WEEKLY
Re-tested on a standing cycle
Not when you remember to look. Every week, whether or not anything happened.
Investing · $50/mo
Theses that check themselves.
You bought a company for reasons. Eighteen months on you will not have re-read them — there is no time in the week for it, and nothing sends a reminder when a thesis quietly becomes proven falsified. Sentry does, every week, against evidence.
A thesis here is not a paragraph.
It is a short list of conditions that could be proved wrong — each with a source, each with a date, each ratified by you before it counts. Vague conviction cannot be monitored. Falsifiable conditions can.
From then on the desk re-tests every condition weekly and reports per condition, not per company. You find out which leg of the argument moved, not merely that the mood changed.
A thesis nobody has re-tested lately is shown as stale — which is a different thing from healthy.
AVGO · THESISRE-TESTED WEEKLY · LAST 11 AUG
Bought for three reasons. One of them stopped being true on 7 August.
AI revenue compounds above 40% year on year.
HOLDS · 11 AUG · Q3 FILING
Custom-silicon share with two hyperscalers is defensible.
HOLDS · 11 AUG · ANALYST DAY
Software margin recovers to 70% by FY27.
BROKE · 7 AUG · GUIDANCE CUT
You were told the day it happened, not at the next earnings call.
A trim to half weight is staged.
Nothing is submitted without you.
PER CONDITION
You learn which leg moved
A verdict and a date attach to each condition, with the source that produced it.
STALE ≠ HEALTHY
Unverified is shown as unverified
An un-rechecked thesis is never quietly reported as fine. Silence has to be earned.
YOU RATIFY
Nothing becomes a thesis on its own
The desk drafts the conditions. You accept them, or rewrite them, before anything is monitored.