Podcast · Situs, episode 3 of 3

Twenty Retractions

What we believed about this client that was not true — and what that changes about build order.

Source brief written from the 44,000-word whiteboard transcript and the day’s threads, then fact-checked against the evidence. Internal only.

Twenty Retractions

Deep dive · two hosts · 45:34 · recorded 2026-09-02

What it covers

The mechanism before the list: a single-source statement gets promoted to a house rule, acquires an unfalsifiable verb like always or locked, and then sizes a build. That is how twenty claims survived long enough to be quoted back at us.

The sharpest ones. Pricing that was never discussed. A unit count wrong by roughly half. A language statistic with no field behind it. A vendor name that turned out to be an artifact of speech-to-text. A question that cost six days and had no consequence either way.

What survives: the gate written as things that must be refused rather than milestones to hit, and why phase one is after-hours leasing even though the buyer names renewals as her highest-ROI item.

And the regime that replaced the method: a provenance mark on every load-bearing claim, one home per fact, an owner and an expiry.

Before you quote anything from this

The underlying session transcript had no speaker labels — they were reconstructed by AI, and person names are garbled throughout. Nothing here attributes a quote to a named person, and neither should the episode. Numbers are traceable to the evidence base; where a figure is inferred rather than measured, the brief says so.

PropFlow Docs