By: syd <syd@kronor.io>
Quantify the scaling caveat: package size barely matters FINDINGS said the scaling measurement was synthetic with uniformly small packages, and that a real workspace with uneven sizes might evaluate differently. Now measured rather than hedged. The generator takes a "varied" mode giving packages 2 to 9 modules, 1 to 3 test files, and files of 20 to 260 lines. At the same project count that is 5.0 MB of source against 177 KB, and evaluation goes 121 ms to 132 ms: twenty-eight times the source for 9%. So evaluation tracks project count weakly and source size weaklier still. What remains untested is a deep cross-package import graph and tsconfig project references; the generated workspace is layered but shallow, and the caveat now says only that. The scale table is refreshed from the same run, which also moved slightly: 98 ms at 24 projects to 143 ms at 389.