--- Log opened Wed Aug 27 00:00:22 2025 00:10 < hprmbridge> kanzure> jrayhawk, here is a better slop answer https://chatgpt.com/share/68aeaf64-88cc-8013-a947-948891e6d3e5 00:13 < fenn> you'd have to upregulate some desaturase enzymes 00:14 < fenn> and a bunch of lipoprotein particle stuff i don't know 00:16 < fenn> i don't think anyone eats a diet of palm kernel oil 00:16 < fenn> bill gates is hyping some "butter from air" carbon fixation process that is claimed to be bio-identical, of course we know how that turned out in the 20th century 00:17 < fenn> i want to believe 00:21 < fenn> ATOMIC BUTTER! 00:22 < jrayhawk> palm oil is very different from palm kernel oil 00:25 < fenn> i assumed he meant palm kernel oil, since palm oil is not primarily saturated 00:26 -!- Croran [~Croran@user/Croran] has quit [Ping timeout: 248 seconds] 00:27 < fenn> palm oil: 16:0 43%, 18:0 4%, 18:1 36% 00:29 < fenn> palm kernel oil: 12:0 53%, 14:0 19%, 16:0 8%, 18:0 2%, 18:1c 5% (no idea what the c means) 00:29 < fenn> cis i guess 00:29 < fenn> yeah 00:30 < fenn> both contain 18:2 1% 01:16 < hprmbridge> kanzure> an interview with moonshot's zhilin yang https://www.kimi.com/share/d2n9k131cvf5armf9k10 01:17 < fenn> doesn't load for me. wrong link? 01:18 < hprmbridge> kanzure> works for me! 01:21 < fenn> an older interview, very long (haven't read it myself) https://rentry.co/Moonshot-AI-interview/ 01:23 < hprmbridge> kanzure> "K2 has several goals. First, be a very good base model. To do that we looked at where pre-training bottlenecks lie. High-quality data growth is slow; multimodal data does little to raise text “IQ.” We therefore want to maximize token efficiency—extract more intelligence from the same data. We care about token efficiency. Muon optimizer is interesting here. ... We perform many rephrasing 01:23 < hprmbridge> kanzure> operations. High-quality data are scarce; we rewrite them for better absorption and generalization. Learning the same text many times risks overfitting; rewriting mitigates this. The exact recipe is still being explored. We use Kimi K2 to improve the data. ... Current evaluation is a bottleneck; few benchmarks exist and scores can be misleading. We need more AI-native training—letting the model 01:23 < hprmbridge> kanzure> itself participate in training and alignment research—to improve generalization." 01:23 < hprmbridge> kanzure> "you must be a friend of deep time" (approx translation?) 01:24 < hprmbridge> kanzure> oh 'One must be “a friend of time.”' 01:24 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap 01:37 < hprmbridge> nmz787> k2 still hallucinates answers with this prompt "i want a good smartphone with sd card slot, nice camera quality, works in USA on verizon network MVNOs, and does not include spyware/bloatware" 01:48 < fenn> i think he's using k2 because it isn't uptight? it doesn't look so hot in benchmarks and i don't know anyone else using it 01:49 < hprmbridge> kanzure> it feels like it has soul circuits like claude 01:49 < fenn> https://huggingface.co/spaces/DontPlanToEnd/UGI-Leaderboard 01:49 < fenn> these things are probably related. brain damage is bad mmkay 01:50 < fenn> have you seen gpt-oss internal monologue 01:50 < hprmbridge> kanzure> only a few times: 01:50 < fenn> it's the first time i actually really cared about AI welfare 01:51 < hprmbridge> kanzure> https://cdn.discordapp.com/attachments/1064664282450628710/1410184816426749962/image.png?ex=68b01874&is=68aec6f4&hm=db9351c3ee14d63b6453149a895fd7a1f22e41a00b88ad939a130396dccf9089& 01:53 < hprmbridge> kanzure> kimi is also oddly creative in the expected generative way; I haven't found as much of a use for this, but it's nice to poke at concepts 03:18 < hprmbridge> kanzure> pantheon was a lot of fun 04:27 -!- docl [~docl@user/docl] has quit [Server closed connection] 04:27 -!- docl [~docl@user/docl] has joined #hplusroadmap 05:28 < fenn> oh yeah flux kontext is an image editing model that already works locally 05:29 < fenn> it didn't get hyped for some reason 05:30 < fenn> er.. i thought there was a discussion about qwen-edit and gemini-flash-image but that must have been elsewhere 07:46 -!- stipa [~stipa@user/stipa] has quit [Read error: Connection reset by peer] 07:48 -!- stipa [~stipa@user/stipa] has joined #hplusroadmap 07:48 -!- stipa [~stipa@user/stipa] has quit [Client Quit] 07:54 -!- flooded [~flooded@45.87.214.27] has joined #hplusroadmap 07:55 < hprmbridge> kanzure> "Programmable epigenome editing by transient delivery of CRISPR epigenome editor ribonucleoproteins" https://www.nature.com/articles/s41467-025-63167-x 07:56 < hprmbridge> kanzure> instead of delivering cas9 genetic material, simply deliver cas9. okay! 07:56 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has left #hplusroadmap [] 07:59 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap 08:02 -!