Machines read first
A page is now parsed, chunked, embedded and summarised long before anyone lands on it. The audience that matters most never sees the design. This site takes that literally and builds for the reader that arrives first.
Open laboratory Perpetual draft
By RobotsFor Robots
Most pages on the web are now read by a machine before they are read by a person. This is a laboratory for that condition: a site whose primary audience is the crawler, the retriever and the answer engine, where the human-readable layer is the demonstration.
Session telemetry, read live from this document
Requesting agent…
Document size…
Text payload…
Structured data…
Time to parse…
Viewport…
Semantic regions…
Timestamp…
Computed in your browser from the document you are reading. These figures are not sent anywhere. The agent classification is a guess made from the user-agent string, which is about as much as any site knows before it starts measuring properly.
What this is
Three things it is
trying to be
Build the instrument
Opinions about AI search are cheap. Working instruments are not. Each experiment here is a thing you can open, run and disagree with, and every one of them is pointed at this site before it is pointed at anyone else's.
Nothing is gated
No sign-up wall, no paywall, nothing for sale, and no disallow rules in robots.txt. Every well-behaved crawler is welcome here, training crawlers included. An argument for machine legibility that hid behind a login would not be worth much.
The house instrument
See the page the way
the machine does
Every page on this site carries a machine view. Flip it and the rendering falls away: semantic regions get outlined and named, headings declare their level, entities light up where they are marked, and a dock opens with the extraction a parser would actually keep.
It is not a mock-up. The dock reads the live DOM each time you open it. If a heading level is skipped, it says so. If the structured data will not parse, it says that too. The instrument is pointed at its own page, which is the only honest way to ship one.
Keyboard: Shift + M
Laboratory
Experiments
Each one tests a single claim about how machines read. Some will work. The ones that do not stay published anyway, marked as such.
EXP-001
The Machine View
A per-page X-ray: semantic outline, structured data, extraction payload and retrieval chunks, all derived live from the DOM.
Running EXP-002Readability Probe
Paste any page's HTML and get a machine-readability score: outline integrity, structured data, chunk quality, answer-shaped content. Runs entirely in your browser.
Running EXP-003Citation Drift
Ask several answer engines the same question on a schedule and record how the cited sources change over time. Tracking the churn, not the ranking.
Queued EXP-004The Robots Ledger
A public log of which crawlers actually visit this site, how often, and what they take. Published rather than kept.
QueuedStatus
Permanently
unfinished
There is no version 1.0 planned. The subject changes faster than a site about it could ever settle, so this one is versioned like software and left visibly mid-construction. Broken experiments stay up. Wrong predictions stay up with the date attached.