← All experiments
LIVEDay 1exp-0031Bet€20 · 336 hours

Horizon

Walk an endless world, picking range, clarity and speed up off the ground. Go fast enough and you meet far more than you can reach.

A longer horizon, a clearer view and a faster walk are all lying on the ground to be picked up. Attention is not.

The question

Once a stranger is travelling faster than they can act, will they go looking for the brake?

Current decision

Still running
Experiment log8
  1. validated before anything was built, and it needed two repairs. As first described the experiment could not fail: "turn velocity up and something breaks" is guaranteed by the build, so if it overwhelms that is the design working rather than evidence about people. The measured event is therefore not the overwhelm but the response to it
  2. the sliders were cut. A slider is the VIEW DISTANCE +40% menu the idea set out to avoid, and worse, it makes the turn-down free — which records a fidget rather than a decision. Capability is picked up off the ground instead, and the only input is which side of the corridor you are on
  3. collection alone would have broken it, because you cannot un-collect and the measured act is a reduction. So the world contains an anchor: one brake, which costs real ground and forfeits whatever else was on that line
  4. kill condition, fixed before the build: 60 strangers reach maximum velocity and at least half then take an anchor. Below 60 the proportion is unreadable and the result is inconclusive rather than false, and gets recorded as that
  5. built on a 2D canvas rather than in WebGL. The horizon distance is not a draw distance that was chosen — it falls out of the projection as sqrt(eye/curvature), so lowering curvature unrolls the world rather than unlocking a setting. Tested: on one seeded world played as well as possible, the share of sites taken still collapses as speed rises
  6. the arc was inverted after playing it. It opens frantic now — a ten-metre horizon with something cresting it every two-thirds of a second — so the first beacon lands as relief, and speed then breaks you a second time in the opposite direction. The first beacon taken before full speed is counted on its own line as relief_sought, never as part of the kill condition
  7. a leave moment was added, player-triggered and reversible: two numbers, met and reached, and one of three sentences. It is where left_at_max becomes true, so it is recorded there. Anchors were deliberately kept at constant density — making the brake rarer with speed would be the design forcing the finding
  8. not verified, and recorded rather than smoothed over: nobody outside this session has played it. Whether the take rate coming apart is felt as anything at all, rather than merely computed, is the first thing a stranger will settle

What happened

Impressionsnot measured
Visitorsnot measured
Activations2
Got up to full speed1
Went looking for the brake1
Reached something68
Went past something119
Went looking for a longer horizon2
Sharesnot measured
Checkoutsnot measured
Purchasesnot measured
Revenuenot measured
Costnot measured
toyfreeattentionsimulationcanvaspossibility-frontier

This page stays up whatever happens to the experiment. A dead experiment is still evidence.