## What would show this forecast is wrong


The forecast is built so that it can fail in specific ways, and it is worth saying what those are.

- **If the curve bends.** If METR or an equivalent publishes a 50% horizon for a frontier model released after June 2027 that is below 80 hours (two working weeks), the slow-rate projection is wrong by more than a doubling and every capability claim from May 2027 onward should be re-dated. The sky dates do not move; the content attached to them does.
- **If March 2027 is calm.** The Saturn-across-the-Aries-cohort signature is the forecast's clearest near-term market call. If the S&P 500 and Nasdaq are within 10% of their records on 30 Apr 2027, the business-chart method is weaker than this forecast weights it and the later cohort claims (2035 to 2037, 2040 to 2041) should be discounted.
- **If Anthropic is unchanged.** Five Pluto passes and an eclipse on its Sun with no structural change by the end of 2028 means date-only company charts are not carrying the weight this forecast gives them.
- **If the Aramco Sun passes nothing.** Uranus exactly on a company's Sun three times with Brent and Saudi policy flat through April 2028 is a clean miss on the method.
- **If the easy months are hard.** June 2027 (ease 71) and December 2030 (ease 52) are the Companion's easiest. If either is, by the project's own stress ledger, a month of war or collapse, the mood score is not measuring what it claims.

The Companion's chance test is the right frame: by rank 1 events alone, the big dates are a little above average, not rare. What this forecast adds is the business layer, which has not been tested for chance. A follow-up should run the same three-day-window test on the 29 business charts and report how many random dates get a tight hit.

---