A console for games that need a judge.
Zebra is three things: a small piece of hardware, a growing library of games, and a platform other designers can build on. Here’s how that works as a product, and the order to build it in.
What you’d buy.
See the packagingWhy a device, not an app.
Phones at the table end game night.
One object in the middle keeps everyone looking at each other instead of at their screens.
A neutral referee should be neutral hardware.
It isn't anyone's phone or anyone's account, so nobody's device is the one deciding.
Listening needs visible trust.
A lit ring, no cloud, a physical shutter. Everyone can see when it's on.
Identical hardware keeps rankings fair.
The same frozen model on the same chip is what lets a score in Ohio mean the same thing in Osaka.
It lives on the game shelf.
It's a gift, it sits next to the board games, and it sells where party games already sell.
It still starts as an app.
The fastest honest test is a phone in the middle of the table. Hardware comes after people accept the calls.
The platform.
Game designers write rules in a small scripting layer that calls the eleven judging primitives. That turns Zebra from a product into a place to publish games, the way a console is a place to publish video games.
- Rules modules ship on NFC game cards or as downloads.
- Designers test against the frozen judge before release, so what they tune is exactly what players get.
- The judge answers questions. The designer decides what the answers are worth.
- Revenue share with designers: [TERMS]
# Categories+ rules module (sketch)
on answer(player, text):
check = judge.counts(text, rule=f"{category}, starts with {letter}")
if check.borderline:
check = defend(player, seconds=10)
if not check.valid:
return call(0, check.reason)
bits = judge.rarity(text, given=prompt)
return call(clamp(round(bits), 1, 5), check.reason)
on challenge(player, ruling):
second = judge.rehear(ruling, argument=player.argument)
if second.stands:
player.score -= 1
else:
player.keep_flag()
ruling.overturn(second)How it spreads.
See the share cardsCalls are shareable
Any ruling can become a card: “Zebra ruled a hot dog is not a sandwich. The crowd put it at 38.”
The Daily
A reason to open the box every day and compare scores with friends who have one.
Venue leagues
Puts Zebra in front of strangers every week, at trivia nights and bars.
Streams
A judge on stream is content by itself. Chat can play along with the Daily.
Zebra-ready editions
Classic party games ship with a rulebook cartridge in the box.
The order to build it in.
- Phase 0, [WEEKS]Phone prototype
A phone in the middle of the table running Categories+, Outlier and Survey Says against a hosted model. One question: do people accept the calls?
- Phase 1, [MONTHS]Dev kit
Off-the-shelf mic array and compute, the model running on the device, eight games, judge edition 1 frozen. Playtests in [N] households.
- Phase 2, [DATE]Launch
The Starter set, two game packs, and the Daily.
- Phase 3, [DATE]Platform
The designer toolkit, rulebook cartridges with publishers, and venue leagues.
What to measure in playtests.
Targets for each get set after Phase 0, once there’s a baseline to compare against.
Share of players who agree the calls were fair after a game. The number everything else serves.
Challenges per 20 calls. Near zero can mean people gave up on arguing; high means the judge is off.
Share of challenges that win. The most honest measure of judge error.
From the end of an answer to the ruling. Anything that feels like waiting breaks the rhythm.
Households that play again within a week.
Calls turned into cards and sent somewhere. The growth loop.
Risks.
| Risk | Response |
|---|---|
| The judge feels arbitrary. | Reasons on every call, the challenge system, and playtesting until the felt-fair rate clears the bar. |
| Small on-device models get things wrong. | Keep facts in curated data. Ask the model only fuzzy questions, and let code do the scoring. |
| Hardware is expensive and slow to build. | Prove the games on a phone first, then build the dev kit from off-the-shelf parts. |
| A big publisher builds its own. | Move first on the platform and the cross-table leaderboard, and license rather than compete. |
| Kids plus an always-listening device. | No cloud, no recording, a strict kid mode, and a privacy review before anything ships. |
| The name. | “Zebra” is a working name. Zebra Technologies already holds the mark in electronics, so a trademark search comes before anything public. |
Still open.
- Voice only, or voice plus a companion screen for writing games like Ghostwriter?
- One judge personality, or swappable voices: stern umpire, game-show host, grandma?
- How much content should the judge generate, and how much should ship on printed cards?
- What does the Starter cost once on-device compute is priced in? [BOM TARGET]
- Which three games prove the concept fastest in Phase 0?
- License the classics first, or lead with originals and license later?