Day changed to 2026-07-01
[08:35] Diana Coman: Vivian Sporepress, I'm making more cft even now but in any case, I have 200cft in storage at q19 currently, so ping me when you get back to PPP and we see
[21:14] Vivian Sporepress: Diana Coman: I'm here at your storage
[23:14] Vivian Sporepress: I think there's really something funny going on vis-a-vis that peculiar lbn column at left edge ppp, http://ossasepia.com/2026/04/01/Eulora-2-Logs-for-Apr-2026/#5396 - I've seen similar effects in other near-deserted sectors now too
[23:14] Ulrich Logfetch: Quoting 2026-04-20 (Eulora 2) Vivian Sporepress: I've seen that too on occasion, even among the cr in ppp from my old data. the newer client knowledge based map also shows a whole column of them at 40 unit spacing, all on their own near the left edge in ppp; especially odd since I don't recall it being me finding them, yet how would I have gotten close enough to see them all but not any actual claims nearby...
[23:16] Vivian Sporepress: I've certainly done 40-unit grid exploring runs so it seems possibly some kind of looking at the reflection of my own retina, as it were
[23:32] Vivian Sporepress: two smoking guns from knowledge/res_sect_kw.txt: in the first, there are some true lbn-spots within my 40-unit grid which covers x=-240..280 by z=-240..280. two jump out as unaligned with the grid, perhaps they were manual explores. then there's one at 0,0,0 - a suspicious y-value by itself but then there's a flotsam found in same sector at same x-z origin but y=12. then a column of 10 at left edge, way outside the grid and outside the normal sight horizon of anywhere I'd been, and exactly the same values as those previously noted in PPP.
[23:39] Vivian Sporepress: in the second I'd covered a larger grid and the ones at z=-320 and -360 appear legit, but again there's the suspicious all-zeros, this time shadowing a CDG at 0,91,0; I recall it and the -440,12,-79 popping up on the plot early in my exploration before I'd got anywhere near x=-440.
[23:45] Vivian Sporepress: doing a text search on that most peculiar -440.00;12.00;-79.00 hits LBN (and only LBN) in 8 sectors including PPP.
[23:46] Vivian Sporepress: that's out of 19 total sectors for which I have LBN knowledge entries.
[23:52] Vivian Sporepress: a similar but less popular one found by visual means: -400.00;0.00;40.00 LBN in 4 sectors including PPP
Day changed to 2026-07-02
[00:02] Vivian Sporepress: another somewhat more complex: -280.00;100.00;-320.00 shows as LBN in 3 sectors including PPP but CDG in sector (-1;0), yet -280.00;67.00;-320.00 is CDG in PPP. Apologies to bystanders for the log-dump, I wasn't expecting to turn up the further clues.
[00:07] Felicia Solecrimp: Diana Coman okay
[00:07] Felicia Solecrimp: Ive been reading thats what satisfied me at most to stop thinking about it
[08:02] Diana Coman: Vivian Sporepress, it seems we didn't quite sync for the cft. Today I'll likely be around in the evening for a couple of hours, from 7pm utc, hopefully that works. Otherwise I guess the fallback time remains the usual on Monday since it's fixed
[08:08] Diana Coman: thanks for the lbn/resource data feedback, I noticed that there are indeed some wrong-sector placements, from what I can tell mostly happening at a sector switch (always the more troublesome point). Not quite sure yet as to what would be a cure for it that is not worse than the problem though. In principle at least I could look to enforce more strictly a clean slate clientside whenever the sector changes, perhaps.
[17:58] Vivian Sporepress: Diana Coman: 7pm works. sector switch seems a plausible explanation, except isn't there a sector ID -> item ID parent-child relationship that would keep it clear even if some no-longer-visible items linger in the object cache?
[20:06] Diana Coman: Vivian Sporepress, there is the hierarchical relationship, I guess an atomic check on that would be the most straightforward thing to add in order to address this. Seemed rather overkill for resource knowledge but most likely unavoidable longer-term anyhow
[21:06] Vivian Sporepress: ready when you are, Diana Coman
[21:08] Diana Coman: Vivian Sporepress, I have 200 cft q19, q23 or q24, which one?
