_> <heinrich5991> plain text password transfer
14:14 <+bridge_> <avolicious> Time to move to QUIC
14:14 <+bridge_> <Assa> well yeah a MITM is always hard
14:14 <+bridge_> <avolicious> Then accounts
14:15 <+bridge_> <avolicious> Then getting rid of the old protocol entirely
14:15 <+bridge_> <Assa> Yeah, I give AI heinrichs PR and say "implement the rest of it"
14:15 <+bridge_> <avolicious> Sounds like a good plan
14:15 <+bridge_> <Assa> who wants to review 12K LOC?
14:16 <+bridge_> <avolicious> Just add "Keep it simple, stupid" in the prompt
14:17 <+bridge_> <heinrich5991> like I recently said to Learath2:
14:17 <+bridge_> <heinrich5991> > there's some unsexy stuff that needs to be done, e.g. writing build files, making sure it works on windows and while cross-compiling
14:17 <+bridge_> <heinrich5991> the latter part is probably not doable by LLMs on their own
14:17 <+bridge_> <heinrich5991> since it's integration work
14:18 <+bridge_> <avolicious> If it can delete prod databases, it can fix this too ^^
14:18 <+bridge_> <Assa> I wouldn't be that sure about it :justatest:
14:18 <+bridge_> <heinrich5991> claude code doesn't run on windows, as a starter
14:18 <+bridge_> <avolicious> WSL
14:19 <+bridge_> <heinrich5991> yes, but then you're not in the environment you want to work in anymore
14:19 <+bridge_> <heinrich5991> if you give me a WSL shell, I'm going to have a really hard time checking windows compiles
14:19 <+bridge_> <avolicious> ?
14:19 <+bridge_> <Assa> it doesn't need to - you only need a tool on windows that goes against the existing openapi spec - i.e. opencode
14:19 <+bridge_> <heinrich5991> by all means, do it if you think it's easy
14:19 <+bridge_> <Assa> you can host the server fully remotely like we do at work
14:20 <+bridge_> <Assa> and you can host all models afaict in a docker container, which should also work on windows (which also makes use of WSL)
14:20 <+bridge_> <Assa> just be warned about the 12 GB used memory
14:22 <+bridge_> <heinrich5991> @avolicious @essigautomat when can I expect the report on how to make the PR build on windows?
14:23 <+bridge_> <avolicious> I am currently working on something else, but I'll add it on the ToDo
14:24 <+bridge_> <heinrich5991> please make it a short report on what needs to be changed and not a +10000 lines PR
14:24 <+bridge_> <heinrich5991> I'd really appreciate it
14:24 <+bridge_> <Assa> I am so overworked right now - I'll also put it on my todo right after fixing my homeserver and having my resin 3D printer in a usable state :justatest:
14:24 <+bridge_> <avolicious> Sure, the report is a PoC, not the PR itself
14:25 <+bridge_> <heinrich5991> and ideally understand the LLM solution yourself so I'll not be the human interface for the LLM to fix up its solution
14:25 <+bridge_> <Assa> I think I'll ask the AI first to understand your PR
14:55 <+bridge_> <learath2> Software peoples burning desire to be replaced by AI models ASAP will be studied in the coming decades
15:03 <+bridge_> <pioooooo> Do you mean the quic one? It's blocked by building on windows?
15:05 <+bridge_> <heinrich5991> see checkboxes at https://github.com/ddnet/ddnet/pull/6961
15:33 <+bridge_> <pioooooo> I see
15:52 <+bridge_> <remakepower> Are there any advantages to move to QUIC?
15:55 <+bridge_> <remakepower> Ah I saw
15:55 <+bridge_> <remakepower> Depends on Rust creates
15:55 <+bridge_> <remakepower> :frozen:
16:04 <+bridge_> <heinrich5991> 1) encrypted connection 2) a way to make accounts work securely
16:06 <+bridge_> <remakepower> :thonk:
16:06 <+bridge_> <heinrich5991> I guess 3) no longer implementing our own reliable messaging
16:08 <+bridge_> <remakepower> I always felt that ddnet was aggressive in these regards.
16:09 <+bridge_> <heinrich5991> in what regards?
16:10 <+bridge_> <remakepower> e.g. QUIC.
16:12 <+bridge_> <remakepower> There are more things which have been the parts of DDNet.
16:16 <+bridge_> <remakepower> :thonk:
16:16 <+bridge_> <remakepower> ah
16:17 <+bridge_> <heinrich5991> I think we're moving exceptionally slow with QUIC tbh
16:17 <+bridge_> <remakepower> That's my word issue.
16:18 <+bridge_> <remakepower> `aggressive` should be `radical`
16:19 <+bridge_> <remakepower> also I think DDNet is the first game project I have seen that use HTTP master server.
16:20 <+bridge_> <heinrich5991> I'd guess most games fetch most their stuff via HTTPS these days
16:21 <+bridge_> <remakepower> At least we could see that some Steam games community servers are fetched not by HTTPS.
16:23 <+bridge_> <remakepower> And
16:24 <+bridge_> <remakepower> I think it has been 2 years.
16:24 <+bridge_> <remakepower> however, nobody cares #5156
16:24 <+bridge_> <DDNet> https://github.com/ddnet/ddnet/issues/5156
16:25 <+bridge_> <heinrich5991> yea, no one cared enough to solve the issue
16:26 <+bridge_> <heinrich5991> if you want to solve the issue, we can have a chat about how it could be implemented
16:28 <+bridge_> <remakepower> Maybe could send a heartbeat to verify the latency and if the master is alive
16:29 <+bridge_> <heinrich5991> yea, could try registering with a random and if it works, continue to use that one
16:29 <+bridge_> <heinrich5991> if it fails, select the next one, I guess?
