Looking back at our own AI Companies coverage from a year ago surfaces a few calls that did not hold up the way we expected, and we think naming them is more useful than letting them quietly disappear.
We underestimated how fast the consensus could shift
A view that seemed like a safe, widely shared assumption in AI Companies coverage a year ago has since been complicated by developments we did not see coming, which is a reminder that durability is not the same thing as correctness.
We overweighted one kind of source
We leaned more heavily on a particular type of source than the story probably warranted, and in hindsight a more balanced mix of perspectives would have caught the gap sooner.
Why we are saying this out loud
A correction buried in an update note does not do much to build trust. Naming what we got wrong in AI Companies coverage, and why, is a better test of whether our process is actually improving.