[21:08] Vivian Sporepress: ah, I'll stick with the q19
[21:10] Vivian Sporepress: tyvm
[21:10] Diana Coman: enjoy!
[21:11] Diana Coman: meanwhile my cft seems to keep pushing upwards quality wise, dunno if I should keep some lower q separate
[21:13] Diana Coman: the mixing storage is most useful but since they end up all sorts of quality, it's either a pain to keep them separate or everything ends up around the same quality point, mostly pushed up or down by the crafting outputs
[21:25] Vivian Sporepress: it's a sort of entropy for quality levels.
[21:25] Vivian Sporepress: I just got 130 nt q1 from a small followed by 1505 nt q1 from a tiny, go figure.
[21:26] Vivian Sporepress: the small cost more but at least returned more experience
Day changed to 2026-07-04
[07:53] Diana Coman: heh, that 1505 nt q1 must have been quite the tiny!
[08:07] Diana Coman: Felicia Solecrimp, talking of AI, there is actually a more promising direction than the most visible hype that you get "following" all the usual (aka most visible/pushed) channels - quite ditching in fact the focus on blind "prediction" by means of "what came before in a lot of data" and aiming instead to extract abstractions on which to base some form of inferrence of outcome. It's only just getting the usual ton of funding so it remains to be seen how it's actually done, as always, but it's certainly a more promising direction and it seems to be getting a ton of funding currently too so it's quite likely to be heard of some more in the not so distant future.
[08:17] Diana Coman: for completeness if we are at it, there were earlier attempts for some form of machine learning aimed at extracting essential features to generate rather than merely predict, e.g. Ha and Schmidhuber's "World Models" from 2018 (https://arxiv.org/pdf/1803.10122) but most likely it's a more complex problem than the blind prediction and so it was the prediction that got pushed instead. Possibly the generative comes next, finally.
[08:35] Diana Coman: Vivian Sporepress, I'm making more cft even now but in any case, I have 200cft in storage at q19 currently, so ping me when you get back to PPP and we see
[21:14] Vivian Sporepress: Diana Coman: I'm here at your storage
[23:14] Vivian Sporepress: I think there's really something funny going on vis-a-vis that peculiar lbn column at left edge ppp, http://ossasepia.com/2026/04/01/Eulora-2-Logs-for-Apr-2026/#5396 - I've seen similar effects in other near-deserted sectors now too
[23:14] Ulrich Logfetch: Quoting 2026-04-20 (Eulora 2) Vivian Sporepress: I've seen that too on occasion, even among the cr in ppp from my old data. the newer client knowledge based map also shows a whole column of them at 40 unit spacing, all on their own near the left edge in ppp; especially odd since I don't recall it being me finding them, yet how would I have gotten close enough to see them all but not any actual claims nearby...
[23:16] Vivian Sporepress: I've certainly done 40-unit grid exploring runs so it seems possibly some kind of looking at the reflection of my own retina, as it were
[23:32] Vivian Sporepress: two smoking guns from knowledge/res_sect_kw.txt: in the first, there are some true lbn-spots within my 40-unit grid which covers x=-240..280 by z=-240..280. two jump out as unaligned with the grid, perhaps they were manual explores. then there's one at 0,0,0 - a suspicious y-value by itself but then there's a flotsam found in same sector at same x-z origin but y=12. then a column of 10 at left edge, way outside the grid and outside the normal sight horizon of anywhere I'd been, and exactly the same values as those previously noted in PPP.
[23:39] Vivian Sporepress: in the second I'd covered a larger grid and the ones at z=-320 and -360 appear legit, but again there's the suspicious all-zeros, this time shadowing a CDG at 0,91,0; I recall it and the -440,12,-79 popping up on the plot early in my exploration before I'd got anywhere near x=-440.
