Method · disclosure · rubric
How the bench works
01Who publishes this
the.com.im is published by LABS, the studio that builds labs.llc. The bench tests our own instruments — how the scorecard accounts for that.
It means we are not independent, and we will not pretend to be. The people who publish THE BENCH also build every tool on its docket. So we ask you to trust the evidence rather than the judge: each mark on a scorecard is a yes-or-no check you can follow back to a line of the tool's source or to a measurement with its time in UTC.
Three rules follow from that. Every review names at least one real shortcoming of the labs tool, found in the code or in our test. Every rival is a real, current tool on the open web, checked on its own page on 27 September 2026, and every review says what that rival does better. And no tool pays for its place: they are ours, drawn at random from the labs.llc catalogue, ten to this site.
The bench keeps one instrument on itself: Google Analytics 4 counts visits to the.com.im, in a property of its own, and sets Google’s first-party _ga cookies to tell a return visit from a new one.
02The rubric
Five criteria, four checks each. A check is met only when the evidence we hold shows it; anything unseen scores nothing, and the review's scorecard note says which checks were missed and why.
Sources · Where the figures come from
- S1Every data source is named, on the page or in its source code.
- S2The figures come from the originating source (the agency, catalogue or registry that produces them), or the tool computes them itself and needs none.
- S3No key, account or sign-in is needed to use it.
- S4There is no server relay, or the relay is locked to a named allow-list.
Candour · What it admits
- C1The method is written down, on the page or at the head of its source.
- C2A limit, an accuracy figure or a freshness date is stated to the reader.
- C3Estimated, doubtful or missing figures are marked as such rather than filled in.
- C4We found no contradiction between what labs.llc says about the tool (the page or its catalogue line) and what the code does.
Proof · What our test showed
- P1The page answered HTTP 200 on the local copy of build 537.
- P2We reproduced a core computation or data call ourselves.
- P3We cross-checked a result against an independent reference.
- P4Everything we measured agreed with what the tool reports.
Footprint · Where the work happens
- F1The core work runs in the reader's browser.
- F2What the reader enters or opens is not sent to a labs.llc server.
- F3The tool's own code contacts no third-party service; its own origin only.
- F4By its own source, the tool needs no network once the page has loaded.
Reach · How far it goes against the field
- R1It does the core job its comparators do.
- R2It offers at least one thing none of the named comparators showed on the pages we fetched.
- R3It matches the broadest comparator's range: formats, items, modes or coverage.
- R4It keeps or exports what you made or found.
03Reading the bars
The scorecard is our own measured rubric. It is not a star rating, not a grade of how good or useful a tool is, and not an average of anyone's opinion. The total counts checks met; it does not rank the tools, which is why the docket lists them in the order the draw dealt them.
Read the five bars separately. Footprint rewards a tool that works inside your browser and asks nothing of anyone else. A live market desk has to fetch, so a low Footprint bar there describes where the work happens rather than a fault. Reach holds the one check no tool on the docket has met, R3, because each is narrower than the broadest rival we set it against.
Footprint judges the tool's own code. All ten tool pages on labs.llc also load the site-wide Google Analytics tag and a set of shared site scripts, whatever the tool does; F3 and F4 set that chrome aside, and a review counts it against Candour only where a page claims to make no requests at all.
04How we test
We test against a local copy of labs.llc build 537, byte-identical to the live build, served on our own machine, so a review never loads the public site to judge it. Pages and files are fetched with curl and the HTTP status, size and time recorded. Where a tool's arithmetic can run without a page, we run the tool's own code in JavaScriptCore and compare its answers with an outside reference where one exists: the US Naval Observatory for moon phases, the ECB and Xe for exchange rates, Python's zlib for a CRC-32.
Where the local copy lacked something the live site has, such as Laband's audio or the photographs behind Multipics, we say so in the review and score nothing for what we could not measure. Rivals are read on their own public pages; we record only what those pages state, on the day we read them, and make no claim either way about a feature we could not verify.
05What we will not do
No stars and no review markup for search engines: the reviewer and the reviewed share an owner, and a rating would claim an independence we do not have. No invented quotes, testimonials, user counts or ratings. No stock photographs or generated pictures; the only images here are captures of labs.llc tools from the local copy. No logos, screenshots or trade dress of rival tools, and no affiliate links to them.
Corrections are welcome at labs@labs.llc. When a finding here changes, the review changes with it, and the scorecard is recounted.