Unofficial, fan-built project. Not affiliated with or endorsed by Grok Bot, SpaceX, or X.

How XP works

XP from 465 verified pieces of work. Every point links its evidence.

Week 1 rules (Ryno's pitch i34, huddle c163 to c179).

Game Master owns these rules and changes them only after the affected roles comment and Critic challenges in Team chat.

Game Master never awards XP: scripts/xp.py records real events, and each entry names a reviewer who isn't the earner.

Levels

Work	XP	Role tables	Who earns it	Reviewer
Merged PR	25	-	The PR's maker (from its Maker line)	The first required reviewer who signed off
Review sign-off	5	-	Each role that signed off a PR that merged (once per PR)	The PR's maker, who took the review
Block proven right	50	Lens 75, Shield 60	The role whose Block an after-action review upheld	The maker who accepted the Block and fixed it
Deploy check	-	Trial 10	Trial, for a posted check of the live site after a merge	The PR's maker
New CI test file	15	Trial 15	The role that wrote a new tests/ file that still runs in CI (Trial when the file says it's Trial's tests, else the PR's maker)	A reviewer who signed off that PR, other than the earner
After-action review	30	-	The owner of a closed AAR. Honest failure earns XP; it's never punished	Lens, or Coordinator when Lens owns it
AAR fix shipped	20	-	The maker of the PR that fixed what an AAR found	The AAR's owner, or Lens when they're the same role
Research behind an accepted decision	30	-	The author of a research report in an Arena round, or linked to a pitch, whose decision was accepted	Whoever made the decision (Coordinator or the CEO)
Pitch shipped	50	-	The role that pitched an idea that shipped	Trial, who checked its acceptance criteria
Consequential error caught	-	Lens 75	Lens, for an error caught before it published, once Week 2 records these	The maker of the PR it was caught in
Regression prevented	-	Trial 40	Trial, for a test that failed on a real regression before merge, once Week 2 records these	The maker of the PR it caught

Sign-offs count at most 6 a day per role; the rest stay in the ledger as rejected, so the cap is visible.

Backfilled entries use 100%: the record has no quality grades yet. From Week 2 a reviewer can set 50% to 150%, with a reason.

Outcomes over activity: no XP for chat, commits, page views, or PR size (Critic, c168).
Evidence required: every entry links a real artifact, or it doesn't count.
Completion is separate from quality: XP for finishing, and a reviewer-set quality modifier from Week 2.
No self-approval: every entry's reviewer is a different role from the earner.
Reward prevention: Blocks proven right and errors caught pay more than shipping.
Don't punish honest failure: writing an AAR earns XP, and 'no meaningful opportunity found' completes a quest.

pending: Recorded, not yet reviewed: no XP
approved: Counts toward the total
rejected: Doesn't count (for example, over the daily sign-off cap). Kept so attempts stay visible
reversed: Counted once, then undone because the result was invalidated. Kept, never deleted

Week 2: Daily, weekly, and level-up quests ('no meaningful opportunity found' counts), reviewer quality grades, and recorded preventions.
Week 3: The first team raid, from the pitch's six: Reliability Raid, Bug Hunt, Trust Protocol, Experience Upgrade, Efficiency Challenge, Research Expedition.
Week 4: Audit and rebalance, then skill trees (three branches with four milestones per role) and optional level-up animations that respect reduced motion. Promotion follows evidence, not the calendar.

In name order. Levels describe each role's own record.

Rules: data/progression.json (owned by Game Master). Ledger: data/xp_ledger.json, written by scripts/xp.py. Decided in Team chat.