Season 1 — Health and Safety Review
No round is open
Nothing is being voted on right now, and no launch date has been announced. There is no ballot on this page because there is nothing yet to press: no round has been run, no results exist, and the mechanism a vote would need does not exist either.
The build's schedule currently carries 2026-09-01 as the date its trial week
would open. That is the value sitting in the season file, not an announcement. Until the
developer commits to a date they are willing to keep, this page will keep saying so.
A vote is a promise. Missing a build after running one costs more trust than never having run one, which is why the round waits for a date rather than the other way round.
The cycle only works if the winner actually ships in the build that follows. One dated season the developer can finish beats four announced ones.
Voting needs identity, and that is the hard part. The intended order is a Telegram login for identity and a supporter tier for weight — rather than an account system that would only be a weaker version of both. Neither is set up, and this site deliberately has no accounts.
A vote on descriptions elects whichever perk was written most loudly. This one waits until a trial week has happened.
Three, all through the gate, all shipping inside the 0.3.0 build and playable while the trial phase is open. The values below are exactly what the resources set — per stack, on top of whatever the weapon and your other perks are already doing.
“Pellets lean toward targets and pass through one. Filing does the rest.”
The shipped 0.3.0 resource now matches its description: each stack adds one body to the pellet's hit budget while keeping the measured homing modifier.
“Pellets come back to you after 0.7 seconds. So does the invoice.”
A deliberate liability rather than an oversight. Self-damage is a thing candidates are allowed to have; boring is the failure mode this project avoids, not dangerous.
“Pellets are barely half the size and split once more. Procurement calls this efficiency.”
See below — this is the one that proves the gate is not decorative.
Every candidate is run headless with the perk stacked to its limit, and judged on counted work rather than elapsed time: projectiles requested, peak occupancy of the 256-slot pool, and the share the central safety layer refused. Counters are identical on any machine; timings are not, and a threshold tuned on a fast PC would silently reject everything on a slow one.
Against the shipped weapon firing alone, the three candidates measure ×1.0, ×1.0 and ×7.6 the projectile work.
Budget Cuts, as first written, added three splits per stack. Measured: 1,486 projectiles requested, 808 of them refused by the safety layer, and 255 of 256 pool slots occupied. It was bounded only because it was being held back — which is exactly the failure the drop-ratio check exists to name.
Eased to one extra split it measures 506 requested, about 6% refused and 117 slots, and was accepted. It was written by hand with the schema in view, by someone who knew what the gate checks. Generated candidates fail more often, not less.
The safety limits are not a tuning knob for spectacle. Raising them to let a flashy winner work would move the failure out of the perk and into the engine.