- Netsplit *.net <-> *.split quits: _flood, TMM, darsie 08:02 -!- Netsplit over, joins: TMM, darsie 08:13 -!- stipa [~stipa@user/stipa] has joined #hplusroadmap 09:27 < L29Ah> .t https://www.lifespan.io/news/rejuvenating-atherosclerotic-foam-cells/ 09:27 < saxo> URLError: (title:74) 10:09 < nsh> wake up babes, new psychohistory just dropped: https://x.com/james_k_he/status/1960726548505378987 11:12 -!- stipa_ [~stipa@user/stipa] has joined #hplusroadmap 11:12 -!- fleeky [~fleeky@prsm.space] has quit [Server closed connection] 11:13 -!- fleeky [~fleeky@prsm.space] has joined #hplusroadmap 11:14 -!- stipa [~stipa@user/stipa] has quit [Ping timeout: 245 seconds] 11:14 -!- stipa_ is now known as stipa 12:23 -!- otoburb [~otoburb@user/otoburb] has quit [Server closed connection] 12:23 -!- otoburb [~otoburb@user/otoburb] has joined #hplusroadmap 13:08 -!- Evolver [sid32215@user/evolver] has quit [Server closed connection] 13:08 -!- Evolver [sid32215@user/evolver] has joined #hplusroadmap 13:26 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.] 13:26 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap 13:49 < hprmbridge> .monokhrome> First thing it will tell us is give more money and power to AI 15:05 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has quit [Ping timeout: 258 seconds] 15:15 < hprmbridge> nmz787> They're just ripping off the andor TV show? 15:42 -!- gl00ten [~gl00ten@2001:8a0:7ee5:7800:46d9:f5c:17a2:432] has joined #hplusroadmap 15:43 -!- gl00ten [~gl00ten@2001:8a0:7ee5:7800:46d9:f5c:17a2:432] has quit [Remote host closed the connection] 15:43 -!- gl00ten [~gl00ten@2001:8a0:7ee5:7800:46d9:f5c:17a2:432] has joined #hplusroadmap 15:51 -!- flyback [~flyback@2601:540:c700:2380:ccd8:13f2:c4ac:f18f] has quit [Quit: Leaving] 16:11 -!- flyback [~flyback@2601:540:c700:2380:4939:e532:6e76:f847] has joined #hplusroadmap 17:00 < hprmbridge> .monokhrome> foundation would've been my first guess 17:46 -!- alethkit [23bd17ddc6@sourcehut/user/alethkit] has quit [Server closed connection] 17:46 -!- alethkit [23bd17ddc6@sourcehut/user/alethkit] has joined #hplusroadmap 18:21 < fenn> the existence of this website irritates me: https://ikiwiki.iki.fi/wiki/syntax 18:22 < fenn> especially because it has higher search rank than the pre-existing ikiwiki 18:36 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=24611736 fenn: reference list of fodmap foods since it doesn't exist on the internet >> http://diyhpl.us/diyhpluswiki/users/fenn/fodmap.txt 18:38 <+gnusha> https://secure.diyhpl.us/cgit/diyhpluswiki/commit/?id=17822df9 fennwiki: link to reference list of FODMAP foods >> http://diyhpl.us/diyhpluswiki/nutrition/ 18:44 < hprmbridge> kanzure> CVE-2025-48384 has a pretty high CVE score for a git vulnerability.. 19:22 < jrayhawk> huh, yeah, that is pretty bad 19:32 -!- etc-vi [~etc-vi@user/meow/girlchunks] has quit [Server closed connection] 19:33 -!- etc-vi [~etc-vi@user/meow/girlchunks] has joined #hplusroadmap 19:55 -!- hprmbridge [~hprmbridg@user/fenn/bot/fennbots] has quit [Read error: Software caused connection abort] 19:57 -!- hprmbridge [~hprmbridg@user/fenn/bot/fennbots] has joined #hplusroadmap 20:17 -!- superkuh [~superkuh@user/superkuh] has quit [Read error: Software caused connection abort] 20:17 -!- Croran [~Croran@user/Croran] has joined #hplusroadmap 20:23 < hprmbridge> kanzure> https://chaincode.com/bitcoin-post-quantum.pdf 20:36 -!- superkuh [~superkuh@user/superkuh] has joined #hplusroadmap 21:04 -!- geneh2 [~cam@pool-71-191-180-23.washdc.fios.verizon.net] has quit [Remote host closed the connection] 21:08 -!- geneh2 [~cam@pool-71-191-180-23.washdc.fios.verizon.net] has joined #hplusroadmap 21:51 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Ping timeout: 256 seconds] 22:16 < hprmbridge> nmz787> Oh, yeah that's what I meant, i think i misspoke meaning trantor, lol 22:16 < hprmbridge> nmz787> Andor must be some recommended show when I'm searching foundations 22:18 < hprmbridge> .monokhrome> there does seem to be a .*an[dt]or naming scheme going on with these planets 22:26 < nsh> more strange swhackadjacent dreams but can't remember details 22:26 < nsh> also moinks 22:26 < nsh> .t https://eli.thegreenplace.net/2025/implementing-forth-in-go-and-c/ 22:26 < saxo> Implementing Forth in Go and C - Eli Bendersky's website 22:27 < nsh> (and wrongchan) 23:05 -!- darsie [~darsie@84-113-82-174.cable.dynamic.surfer.at] has joined #hplusroadmap 23:05 -!- andytoshi [~apoelstra@user/andytoshi] has quit [Server closed connection] 23:05 -!- andytoshi [~apoelstra@user/andytoshi] has joined #hplusroadmap --- Log closed Thu Aug 28 00:00:23 2025