--- Log opened Mon Jul 08 00:00:47 2024
01:08 -!- tinwhiskers [~tinwhiske@user/tinwhiskers] has quit [Quit: No Ping reply in 180 seconds.]
01:09 -!- tinwhiskers [~tinwhiske@user/tinwhiskers] has joined #hplusroadmap
03:02 -!- gl00ten [~gl00ten@bl11-102-149.dsl.telepac.pt] has joined #hplusroadmap
03:23 -!- L29Ah [~L29Ah@wikipedia/L29Ah] has joined #hplusroadmap
06:24 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has quit [Remote host closed the connection]
06:24 -!- justanotheruser [~justanoth@gateway/tor-sasl/justanotheruser] has joined #hplusroadmap
09:12 < geneh2> state space explosion isn't a problem if you have gradients
09:29 < geneh2> the space of possible neural networks is huge, but we can get gradients
10:13 < hprmbridge> nmz787> I'm not sure how gradients impact this conversation. That sounds like something related to optimization not satisfying constraints. Related, but sort of icing on the cake. What I mean is, you might have a functional block laid out, but later when the ass-end of the chip is getting laid out, it might have some failing constraint that upends the earliest block your tool laid out, cause it to ve
10:13 < hprmbridge> nmz787> invalidated and possibly invalidating many other connections and layout features.
10:48 < TMA> as an amusing sidenote to the EDA remark on HN: morphle, the author of that remark has currently karma exactly equal to 666
11:10 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
11:11 -!- gl00ten [~gl00ten@bl11-102-149.dsl.telepac.pt] has quit [Ping timeout: 272 seconds]
11:13 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 268 seconds]
11:14 -!- test_ is now known as _flood
11:59 -!- ike8 [e8f913dbdf@irc.cheogram.com] has quit [Quit: Gateway shutdown]
12:17 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
12:18 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
12:25 < docl> pasky: I've been rewriting your claude.vim for my own use. planning to paste to my website soon if I don't get sidetracked
12:26 < docl> the temptation is to write my own paste mechanism to avoid githubbing. hmm. I should probably just stick the file up for now
12:33 -!- ike8 [e8f913dbdf@irc.cheogram.com] has joined #hplusroadmap
13:18 -!- streety [~streety@176-58-127-93.ip.linodeusercontent.com] has quit [Quit: ZNC 1.8.2 - https://znc.in]
13:20 -!- streety [~streety@176-58-127-93.ip.linodeusercontent.com] has joined #hplusroadmap
13:44 -!- gl00ten [~gl00ten@bl11-102-149.dsl.telepac.pt] has joined #hplusroadmap
13:55 < pasky> docl: or make a PR if there's something interesting :)
13:56 < pasky> I wanted to add tool use and thread of thought prompt but got sidetracked myself
14:26 < docl> for the time being, current version is here: https://moredoingstuff.com/pastes/claudesuper.vim
14:29 < docl> if nothing else, it's a decent adhd mitigation tool for sustaining hyperfocus. the suggestions are so crappy sometimes though
14:29 < docl> I burned through 1300 queries over the past week
14:30 < docl> instead of using the chat, I've been often just adding comments
14:31 < docl> lots of good vim practice. hadn't been using vim diff before
14:34 < docl> in quite a few cases I wound up recoding stuff by hand for brevity and readability. like figure out the necessary keystrokes and put it after a normal!
15:07 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
15:11 -!- _flood [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 252 seconds]
15:34 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
15:38 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 268 seconds]
15:39 -!- gl00ten [~gl00ten@bl11-102-149.dsl.telepac.pt] has quit [Ping timeout: 252 seconds]
15:46 -!- gl00ten [~gl00ten@bl11-102-149.dsl.telepac.pt] has joined #hplusroadmap
15:49 -!- test__ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
15:51 -!- gl00ten [~gl00ten@bl11-102-149.dsl.telepac.pt] has quit [Ping timeout: 252 seconds]
15:53 -!- test_ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 268 seconds]
16:09 < pasky> docl: yeah the suggestions aren't always the best :) I'm lowkey convinced that it ought to be possible to cleverly instrument existing top llms to be much more thoughtful / powerful, but with substantial prompting (after all the default claude.ai system prompt is many thousands of tokens.. but it still comes short)
16:13 < pasky> docl: hmm interesting, having a permanent proposed changes window might make sense on bigger displays
17:33 -!- flyback [~flyback@2601:540:c781:7f90:857b:6ee7:e76a:e703] has quit [Ping timeout: 260 seconds]
17:45 -!- flyback [~flyback@2601:540:c781:7f90:2ea:4995:ad03:87fd] has joined #hplusroadmap
18:28 < docl> pasky: it might just be personal preference, but I was getting annoyed by the changing width when it went into diff mode. my screen is fairly wide these days. could possibly be worth switching the chat buffer out with the proposed changes buffer to have there be 2 windows
18:31 < docl> I'm kind of happy to see vim being useful as a textual gui / tui for this
18:35 -!- flooded [flooded@gateway/vpn/protonvpn/flood/x-43489060] has joined #hplusroadmap
18:37 -!- test__ [flooded@gateway/vpn/protonvpn/flood/x-43489060] has quit [Ping timeout: 268 seconds]
19:22 < fenn> stupid billion dollar startup idea: train stable diffusion on fully optimized chunks of chip fab IC artwork. optimize chunks separately with all the multiphysics and optical serifs, then glue them together with img2img keeping the original data fixed
19:23 < fenn> i know they're already doing "machine learning accelerated EDA" but who knows if it's as good as the stupid thing
19:35 -!- Guest72 [~Guest72@cst-prg-26-13.cust.vodafone.cz] has joined #hplusroadmap
19:35 -!- Guest72 [~Guest72@cst-prg-26-13.cust.vodafone.cz] has quit [Client Quit]
20:00 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has quit [Remote host closed the connection]
20:00 -!- Gooberpatrol66 [~Gooberpat@user/gooberpatrol66] has joined #hplusroadmap
20:10 < hprmbridge> nmz787> I've been talking with my team about that fenn, the problem is going from one process node to another things can drastically change
20:14 < hprmbridge> nmz787> And IP blocks are already copy-pastable once laid out
20:20 -!- mxz_ [~mxz@user/mxz] has quit [Ping timeout: 264 seconds]
20:21 -!- mxz__ [~mxz@user/mxz] has joined #hplusroadmap
20:23 -!- mxz [~mxz@user/mxz] has quit [Ping timeout: 264 seconds]
20:23 -!- mxz__ is now known as mxz
20:38 -!- Guest94 [~Guest94@cst2-167-47.cust.vodafone.cz] has joined #hplusroadmap
20:42 -!- Guest94 [~Guest94@cst2-167-47.cust.vodafone.cz] has quit [Client Quit]
22:28 -!- mxz_ [~mxz@user/mxz] has joined #hplusroadmap
23:50 -!- TMM [hp@amanda.tmm.cx] has quit [Quit: https://quassel-irc.org - Chat comfortably. Anywhere.]
23:51 -!- TMM [hp@amanda.tmm.cx] has joined #hplusroadmap
--- Log closed Tue Jul 09 00:00:47 2024