A Tuesday with an adoption number that deserves a careful read and a few tools worth keeping on the radar.

When someone asks me how much companies already use AI, I notice I almost always answer off the top of my head, without a source.

Today I read an analysis by Simon Willison of the US Census Bureau’s biweekly survey, which covers about 200,000 businesses. From what he shows, among companies with 250 or more employees, the “yes” for AI use was close to 30% in February 2026 and close to 40% in October.

But there is a detail that interests me more than the number. In November 2025 the question changed from “producing goods or services” to “any business function”. And, since the October 19 collection, the questionnaire has used the term “Super Intelligence” in place of “Artificial Intelligence”. In other words, the same curve carries question changes along the way.

On the product side, this is a familiar lesson. Before presenting a trend in the roadmap or in the pitch, it’s worth asking how the data was collected and whether the definition held. A series that looks clear may be comparing different things. For a starting point on the topic, the guide on AI for Product Managers and the overview of the state of AI in product management in 2026 help put the numbers in context.

I remain optimistic about the speed of adoption. I just think product decisions get better when we understand the yardstick used to measure.

For those who are curious and want to see the full analysis, here is the link.

The rest of the radar

  • Momentic launches Mo, an AI QA engineer — end-to-end tests running on every pull request take QA off the critical delivery path; the benchmark cited is from the company itself. Momentic.
  • Article asks the SEC to block Anthropic’s IPO — opinion, but the spending commitments cited help think about the risk of depending on a single model vendor. This belongs in your product’s AI governance. Tech Policy Press.
  • Claude Code skill for product demo videos — demos built by the team itself from a brief, without depending on video editing. GitHub.
  • LongCat-2.5, focused on multimodality for agents — another model competing in the agent space; the details were not verified, so check the announcement before citing it. LongCat.

That’s it for today. More tomorrow.