Write-ups of the real systems we build: what broke, what we changed, and why. Published here in full; also on Substack for subscribers.
Why percentage-based retention charts mislead across videos of different lengths, and the absolute-seconds method that reveals the real problem.
Why a keyword filter can't tell a polite decline from real interest, the actual classifier logic that can, and an honest look at the one case it probably got wrong.
What happens when your AI avatar video comes back wrong: the real retry logic behind the daily pipeline, and the three questions worth asking about any AI system that matters.
No spam, no paid tier, unsubscribe whenever. This form posts straight to Substack. You never leave this page.