2026-01-15 01:38:34 <cleobuline  > !uptime
2026-01-15 01:38:34 <LispBot758  > 298h 57min 18s
2026-01-15 02:52:12 <KipIngram   > Holy cow.  I just discovered that we've barely had a "written down" value of the the gravitational constant G for much over a century.  I had no idea that was such a late development.
2026-01-15 02:53:00 <KipIngram   > Of course we didn't know that galaxies beyond our own existed until about a hundred years ago, so I guess there's a fair bit of late breaking knowledge.
2026-01-15 09:34:56 <veltas      > I guess because it's a weak force so difficult to measure
2026-01-15 09:42:07 <veltas      > Newton made an educated guess that was bloody close, I think he must have cheated
2026-01-15 09:43:57 <veltas      > KipIngram are you sure that's right, Cavendish measured it to about 3sf in 1798
2026-01-15 09:44:13 <veltas      > Not even sure what you're saying really
2026-01-15 15:31:31 <KipIngram   > What Cavendish measured and reported was the density of Earth.  I'm referring to the explicit use of the constant G.  For a long time they avoided making it explicit.
2026-01-15 15:31:37 <KipIngram   > That's what surprised me so much.
2026-01-15 15:32:19 <KipIngram   > Cavidish definitely had the "essence" of it in hand, but just didn't express it that way.
2026-01-15 16:57:30 <tpbsd       > interesting article about TSMC's node roadmap https://www.culpium.com/p/exclusiveapple-is-fighting-for-tsmc
2026-01-15 16:59:29 <tpbsd       > they have a chart that shown a linear reduction in nodesize to about 2018 (7nm) and then it flattens out
2026-01-15 17:00:47 <tpbsd       > of course embedded is still at 40nm (analog parts are much larger node size ... higher voltages etc) which they hit in 2008
2026-01-15 17:10:30 <tpbsd       > Confusingly, N4P is sometimes called a 4nm node and sometimes 5nm, but it’s part of the broader 5nm family of processes — TSMC calls it an enhanced version of 5nm. Call it 4nm if you like
2026-01-15 17:58:32 <amby        > and of course no measurement of a 4nm chip is actually 4nm
2026-01-15 18:14:56 <tpbsd       > amby, I believe that as the design of the transistors has changed that it's become harder to quantify their size
2026-01-15 18:19:01 <amby        > theyre definitely bigger than 4nm though
2026-01-15 18:20:54 <tpbsd       > yeah, the node size is actually the 'process name' I believe, it's not really related to the actual size, it's mainly marketing speak
2026-01-15 18:22:31 <tpbsd       > probably a quantification of 'transistors per mm' or something like that would be more accurate an indicator ?
2026-01-15 18:24:21 <amby        > wikipedia lists megatransistors per square millimetre for various peoples processes
2026-01-15 23:45:16 <KipIngram   > :-)  All of it just translates to "a LOT" for me.
2026-01-15 23:46:00 <KipIngram   > It's a shame we devote half or more of those transistors to features aimed at making legacy code run fast without re-compiling.
2026-01-15 23:46:35 <KipIngram   > I mean, just think about instruction re-ordering.  If you were willing to re-compile, you could put that intelligence in the compiler and get the damn instructions in the right order to begin with.
2026-01-15 23:47:01 <KipIngram   > But no - we re-order them EVERY SINGLE TIME THEY EXECUTE, on EVERY PROCESSOR IN THE WORLD.  Just think for a minute about how wasteful that is.
2026-01-15 23:59:07 <KipIngram   > Register renaming.  Just expose the resources and let the compiler figure it out.  Same reasoning - do it once when the vendor compiles the software instead of every time instructions execute on every processor.