Show Idle (>14 d.) Chans


← 2025-11-20 | 2025-11-22 → ↓
00:04 asciilifeform when folx w/ an actual, working industrial base print 1e6 tonned of paperola, would expect result would be diff. vs when folx who manufacture only tentacle pr0n do it
00:05 asciilifeform sorta like a 17yo student going on a bender, vs. 60yo eltsin with coupla crumbs of liver remaining
~ 9 hours 48 minutes ~
09:53 s1gnp0st this session making tempest was pretty interesting. only a couple of days to get to this point. first time I've enjoyed barking a nontrivial program into an LLM. gpt-5.1-codex-max. I jammed asciilifeform's 0xFA spec directly into the system prompt via AGENTS.md. had the thing write an enormous test suite from the spec.
09:56 s1gnp0st rather than get hung up about whether it can "think", it's perfectly useful if what it did was create a faithful *projection* of well-written thought from the spec to C
~ 47 minutes ~
10:43 asciilifeform s1gnp0st: will be interesting to read (esp. if commented re: how much massage/what kind was req'd to make it go)
~ 1 hours 32 minutes ~
12:16 s1gnp0st will post the full sauces.
~ 4 hours 32 minutes ~
16:48 phf http://logs.nosuchlabs.com/log/pest/2025-11-19#1035622 << i'll be honest, my pest doesn't even attempt to keep chains longterm. every time i reload and reconnect, it's a fresh chain.
16:50 phf i think o.g. spec required you to spit warnings and errors when there's chain shenanigans, but afaiu the new one doesn't. still i suspect if you were to make an llm /really/ stick it to a spec, a clever little bot will make you an pest that will be rather unhappy with what we actually have in the logchain
16:50 phf *afair rather
16:54 phf 􏿽(apropos pest wishlist, i really wish public logotrons would support multipart messages. or at least did when there was active conversation last time. i think mine still uses the previous multipart messages spec, with the special character, because it's backwards compatible. i thin
16:54 phf 􏿽k the new multipart requires support from everyone, otherwise messages will be lost, and i don't think everyone supports it. but i also don't really know the state of things. i'm writing a bunch of extra stuff to trigger multipart. foobar)
~ 22 minutes ~
17:16 s1gnp0st I'm receiving those with the continuation indicator just ending up in weechat as garbage on a separate message
17:18 * s1gnp0st backed off the autistic bot's urge to conform the world to an impl that stitches what it gets together at best effort.
17:20 s1gnp0st de facto required at this point that zeroed netchains and selfchains be tolerated
17:31 s1gnp0st "everything not saved will be lost"
17:32 s1gnp0st if nobody's selfchain can get to the chunks such a station laid down, gone eh? and even then, one could only walk from where such a speaker still maintaining a chain spoke
~ 1 hours 27 minutes ~
18:59 awt_akris phf: akris supports the new multipart (and old multipart). I don't know about any other stations.
19:00 awt_akris 100% multipart support would be very nice
~ 59 minutes ~
19:59 asciilifeform http://logs.bitdash.io/pest/2025-11-21#1035703 << chains, recall, came outta asciilifeform's orig attempt to 'square the circle' of getting reasonable certainty that it's the same $fella-who-aint-in-your-l1 talking at time t and t+n
19:59 bitbot Logged on 2025-11-21 16:48:56 phf[jonsykkel]: http://logs.nosuchlabs.com/log/pest/2025-11-19#1035622 << i'll be honest, my pest doesn't even attempt to keep chains longterm. every time i reload and reconnect, it's a fresh chain.
19:59 asciilifeform (this is before we even got retrieval of old msgs going)
20:00 asciilifeform theoretically they also let you have usenet-style replies/threads, but afaik this aint in any current pestron
20:00 asciilifeform (possibly 1 day somebody will bake...)
20:01 asciilifeform http://logs.bitdash.io/pest/2025-11-21#1035712 << correct
20:01 bitbot Logged on 2025-11-21 17:32:20 s1gnp0st[jonsykkel]: if nobody's selfchain can get to the chunks such a station laid down, gone eh? and even then, one could only walk from where such a speaker still maintaining a chain spoke
20:03 asciilifeform http://logs.bitdash.io/pest/2025-11-21#1035710 << zeroed selfchains always gotta work (how else new speaker?) , zeroed netchains leave you with a message that can only be put in the correct place by timestamp (yours or his) which introduces lulz potential (somebody can stick a msg 'backwards in time' using the getdata mech
20:03 bitbot Logged on 2025-11-21 17:20:15 s1gnp0st[jonsykkel]: de facto required at this point that zeroed netchains and selfchains be tolerated
20:03 asciilifeform anism)
20:07 * asciilifeform is impressed that there are afaik 5 (!!!) pestrons, by 4 people, that don't 100% implement the protocol, but still work!
20:07 * asciilifeform claps
20:08 asciilifeform iirc there was even another ( punkman ? gregory_nyssa ? or both ? ) attempt, not published but mentioned by author, in 1st yr of pestnet
20:09 * asciilifeform had own attempt, currently on ice, because it sucked arse
20:12 asciilifeform imho much of what's not implemented currently , is easy to do without because we have, what, 6 people who talk with any regularity
20:13 asciilifeform if we had the traffic of e.g. #t at its heyday -- might be different story
20:14 asciilifeform btw the 'inv' cmd in 0xfa spec is specifically to make retro-retrieval of the null-netchained msgs at least theoretically possible.
20:15 asciilifeform (i.e. a peer can offer'em to you, and you'll get'em, even if there was otherwise no way for you to know that they had existed)
20:17 * asciilifeform intended from the start to abolish not only need for centralized relay a la irc, but centralized loggers, in principle anybody plugged into a pestnet ooughta be able to stand up a logger just by switching on a 'http logger at port xyz' in his config etc
~ 43 minutes ~
21:01 s1gnp0st exported a netchain-ordered log with a little svg lineage visualizer.
21:01 s1gnp0st ~40mb chonker
← 2025-11-20 | 2025-11-22 → ↑