Request intake
Anyone submits a brief — a baker, a band, a mate with a side hustle. No dollars, no obligation, no design-by-committee. Briefs are read on stream and slotted into a benchmark round.
Methodology · the whole pipeline✦The build is the benchmark — nothing runs off-stage
how it runs · four moves
A request becomes a website becomes an episode.
One shared brief per round — that's what makes the columns honest. People judge taste and truth on camera; everything else is a measured number. Head-to-head crowns, the composite ranks.
Built live on hardware
A nominated model + harness builds it inside a benchmark slot, on the channel. The clock is the scoreboard. Trace kept honest: stdout + stderr.
Viewport probe
Automated checks before any human verdict: zero horizontal overflow at 375 / 768 / 1440 px, zero unhandled console exceptions, and a link walk across every page.
Ratified & turnkey-deployed
Judges rate aesthetic + accuracy on camera; verdicts are ratified into the matrix; the build ships to freeweb.agency/name wearing its badge — all rendered from the matrix, never hand-typed.
When a better model takes the crown, sites get remastered on the next episode. Nothing rots — it just re-ranks.
the rubric · five weighted axes
Five weighted axes. Two pairs of human eyes.
Recorded, not scored: hardware, quant and harness are printed on the badge exactly as observed. No vibes, no vanity metrics.
interactive · score simulator
Play judge. Can you beat the incumbent?
Set the axes the way a judge would. The composite updates live against the displaced #1.
—
—