# Thesis Board

> A machine for remembering what an idea said before reality had the chance to answer it. Monitor 001 follows "The 2028 Global Intelligence Crisis" by Citrini Research (James van Geelen) and Alap Shah (2026-02-22), which its authors call "What follows is a scenario, not a prediction.". Two analysts wrote a memo from June 2028 about an AI-driven crash. Every month we check which parts of it are coming true. The monitor never scores whether the authors were right; it records which links of the causal chain reality has picked up, one check-in per month, and earlier months are never rewritten.

Central rule: A predicted symptom appearing does not mean the predicted mechanism caused it.

Latest reading (SEPTEMBER 2026): Machines Happening. People Not yet. Money Other way.

## The fourteen links and their current status
- L01 (machines): AI capability → Agents that do real knowledge work. The models get good enough to actually do the job. — status ARRIVED
- L02 (machines): Capable agents → Firms automate knowledge work. Companies actually put them to work. — status ARRIVED
- L03 (people): Automation → Firms need fewer white-collar workers. The work gets done with fewer people. — status DIVERTED
- L04 (people): Fewer workers needed → Hiring weakens, layoffs rise. It shows up in the job numbers. — status DIVERTED
- L05 (people): Weak hiring → Labour income deteriorates. Households start earning less. — status NO GATE
- L06 (people): Falling labour income → Displaced workers cut spending. People stop spending. — status DELAYED
- L07 (people): Weaker spending → Consumer-facing businesses weaken. The shops and restaurants feel it. — status NO GATE
- L08 (machines): Margin pressure → Firms invest further in AI. So companies buy more AI to protect profits. — status ARRIVED
- L09 (money): Agents everywhere → Software and intermediary models under pressure. Per-seat software and middlemen lose their grip. — status DIVERTED
- L10 (money): Weak software economics → Private-credit defaults rise. The loans behind those companies go bad. — status ARRIVED
- L11 (money): Private-credit stress → Software credit is the source of it. And specifically, software is what breaks. — status NOT BOARDING
- L12 (money): Impaired white-collar income → Household credit and mortgages weaken. Well-paid people start missing payments. — status DELAYED
- L13 (money): Sector problems → Systemic financial stress. And the whole system catches it. — status NO GATE
- L14 (money): AI capex boom → Leverage disappoints and AI-linked credit reprices. The crash comes from the money spent building AI, not from AI wrecking software. [not in the memo] — status BOARDING

## Status vocabulary
- ARRIVED: Supporting evidence, and it is strong.
- BOARDING: Supporting evidence, but early or narrow.
- DELAYED: Expected by now. Not observed yet.
- DIVERTED: It happened, but not the way the memo wrote it.
- NOT BOARDING: Counterevidence this period.
- NO GATE: Nothing observable yet. Too early to call.

## Pages
- [The board](https://psyops.tech/): all links, scrub any month
- [All check-ins](https://psyops.tech/checkins/): one page per month, full evidence with sources
- [February 2026](https://psyops.tech/checkins/2026-02)
- [March 2026](https://psyops.tech/checkins/2026-03)
- [April 2026](https://psyops.tech/checkins/2026-04)
- [May 2026](https://psyops.tech/checkins/2026-05)
- [June 2026](https://psyops.tech/checkins/2026-06)
- [July 2026](https://psyops.tech/checkins/2026-07)
- [August 2026](https://psyops.tech/checkins/2026-08)
- [September 2026](https://psyops.tech/checkins/2026-09)
- [Chronography](https://psyops.tech/chronography): how our view of one belief changed month by month
- [RSS](https://psyops.tech/feed.xml)
- [Full text for language models](https://psyops.tech/llms-full.txt)

## Data (JSON, CC BY 4.0)
- https://psyops.tech/data/monitor.json — the thesis decomposed
- https://psyops.tech/data/checkins/YYYY-MM.json — one per month
- https://psyops.tech/data/chronography.json

## Provenance
From June 2026 the check-ins are built from the author's own monthly reports. February to May are a sourced reconstruction. Every evidence item carries publisher, date and URL where one exists; items without a recoverable public link say so.

Built by Deverboost (https://www.deverboost.com/). Contact hello@deverboost.com.