16:29 <+bridge_> <remakepower> Also
16:29 <+bridge_> <heinrich5991> what is "failure" here? port forward check included, I guess?
16:29 <+bridge_> <remakepower> if the master reports that it couldn
16:30 <+bridge_> <remakepower> if the master reports that it couldn't go through the firewall, then select the next one.
16:30 <+bridge_> <heinrich5991> yea, sounds good 🙂 want to implement that?
16:31 <+bridge_> <remakepower> I would like to work on the cpp part :3
16:32 <+bridge_> <heinrich5991> it only has a C++ part, I think
16:33 <+bridge_> <remakepower> Shouldn't the master send a heartbeat packet back?
16:33 <+bridge_> <heinrich5991> I'd just try registering immediately
16:34 <+bridge_> <remakepower> there seems to be only one master now.
16:34 <+bridge_> <heinrich5991> yes
16:34 <+bridge_> <heinrich5991> do you need another one for testing?
16:35 <+bridge_> <remakepower> Would be useful after the PR opens.
16:37 <+bridge_> <remakepower> hmm
16:37 <+bridge_> <remakepower> make the default config from
16:37 <+bridge_> <remakepower> `https://master1.ddnet.org/ddnet/15/register` to `https://master%d.ddnet.org/ddnet/15/register`?
16:38 <+bridge_> <heinrich5991> you can do that for now, for the final PR probably `https://register%d.ddnet.org/ddnet/15/register`
16:38 <+bridge_> <heinrich5991> so that we can host the server list and the registration on different domains
16:40 <+bridge_> <remakepower> ah
16:40 <+bridge_> <remakepower> so the best master server is the server could register?
16:40 <+bridge_> <remakepower> (first)
16:40 <+bridge_> <heinrich5991> yes
16:42 <+bridge_> <remakepower> Are there only 4 masters' domains could be used?
16:42 <+bridge_> <heinrich5991> yes
16:52 <+bridge_> <remakepower> hm.. skip after the server reports error three times?
16:53 <+bridge_> <heinrich5991> yea, maybe something like that 🙂
17:20 <+bridge_> <remakepower> oh no,
17:21 <+bridge_> <remakepower> there would be warning
17:30 <+bridge_> <remakepower> master1 reports me The requested URL returned error: 400
17:31 <+bridge_> <remakepower> when register to master4, it reports me: `The requested URL returned error: 400`
17:32 <+bridge_> <heinrich5991> that makes sense, there's no register endpoint on that server, currently
17:32 <+bridge_> <remakepower> I saw a double register problem
17:33 <+bridge_> <remakepower> ```
17:33 <+bridge_> <remakepower> 2026-02-11 00:31:51 I register/6/ipv6: registering...
17:33 <+bridge_> <remakepower> 2026-02-11 00:31:51 I register/6/ipv4: registering...
17:33 <+bridge_> <remakepower> 2026-02-11 00:31:51 I register/7/ipv6: registering...
17:33 <+bridge_> <remakepower> 2026-02-11 00:31:51 I register/7/ipv4: registering...
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 I register/6/ipv6: registering...
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 I register/6/ipv4: registering...
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 I register/7/ipv6: registering...
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 I register/7/ipv4: registering...
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 E register/7/ipv6: ERROR: the master server reports that clients can not connect to this server.
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 E register/7/ipv6: ERROR: configure your firewall/nat to let through udp on port 8303.
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 E register/6/ipv6: ERROR: the master server reports that clients can not connect to this server.
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 E register/6/ipv6: ERROR: configure your firewall/nat to let through udp on port 8303.
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 E register/7/ipv4: ERROR: the master server reports that clients can not connect to this server.
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 E register/7/ipv4: ERROR: configure your firewall/nat to let through udp on port 8303.
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 E register/6/ipv4: ERROR: the master server reports that clients can not connect to this server.
17:33 <+bridge_> <remakepower> 2026-02-11 00:32:06 E register/6/ipv4: ERROR: configure your firewall/nat to let through udp on port 8303.
17:33 <+bridge_> <remakepower> ```
17:36 <+bridge_> <remakepower> hm
17:38 <+bridge_> <remakepower> ```
17:38 <+bridge_> <remakepower> - [ ] Tested the change ingame
17:38 <+bridge_> <remakepower> ```
17:38 <+bridge_> <remakepower> Does this `ingame` mean `in the client`?
17:39 <+bridge_> <heinrich5991> yes, but ignore that part or remove it from the check box
17:39 <+bridge_> <heinrich5991> just tick it if you tested it 🙂
17:42 <+bridge_> <remakepower> How could I test a master server thing in client????
17:43 <+bridge_> <remakepower> How could I test a master server thing in client???? (which is not related)
17:47 <+bridge_> <heinrich5991> ignore that part or remove the word "ingame" from the checkbox ^^
17:47 <+bridge_> <heinrich5991> we don't reject your PR if it doesn't have all boxes ticked ^^
17:51 <+bridge_> <remakepower> Ah right
17:51 <+bridge_> <remakepower> where the next master is
18:20 <+bridge_> <remakepower> @heinrich5991