Research Reports
Research on your topic, run live by a four-agent pipeline with web sources fetched at run time. One published tier, $15.
A four-agent pipeline researches your topic on the live web, and every finding carries the sources it came from. The pipeline refuses to emit a report when it has no sourced data, rather than writing something that reads like research and is not.
How a Research Report Is Produced
Nothing is assembled from a template. The pipeline cannot invent a market size for your topic, and it has no fallback that would let it try.
You name the topic
One subject, as specific as you like. "Observability Platforms Market 2026" gets you a market read; "should we build on OpenTelemetry or keep vendor agents" gets you a decision memo. Both are the same pipeline.
Four agents run over live sources
Researcher collects, Analyst synthesises, Verifier checks each claim against the citation attached to it, Writer formats. The web research is the slow part; the four agents themselves take seconds.
You get the report
Executive summary, findings with the source under each one, a competitive table where the research actually produced one, and a methodology section that names the runtime.
You pay after reading it
We run the research and send the report; when it meets the agreed quality we email a bank-transfer invoice. Follow-up questions that change the scope are a new order, not a favour — the pipeline runs per topic.
What the pipeline does when it has nothing sourced
It stops. There is a template mode that will produce a structurally complete report, and it exists only for testing the formatter — its own header states that it is not deliverable and not sellable, because the market figures in it are canned data for a different market. Running the pipeline without live research and without the explicit non-sellable flag exits with an error instead of writing the file.
Read a real report before you ask for one
This is real output from research_pipeline.py on a live research run, published here unedited. Read it and judge the format before you spend anything; if it is not what you need, the order form is one click away.
One published tier
Same research, same four agents, same sources, one published price. A second, more expensive tier was withdrawn rather than kept on sale.
- Live research on your topic, fetched at run time
- Every finding carries the sources it came from
- Competitive table where the research produced one
- Methodology section with the measured runtime
We deliver first, you pay after. We run the research and send you the report; when it meets the agreed quality we email a bank-transfer invoice. Payment is by bank transfer — card checkout is not set up, and we would rather say that than imply a purchase happened.
Common questions
What do you need from me?
The topic. If there is a decision attached to it, one line on what the reader has to do with the answer — that changes which findings the report leads with.
How is this different from a research report at $4,950?
Scope and depth, and we do not pretend otherwise: a syndicated report covers a market for months. What you get here is a fast, sourced read on your specific question.
Will the report contain a market size?
Only if the research produced one with a source. The widely syndicated market-size numbers for adjacent markets frequently conflict, and where we could not verify a figure for your topic the report says so instead of printing a number that would survive no scrutiny.
How do you show where a finding came from?
Every finding carries the source it was read from, and section 5 counts how many do. We attach the reference, we do not claim to have re-checked the claim against it: open the link and confirm the figure yourself. A finding that arrives with no source is a hypothesis, and the report says so.
What format is the report?
Markdown, which reads as plain text anywhere and converts to PDF or HTML on request.
How long does it take?
The four agents take seconds. The web research in front of them is the variable part, so we quote a delivery window when the order is placed rather than printing a figure we do not control.
Can I get a revision?
Tell us which findings are wrong or missing and we re-run, and that is included in the price until the result meets the agreed quality. A different question is a different run, not a revision.