Thanks for flagging this — you were right, the cube-root curve was still unbounded and a stockpile run could cash out massive patent counts in one publish.
Just shipped v0.9.4 with a softcap above 100 patents/publish. Normal play is untouched (100 K prototypes is still ~92 patents), but past that the curve compresses:
- 1 M prototypes: 200 → 119
- 1 B: 2 000 → 235
- 1 T: 20 000 → 722
- 1 e14: 92 831 → 1 793
- 1 e15: 200 000 → 2 889
So the worst-case overnight farm now caps around 2–3K patents instead of 100K+. Full mastery (~904 patents) still takes the intended 8–12 publishes, and a dedicated stockpile run shaves 2–3 off that rather than ending the meta-layer outright. Existing patents and progress are preserved — only future publishes use the new formula.
Let me know if it still feels too generous after a few runs. If stockpiling is still the dominant strategy I can tighten the softcap further or flatten the curve earlier.