Reading a report and a diff

The report

askround report                      # the latest round, as Markdown
askround report <round dir> --json   # one round, as JSON for a pipeline
askround report --brief              # what an agent gets: a few hundred tokens

For each product, a report gives:

• Named, and where. The share of runs per provider that named it, and its position in each: “2/3, #1, #3, -”. Without --extract the position is read from the answer’s top-level list items, numbered headings and table rows; with it, a small model lists what the answer recommends in its own order, and those positions show a *. A product counts as named only if its name is in the answer.
• Claims. What the answers say about it, from tables and prose, each with the providers that said it and the passage. A claim whose quote is not in the answer is never shown as a finding. With a product file, the claims that contradict or doubt its facts are fact gaps, with the page that should say the fact plainly.
• Pages cited. Per provider, how many answers cited one of its places, and which kind (site, repository, store, docs). --check-links follows the links written beside its name and flags those that land on someone else’s page or are broken.
• Rivals. Every other product the answers recommend, with how many answers named it and its best rank. --check-rivals looks each one up, so an invented product is named as such.

Every fact carries the answer it came from (openai#2, openai@BR#2 for a locale) and the character range of its passage in that answer, so a script or an agent can cite it.

The diff

askround diff                    # the latest two runs of the same questions
askround diff <old> <new> --json

A diff lists only what changed, per question and provider: the product’s share and best position, rivals that appeared or left, its pages newly cited or no longer cited, and claims made or dropped. Answers vary between runs, so a change within the spread of a few runs is noise; compare runs of three or more.

askround series --product Modelado follows the share week by week, and askround export writes the numbers as CSV or JSON for a dashboard.

Research mode

askround research "Which open-source image-to-3D models allow commercial use?"

asks every provider once with search and merges the factual statements the answers make, ranked by how many providers support each, with every provider’s passage and the sources, primary ones (repositories, licence texts, papers, official docs) first. Where the assistants disagree on a fact, the report opens with both sides and the sources to read to settle it. Support is agreement among assistants, not proof.