Answers
Published answers to questions about machine commerce, each tied to a dated research record.
Observability
Discovery and execution
- Do x402 endpoints change their price or payment address?
- How many listed routes can an agent actually construct a call for?
- How many x402 and MPP routes are listed?
- Why can a listed endpoint not be called?
- Do routes that leave the Bazaar catalogue stop quoting?
- Why did paid x402 calls answer HTTP 400 in September 2026?
- Do x402 sellers settle a payment before validating the request?
- Does paying an endpoint deliver what was bought?
Market structure
Buying decisions
- Why does a machine purchase still need human approval, when the machine can pay?
- Who decides what counts as evidence, and why does that choice have economic value?
- What does a seller rating leave out for a machine buyer?
- Why does evidence held inside a platform raise the cost of leaving it?
- Should machine commerce standardize the evidence, or the decision?
- What is a purchase outcome actually evidence of?
- Is evidence portability a question about data formats, or about competition?
- What evidence would let an owner stop approving each machine purchase?
- What would a record of machine purchases have to hold to be useful?
- How would you test whether evidence actually changes a purchasing decision?
- What is the difference between transaction infrastructure and decision infrastructure?
- Why is a low price or a strong track record not enough for a machine buyer?
- What work is left once a machine's payments are automated?
- Do fewer purchase approvals mean the evidence is working?
- Who needs confidence in a machine purchase, and about what?
- Would adding more metrics fix a rating for machine buyers?
- What can a machine buyer do with the evidence a score throws away?
- What is the machine-commerce equivalent of a credit rating?
- What does a history of observed outcomes show a buyer that a credential does not?
- Can two machine buyers read the same record and reach different conclusions?
- Is more evidence always worth it for a machine buyer?
- Why would a platform keep its purchase evidence private?
- Why can an established supplier look unknown to a machine buyer?
- What does a rival have to overcome to win a machine buyer from a platform?
- Does a buyer need to own its purchase evidence to switch platforms?
- Can competing measurement systems constrain whoever defines a measure?
- How would you test whether richer evidence lets an owner delegate more?
- How would you test whether portable evidence lowers the cost of switching?
- How would you test whether the definition of a measure changes a market?
- Does a machine buyer need an agreed verdict on whom to trust?
House rules
- Denominators
- A figure travels with the population it was counted over, in the issue's own words: a share and the denominator it was computed against are shown together.
- Dates
- Each answer names the record it reads from and that record's date. A later issue is a later answer at its own address, so two readings can be compared rather than replaced.
- Corrections
- An answer reads from the register, so a correction to an issue reaches its answers through the record. An issue that is withdrawn takes its answers with it.