Cluster
Shipping vs Experimenting vs None
Bassethound rolls the AI-stack signals into one verdict: shipping, experimenting, or none. This cluster covers where the lines are and how to read them.
5 questions about shipping vs experimenting vs none
- How can you tell if a company's AI is real or just marketing?
You tell real AI from marketing by fingerprinting the stack, not by reading the homepage. Model SDKs, vector-store clients, orchestration frameworks, and machine-facing endpoints leave signatures that a press release does not.
- How do you score AI adoption from the outside?
You score AI adoption from the outside by fingerprinting the AI stack a domain exposes and counting the distinct signal groups that fire, then normalizing that count into a score and a verdict.
- How do you tell if a company is shipping AI in production?
A company is shipping AI when its site exposes production AI infrastructure, not marketing: a live MCP endpoint, a technical llms.txt, a wired-in model provider, and a vector store or orchestration framework.
- What makes a company count as shipping AI?
A company counts as shipping AI when it exposes a machine-facing endpoint, or when at least three distinct AI signal groups fire across its stack. That is the shipping verdict: evidence in production, not a marketing claim.
- What's the difference between shipping and experimenting with AI?
Shipping means AI runs in production behind an exposed machine-facing endpoint or across three-plus signal groups. Experimenting means one or two signals fire with nothing exposed to machines.
Related clusters
Frequently asked questions
What does 'shipping AI' mean?
A company is shipping AI when its site exposes production AI infrastructure (a live MCP endpoint, a technical llms.txt, a wired-in model provider, a vector store or orchestration framework), not just marketing about AI.
How is the verdict scored?
Bassethound counts distinct AI signal groups and weights the ones that imply production. Shipping if a machine-facing endpoint is present or several groups fire, experimenting if at least one does, none if the stack is empty.
Sniff a domain.
Run sniff_domain on any site and read its five-layer dossier in one call.