[23:45] Vivian Sporepress: doing a text search on that most peculiar -440.00;12.00;-79.00 hits LBN (and only LBN) in 8 sectors including PPP.
[23:46] Vivian Sporepress: that's out of 19 total sectors for which I have LBN knowledge entries.
[23:52] Vivian Sporepress: a similar but less popular one found by visual means: -400.00;0.00;40.00 LBN in 4 sectors including PPP
Day changed to 2026-07-02
[00:02] Vivian Sporepress: another somewhat more complex: -280.00;100.00;-320.00 shows as LBN in 3 sectors including PPP but CDG in sector (-1;0), yet -280.00;67.00;-320.00 is CDG in PPP. Apologies to bystanders for the log-dump, I wasn't expecting to turn up the further clues.
[00:07] Felicia Solecrimp: Diana Coman okay
[00:07] Felicia Solecrimp: Ive been reading thats what satisfied me at most to stop thinking about it
[08:02] Diana Coman: Vivian Sporepress, it seems we didn't quite sync for the cft. Today I'll likely be around in the evening for a couple of hours, from 7pm utc, hopefully that works. Otherwise I guess the fallback time remains the usual on Monday since it's fixed
[08:08] Diana Coman: thanks for the lbn/resource data feedback, I noticed that there are indeed some wrong-sector placements, from what I can tell mostly happening at a sector switch (always the more troublesome point). Not quite sure yet as to what would be a cure for it that is not worse than the problem though. In principle at least I could look to enforce more strictly a clean slate clientside whenever the sector changes, perhaps.
[17:58] Vivian Sporepress: Diana Coman: 7pm works. sector switch seems a plausible explanation, except isn't there a sector ID -> item ID parent-child relationship that would keep it clear even if some no-longer-visible items linger in the object cache?
[20:06] Diana Coman: Vivian Sporepress, there is the hierarchical relationship, I guess an atomic check on that would be the most straightforward thing to add in order to address this. Seemed rather overkill for resource knowledge but most likely unavoidable longer-term anyhow
[21:06] Vivian Sporepress: ready when you are, Diana Coman
[21:08] Diana Coman: Vivian Sporepress, I have 200 cft q19, q23 or q24, which one?
[21:08] Vivian Sporepress: ah, I'll stick with the q19
[21:10] Vivian Sporepress: tyvm
[21:10] Diana Coman: enjoy!
[21:11] Diana Coman: meanwhile my cft seems to keep pushing upwards quality wise, dunno if I should keep some lower q separate
[21:13] Diana Coman: the mixing storage is most useful but since they end up all sorts of quality, it's either a pain to keep them separate or everything ends up around the same quality point, mostly pushed up or down by the crafting outputs
[21:25] Vivian Sporepress: it's a sort of entropy for quality levels.
[21:25] Vivian Sporepress: I just got 130 nt q1 from a small followed by 1505 nt q1 from a tiny, go figure.
[21:26] Vivian Sporepress: the small cost more but at least returned more experience
Day changed to 2026-07-04
[07:53] Diana Coman: heh, that 1505 nt q1 must have been quite the tiny!
[08:07] Diana Coman: Felicia Solecrimp, talking of AI, there is actually a more promising direction than the most visible hype that you get "following" all the usual (aka most visible/pushed) channels - quite ditching in fact the focus on blind "prediction" by means of "what came before in a lot of data" and aiming instead to extract abstractions on which to base some form of inferrence of outcome. It's only just getting the usual ton of funding so it remains to be seen how it's actually done, as always, but it's certainly a more promising direction and it seems to be getting a ton of funding currently too so it's quite likely to be heard of some more in the not so distant future.
[08:17] Diana Coman: for completeness if we are at it, there were earlier attempts for some form of machine learning aimed at extracting essential features to generate rather than merely predict, e.g. Ha and Schmidhuber's "World Models" from 2018 (https://arxiv.org/pdf/1803.10122) but most likely it's a more complex problem than the blind prediction and so it was the prediction that got pushed instead. Possibly the generative comes next, finally.