Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2511.16958 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Table of Contents:
- We model the cadence of AI product releases, i.e. quiet spells, reversible patches, and rarer pivots, as optimal exercise of strategic real options under reputational learning. A privately observed technical state follows a diffusion. The firm controls two upgrade options with asymmetric costs and reversibility (a cheap patch and a costly pivot) and a publication-frequency clock, a Cox process whose intensity governs when noisy public performance and safety signals are disclosed. For sufficiently low clock costs the optimal policy posts observable clock-off windows around knife-edge regions. These windows shut down the martingale part of public beliefs, eliminate knife-edge mixing, and collapse behavior to a two-rung release ladder with endogenous triggers, jump targets, and no interior mixing. Within stationary Markov strategies we show that this ladder is uniquely characterized by a boundary-value system with value matching and smooth pasting at triggers and target optimality at jump targets. We endogenize market or platform adoption as a threshold rule in public beliefs and show that leverage creates an irreversibility wedge: the gap between first-best and levered surplus is bounded by the takeover switching cost of the least reversible rung. Patches are debt-insensitive; pivots can be distorted, but only up to that bound. The framework predicts telemetry signatures in firm-authored disclosures: a pre-release cadence dip in publication intensity and intra-month dispersion as the clock is shut off before a major reset; two post-release plateaus in disclosed performance, consistent with patch versus pivot jump targets; and debt-insensitive patch timing in high-reversibility regimes, with leverage effects concentrated in pivots. Unlike option-implied volatility spikes, these patterns reflect the firm's own throttling of technical signals rather than market pricing of event risk.