Search demand is a gate, not a verdict
We use Keyword Planner and Trends to decide which choices deserve deeper research. Bucket values are estimates; blank rows stay unknown rather than being converted to zero.
METHOD / AUDIT
Testing methodology
A recommendation is only useful when you can see how it was made, what remains unknown, and the checked date for the underlying evidence.
Current scope
United States demand scan checked 2026-08-18. Product tests are still planned or researching; no final winners are published.
The task is locked before the tools run
Each comparison starts with one repository, fixed constraints, acceptance criteria, and a time boundary. Changing the brief for one tool invalidates the comparison.
Artifacts outrank impressions
We record prompts, files changed, build and test output, failures, corrections, elapsed work, and attributable costs. A polished demo is not enough.
Status is explicit
Planned means the test has not started. Researching means evidence is being collected. Tested means the documented acceptance checks ran against the recorded version.
Every recommendation has a checked date
AI products change quickly. Version drift, pricing changes, and new failure modes can move a page back to Researching until the test is repeated.
Commercial relationships never set rank
This MVP has no affiliate links. If monetization is added later, commercial relationships will be disclosed without changing the evidence threshold.
EVIDENCE / LABELS
Every material claim states what we actually know
Labels prevent a plausible observation or estimate from being presented as a verified fact.
Directly checked
Supported by reproducible output, a primary source, or an artifact checked against stated acceptance criteria.
Seen in practice
Observed during research or testing, but not yet reproduced enough to support a general recommendation.
Calculated with limits
Derived from a disclosed method or bounded source range. The assumptions stay visible beside the value.
Not reliably known
Unavailable, conflicting, or too weak to use. Unknown is never silently converted into zero or a confident claim.
CLAIM POLICY
No superlative without a completed comparison.
“Best,” “fastest,” and “most reliable” require a locked brief, comparable tools, completed acceptance checks, and visible evidence. If the product changes or the evidence expires, the page moves back to Researching until the relevant test is repeated.